gnuplot 4.2
An Interactive Plotting Program
Thomas Williams & Colin Kelley
Version 4.2 organized by: Hans-Bernhard Bröker, Ethan A Merritt, and others
Major contributors (alphabetic order):
Hans-Bernhard Bröker
John Campbell
Robert Cunningham
David Denholm
Gershon Elber
Roger Fearick
Carsten Grammes
Lucas Hart
Lars Hecking
Thomas Koenig
David Kotz
Ed Kubaitis
Russell Lang
Timothée Lecomte
Alexander Lehmann
Alexander Mai
Ethan A Merritt
Petr Mikulík
Carsten Steger
Tom Tkacik
Jos Van der Woude
Alex Woo
James R. Van Zandt
Johannes Zellner
Copyright © 1986 - 1993, 1998, 2004 Thomas Williams, Colin Kelley
Copyright © 2004 - 2009 various authors
Mailing list for comments:
Mailing list for bug reports:
Web access (preferred):

This manual was originally prepared by Dick Crawford.
1 September 2009


I  Gnuplot
1 Copyright
2 Introduction
3 Seeking-assistance
4 New features introduced in version 4.2
 4.1 New plot styles
  4.1.1 Histogram
  4.1.2 Label plots
  4.1.3 Image data
  4.1.4 Filled curves
  4.1.5 Vectors
 4.2 Input from binary data files
 4.3 New plot elements
  4.3.1 RGB colors
  4.3.2 Arbitrary rectangles
 4.4 String handling
  4.4.1 String and text data read from datafiles
  4.4.2 User-defined string variables, operators, and functions
 4.5 Macros
 4.6 Auto-layout of multiple plots on a page
 4.7 Internal variables
 4.8 New or revised terminal drivers
  4.8.1 wxt
  4.8.2 emf
  4.8.3 gif, jpeg, png
  4.8.4 postscript
  4.8.5 ai
  4.8.6 epslatex, pslatex, pstex
  4.8.7 windows
 4.9 Canvas size
5 Backwards compatibility
6 Features introduced in version 4.0
 6.1 Mouse and hotkey support in interactive terminals
 6.2 New terminals
 6.3 New plot style pm3d
 6.4 Filled boxes
 6.5 New plot option smooth frequency
 6.6 Improved text options
 6.7 More text encodings
 6.8 Arrows
 6.9 Data file format
 6.10 New commands
 6.11 Other changes and additions
 6.12 Accompanying documentation
7 Batch/Interactive Operation
8 Command-line-editing
10 Coordinates
11 Datastrings
12 Environment
13 Expressions
 13.1 Functions
  13.1.1 Random number generator
 13.2 Operators
  13.2.1 Unary
  13.2.2 Binary
  13.2.3 Ternary
 13.3 Gnuplot-defined variables
 13.4 User-defined variables and functions
14 Glossary
15 Linetype, colors, and styles
 15.1 Colorspec
  15.1.1 Linecolor variable
16 Mouse input
 16.1 Bind
 16.2 Mouse variables
17 Plotting
18 Start-up
19 String constants and string variables
20 Substitution and Command line macros
 20.1 Substitution of system commands in backquotes
 20.2 Substitution of string variables as macros
 20.3 String variables, macros, and command line substitution
21 Syntax
 21.1 Quote Marks
22 Time/Date data
II  Commands
23 Cd
24 Call
25 Clear
26 Exit
27 Fit
 27.1 Adjustable parameters
 27.2 Short introduction
 27.3 Error estimates
  27.3.1 Statistical overview
  27.3.2 Practical guidelines
 27.4 Control
  27.4.1 Control variables
  27.4.2 Environment variables
 27.5 Multi-branch
 27.6 Starting values
 27.7 Tips
28 Help
29 History
30 If
31 Load
32 Lower
33 Pause
34 Plot
 34.1 Data
  34.1.1 Binary
  34.1.2 Binary general Array Record Format Endian Filetype Avs Edf Keywords Scan Transpose Dx, dy, dz Flipx, flipy, flipz Origin Center Rotate Perpendicular Binary examples
  34.1.3 Every
  34.1.4 Example datafile
  34.1.5 Index
  34.1.6 Smooth Acsplines Bezier Csplines Sbezier Unique Frequency
  34.1.7 Special-filenames
  34.1.8 Thru
  34.1.9 Using Using_examples Using title Xticlabels X2ticlabels Yticlabels Y2ticlabels Zticlabels
 34.2 Errorbars
 34.3 Errorlines
 34.4 Parametric
 34.5 Ranges
 34.6 Title
 34.7 With
35 Print
36 Pwd
37 Quit
38 Raise
39 Replot
40 Reread
41 Reset
42 Save
43 Set-show
 43.1 Angles
 43.2 Arrow
 43.3 Autoscale
  43.3.1 Parametric mode
  43.3.2 Polar mode
 43.4 Bars
 43.5 Bmargin
 43.6 Border
 43.7 Boxwidth
 43.8 Clabel
 43.9 Clip
 43.10 Cntrparam
 43.11 Color box
 43.12 Contour
 43.13 Data style
 43.14 Datafile
  43.14.1 Set datafile fortran
  43.14.2 Set datafile missing
  43.14.3 Set datafile separator
  43.14.4 Set datafile commentschars
  43.14.5 Set datafile binary
 43.15 Decimalsign
 43.16 Dgrid3d
 43.17 Dummy
 43.18 Encoding
 43.19 Fit
 43.20 Fontpath
 43.21 Format
  43.21.1 Gprintf
  43.21.2 Format specifiers
  43.21.3 Time/date specifiers
 43.22 Function style
 43.23 Functions
 43.24 Grid
 43.25 Hidden3d
 43.26 Historysize
 43.27 Isosamples
 43.28 Key
 43.29 Label
 43.30 Lmargin
 43.31 Loadpath
 43.32 Locale
 43.33 Logscale
 43.34 Macros
 43.35 Mapping
 43.36 Margin
 43.37 Mouse
  43.37.1 X11 mouse
 43.38 Multiplot
 43.39 Mx2tics
 43.40 Mxtics
 43.41 My2tics
 43.42 Mytics
 43.43 Mztics
 43.44 Offsets
 43.45 Origin
 43.46 Output
 43.47 Parametric
 43.48 Plot
 43.49 Pm3d
  43.49.1 Depthorder
 43.50 Palette
  43.50.1 Rgbformulae
  43.50.2 Defined
  43.50.3 Functions
  43.50.4 File
  43.50.5 Gamma correction
  43.50.6 Postscript
  43.50.7 Colornames
 43.51 Pointsize
 43.52 Polar
 43.53 Print
 43.54 Object
 43.55 Rmargin
 43.56 Rrange
 43.57 Samples
 43.58 Size
 43.59 Style
  43.59.1 Set style arrow
  43.59.2 Set style data
  43.59.3 Set style fill
  43.59.4 Set style function
  43.59.5 Set style increment
  43.59.6 Set style line
  43.59.7 Plotting styles
  43.59.8 Set style rectangle Boxerrorbars Boxes Boxxyerrorbars Candlesticks Dots Filledcurves Financebars Fsteps Histeps Histograms Newhistogram Image Impulses Labels Lines Linespoints Points Steps Rgbimage Vectors Xerrorbars Xyerrorbars Yerrorbars Xerrorlines Xyerrorlines Yerrorlines
 43.60 Surface
 43.61 Table
 43.62 Terminal
 43.63 Termoption
 43.64 Tics
 43.65 Ticslevel
 43.66 Ticscale
 43.67 Timestamp
 43.68 Timefmt
 43.69 Title
 43.70 Tmargin
 43.71 Trange
 43.72 Urange
 43.73 Variables
 43.74 Version
 43.75 View
 43.76 Vrange
 43.77 X2data
 43.78 X2dtics
 43.79 X2label
 43.80 X2mtics
 43.81 X2range
 43.82 X2tics
 43.83 X2zeroaxis
 43.84 Xdata
 43.85 Xdtics
 43.86 Xlabel
 43.87 Xmtics
 43.88 Xrange
 43.89 Xtics
  43.89.1 Xtics time_data
  43.89.2 Xtics rangelimited
 43.90 Xyplane
 43.91 Xzeroaxis
 43.92 Y2data
 43.93 Y2dtics
 43.94 Y2label
 43.95 Y2mtics
 43.96 Y2range
 43.97 Y2tics
 43.98 Y2zeroaxis
 43.99 Ydata
 43.100 Ydtics
 43.101 Ylabel
 43.102 Ymtics
 43.103 Yrange
 43.104 Ytics
 43.105 Yzeroaxis
 43.106 Zdata
 43.107 Zdtics
 43.108 Zzeroaxis
 43.109 Cbdata
 43.110 Cbdtics
 43.111 Zero
 43.112 Zeroaxis
 43.113 Zlabel
 43.114 Zmtics
 43.115 Zrange
 43.116 Ztics
 43.117 Cblabel
 43.118 Cbmtics
 43.119 Cbrange
 43.120 Cbtics
44 Shell
45 Splot
 45.1 Data-file
  45.1.1 Binary matrix
  45.1.2 Example datafile
  45.1.3 Matrix_ascii
  45.1.4 Matrix
 45.2 Grid data
 45.3 Splot overview
46 System
47 Test
48 Undefine
49 Unset
50 Update
III  Terminal types
51 Terminal
 51.1 Aed767
 51.2 Aifm
 51.3 Amiga
 51.4 Apollo
 51.5 Aqua
 51.6 Atari ST (via AES)
 51.7 Be
  51.7.1 Command-line_options
  51.7.2 Monochrome_options
  51.7.3 Color_resources
  51.7.4 Grayscale_resources
  51.7.5 Line_resources
 51.8 Cgi
 51.9 Cgm
  51.9.1 Cgm fonts
  51.9.2 Cgm fontsize
  51.9.3 Cgm linewidth
  51.9.4 Cgm rotate
  51.9.5 Cgm solid
  51.9.6 Cgm size
  51.9.7 Cgm width
  51.9.8 Cgm nofontlist
 51.10 Corel
 51.11 Debug
 51.12 Dospc
 51.13 Dumb
 51.14 Dxf
 51.15 Dxy800a
 51.16 Eepic
 51.17 Emf
 51.18 Emxvga
 51.19 Epslatex
 51.20 Epson-180dpi
 51.21 Excl
 51.22 Fig
 51.23 Ggi
 51.24 Gif
 51.25 Gnugraph(GNU plotutils)
 51.26 Gpic
 51.27 Gpr
 51.28 Grass
 51.29 Hercules
 51.30 Hp2623a
 51.31 Hp2648
 51.32 Hp500c
 51.33 Hpgl
 51.34 Hpljii
 51.35 Hppj
 51.36 Imagen
 51.37 Iris4d
 51.38 Jpeg
 51.39 Kyo
 51.40 Latex
 51.41 Linux
 51.42 Macintosh
 51.43 Mf
  51.43.1 METAFONT Instructions
 51.44 Mgr
 51.45 Mif
 51.46 Mp
  51.46.1 Metapost Instructions
 51.47 Mtos
 51.48 Next
 51.49 Openstep (next)
 51.50 Pbm
 51.51 Pdf
 51.52 Pm
 51.53 Png
 51.54 Postscript
  51.54.1 Enhanced postscript
  51.54.2 Editing postscript
  51.54.3 Postscript fontfile
  51.54.4 Postscript prologue
 51.55 Pslatex and pstex
 51.56 Pstricks
 51.57 Qms
 51.58 Regis
 51.59 Rgip
 51.60 Sun
 51.61 Svg
 51.62 Svga
 51.63 Tek40
 51.64 Tek410x
 51.65 Texdraw
 51.66 Tgif
 51.67 Tkcanvas
 51.68 Tpic
 51.69 Unixpc
 51.70 Unixplot
 51.71 Atari ST (via VDI)
 51.72 Vgagl
 51.73 VWS
 51.74 Vx384
 51.75 Windows
  51.75.1 Graph-menu
  51.75.2 Printing
  51.75.3 Text-menu
  51.75.4 Wgnuplot.ini
 51.76 Wxt
 51.77 X11
  51.77.1 X11_fonts
  51.77.2 Command-line_options
  51.77.3 Monochrome_options
  51.77.4 Color_resources
  51.77.5 Grayscale_resources
  51.77.6 Line_resources
  51.77.7 X11 pm3d_resources
  51.77.8 X11 other_resources
 51.78 Xlib
IV  Graphical User Interfaces
V  Bugs
52 Gnuplot limitations
53 External libraries
VI  Index

Part I

1 Copyright

   Copyright (C) 1986 - 1993, 1998, 2004, 2007  Thomas Williams, Colin Kelley

Permission to use, copy, and distribute this software and its documentation for any purpose with or without fee is hereby granted, provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation.

Permission to modify the software is granted, but not the right to distribute the complete modified source code. Modifications are to be distributed as patches to the released version. Permission to distribute binaries produced by compiling modified sources is granted, provided you

 1. distribute the corresponding source modifications from the  
  released version in the form of a patch file along with the binaries,  
 2. add special version identification to distinguish your version  
  in addition to the base release version number,  
 3. provide your name and address as the primary contact for the  
  support of your modified version, and  
 4. retain our contact information in regard to use of the base  

Permission to distribute the released version of the source code along with corresponding source modifications in the form of a patch file is granted with same provisions 2 through 4 for binary distributions.

This software is provided "as is" without express or implied warranty to the extent permitted by applicable law.


     Original Software:  
        Thomas Williams,  Colin Kelley.

     Gnuplot 2.0 additions:  
        Russell Lang, Dave Kotz, John Campbell.

     Gnuplot 3.0 additions:  
        Gershon Elber and many others.

     Gnuplot 4.0 additions:  
        See list of contributors at head of this document.

2 Introduction

gnuplot is a command-driven interactive function and data plotting program.

Any command-line arguments are assumed to be names of files containing gnuplot commands, with the exception of standard X11 arguments, which are processed first. Each file is loaded with the load command, in the order specified. gnuplot exits after the last file is processed. The special filename "-" is used to denote standard input. When no load files are named, gnuplot enters into an interactive mode. See help for batch/interactive (p. 37) for more details.

gnuplot is case sensitive (commands and function names written in lowercase are not the same as those written in CAPS). All command names may be abbreviated as long as the abbreviation is not ambiguous. Any number of commands may appear on a line (with the exception that load or call must be the final command), separated by semicolons (;). Strings are indicated with quotes. They may be either single or double quotation marks, e.g.,

     load "filename"  
     cd ’dir’

although there are some subtle differences (see syntax (p. 101) for more details).

Many gnuplot commands have multiple options. Version 4 is less sensitive to the order of these options than earlier versions, but some order-dependence remains. If you see error messages about unrecognized options, please try again using the exact order listed in the documentation.

Commands may extend over several input lines by ending each line but the last with a backslash (\). The backslash must be the last character on each line. The effect is as if the backslash and newline were not there. That is, no white space is implied, nor is a comment terminated. Therefore, commenting out a continued line comments out the entire command (see comments (p. 41)). But note that if an error occurs somewhere on a multi-line command, the parser may not be able to locate precisely where the error is and in that case will not necessarily point to the correct line.

In this document, curly braces ({}) denote optional arguments and a vertical bar (|) separates mutually exclusive choices. gnuplot keywords or help topics are indicated by backquotes or boldface (where available). Angle brackets (<>) are used to mark replaceable tokens. In many cases, a default value of the token will be taken for optional arguments if the token is omitted, but these cases are not always denoted with braces around the angle brackets.

For on-line help on any topic, type help followed by the name of the topic or just help or ? to get a menu of available topics.

The new gnuplot user should begin by reading about plotting (if on-line, type help plotting).

See the simple.dem demo, also available together with other demos on the web page

3 Seeking-assistance

There is a mailing list for gnuplot users. Note, however, that the newsgroup

is identical to the mailing list (they both carry the same set of messages). We prefer that you read the messages through the newsgroup rather than subscribing to the mailing list. Instructions for subscribing to gnuplot mailing lists may be found via the gnuplot development website on SourceForge

The address for mailing to list members is:

Bug reports and code contributions should be mailed to:

The list of those interested in beta-test versions is:

There is also the canonical (if occasionally out-of-date) gnuplot web page at

Before seeking help, please check the

FAQ (Frequently Asked Questions) list.

When posting a question, please include full details of the version of gnuplot, the machine, and operating system you are using. A small script demonstrating the problem may be useful. Function plots are preferable to datafile plots. If email-ing to gnuplot-info, please state whether or not you are subscribed to the list, so that users who use news will know to email a reply to you. There is a form for such postings on the WWW site.

4 New features introduced in version 4.2

Gnuplot version 4.2 offers many new features introduced since the preceding official version 4.0. This section lists major additions and gives a partial list of changes and minor new features. For a more exhaustive list, see the NEWS file.

4.1 New plot styles

4.1.1 Histogram

Histograms, or bar charts, can be produced. See histograms (p. 470).

4.1.2 Label plots

In coordination with the new datastrings feature described below, gnuplot can draw a label at each vertex of a curve. See labels (p. 482).

4.1.3 Image data

The image and rgbimage styles allow to plot 2D images (from ascii or binary files) and map them in a 2D or 3D plot. See image (p. 480) and rgbimage (p. 484).

4.1.4 Filled curves

The plot style fillstyle has been augmented to allow to fill the area between two input curves with a color or a pattern. See filledcurves (p. 464).

4.1.5 Vectors

Gnuplot can draw plots with vectors with a small arrowhead, requiring four or six columns of data for 2D or 3D, respectively. See vectors (p. 484).

4.2 Input from binary data files

Gnuplot can now read a generic binary input, including matrix binary and general binary (until now gnuplot supported only its own binary matrix format). Several matrix file formats are autodetected (gpbin, edf, avs). Binary data files are mainly useful for image and rgbimage drawings. See binary (p. 154) and binary general filetype (p. 158).

4.3 New plot elements

4.3.1 RGB colors

Explicit RGB colors can be specified for all plot elements instead of specifying a predefined linetype. See colorspec (p. 68).

4.3.2 Arbitrary rectangles

You can place rectangles with desired fill style and border anywhere in a 2D plot. See set object rectangle (p. 428).

4.4 String handling

4.4.1 String and text data read from datafiles

Gnuplot can now read and process text fields in datafiles. See datastrings (p. 43).

4.4.2 User-defined string variables, operators, and functions

String variables and string functions are introduced. Most gnuplot commands that previously required a string constant will now also accept a string variable, a string expression, or a function that returns a string. See string variables (p. 86).

4.5 Macros

Gnuplot supports command line macro expansion by ’@stringvariablename’. See macros (p. 94).

4.6 Auto-layout of multiple plots on a page

The multiplot mode is now able to layout automatically simple multiplots without having to set the size or the position for each plot. See multiplot (p. 378).

4.7 Internal variables

Gnuplot now exports several "read-only" variables such as GPVAL_TERM, GPVAL_X_MIN, etc. See gnuplot-defined variables (p. 56).

4.8 New or revised terminal drivers

4.8.1 wxt

The wxt terminal is an interactive and cross-platform terminal for on-screen rendering. It uses the wxWidgets library for its user interface, and Cairo associated with Pango for the actual rendering, providing nice plots with antialiasing on lines and text. The terminal supports the full range of gnuplot capabilities, including mousing, pm3d plots, image plots and enhanced text.

4.8.2 emf

The emf terminal generates an Enhanced Metafile Format file. This file format is the metafile standard on MS Win32 Systems. The emf terminal supports pm3d, rgb color, and image plot modes.

4.8.3 gif, jpeg, png

The code for the terminals using the gd library has been consolidated. The gif terminal also knows how to produce an animated gif from a sequence of plots.

4.8.4 postscript

The postscript terminal can load prologue files, which can contain additional user-defined sections with, for example, character encodings. See postscript prologue (p. 700).

4.8.5 ai

The Adobe Illustrator ai driver is outdated. Since Adobe Illustrator understands PostScript files, set terminal post level1 ... should be used instead.

4.8.6 epslatex, pslatex, pstex

The terminals supporting an output to latex augmented by PostScript commands have been consolidated. Many options are the same as in the postscript terminal.

4.8.7 windows

The windows terminal now supports the enhanced text mode.

4.9 Canvas size

In earlier versions of gnuplot, some terminal types used the values from set size to control also the size of the output canvas; others did not. The use of ’set size’ for this purpose was deprecated in version 4.2. In version 4.3 almost all terminals now behave as follows:

set term <terminal_type> size <XX>, <YY> controls the size of the output file, or "canvas". Please see individual terminal documentation for allowed values of the size parameters. By default, the plot will fill this canvas.

set size <XX>, <YY> scales the plot itself relative to the size of the canvas. Scale values less than 1 will cause the plot to not fill the entire canvas. Scale values larger than 1 will cause only a portion of the plot to fit on the canvas. Please be aware that setting scale values larger than 1 may cause problems on some terminal types.

The major exception to this convention is the PostScript driver, which by default continues to act as it has in earlier versions. Be warned that the next version of gnuplot may change the default behaviour of the PostScript driver as well.


     set size 0.5, 0.5  
     set term png size 600, 400  
     set output "figure.png"  
     plot "data" with lines

These commands will produce an output file "figure.png" that is 600 pixels wide and 400 pixels tall. The plot will fill the lower left quarter of this canvas. This is consistent with the way multiplot mode has always worked, however it is a change in the way the png driver worked for single plots in version 4.0.

5 Backwards compatibility

Gnuplot version 4.0 deprecated certain syntax used in earlier versions, but continued to recognize it. This is now under the control of a configuration option, and can be disabled as follows:

     ./configure --disable-backwards-compatibility

Notice: Deprecated syntax items may be disabled permanently in some future version of gnuplot.

One major difference is the introduction of keywords to disambiguate complex commands, particularly commands containing string variables. A notable issue was the use of bare numbers to specify offsets, line and point types. Illustrative examples:


     set title "Old" 0,-1  
     set data linespoints  
     plot 1 2 4               # horizontal line at y=1


     TITLE = "New"  
     set title TITLE offset char 0, char -1  
     set style data linespoints  
     plot 1 linetype 2 pointtype 4

Another compatibility issue is the effect of the command set size outside when not in multiplot mode. In earlier versions, the command set size <xx>, <yy> caused some terminals to change both the size of the plot and the size of the canvas is was drawn on; other terminatls changed only the plot size. The use of set size to change the canvas size is now deprecated.

Please see set size (p. 434), set term size (p. 30) and also the documentation for individual terminals.

6 Features introduced in version 4.0

Gnuplot version 4.0 contained many features introduced since the preceding official version 3.7. These are summarized here.

6.1 Mouse and hotkey support in interactive terminals

Interaction with the current plot via mouse and hotkeys is supported for the X11, OS/2 Presentation Manager, ggi, Windows, and wxWidgets terminals. See mouse input (p. 73) for more information on mousing. See help for bind (p. 73) for information on hotkeys. Also see the documentation for individual mousing terminals ggi (p. 614), pm (p. 684), windows (p. 731), wxt (p. 735) and x11 (p. 737).

Sample script: mousevariables.dem

6.2 New terminals

aqua: New terminal for Mac OS X. Requires AquaTerm 1.0 or later.

epslatex: New terminal. Prepares eps figures for inclusion in LaTeX documents.

gif: Consolidated with png/jpeg terminals. Requires libgd.

ggi: New full-screen interactive terminal for Linux. Interface to the General Graphics Interface Library.

pdf: New terminal exporting Adobe Portable Document Format. Requires libpdf.

png and jpeg: Support for GIF, PNG and JPEG image output is provided by a new driver via libgd. The new driver supports many more features than the old png driver, including TrueType fonts. Requires libgd.

svg: New terminal exporting Scalable Vector Graphics.

6.3 New plot style pm3d

The splot command is now capable of plotting 2D maps and 3D surfaces colored by greyscale or color palettes. See help for set pm3d (p. 390), set palette (p. 400), set cbrange (p. 536), set view map (p. 500), set colorbox (p. 288) and test palette (p. 560).

Sample scripts: pm3d.dem pm3dcolors.dem pm3dgamma.dem

6.4 Filled boxes

A solid color or patterned fill style can be set for any plot style that contains boxes. See boxes (p. 458), boxerrorbars (p. 458), boxxyerrorbars (p. 461), candlesticks (p. 461), set style fill (p. 445).

Sample scripts: fillstyle.dem candlesticks.dem

6.5 New plot option smooth frequency

Input data can be filtered through several built-in routines for interpolation or approximation of data. See smooth (p. 171), frequency (p. 174), unique (p. 174).

Sample scripts: steps.dem mgr.dem

6.6 Improved text options

Most gnuplot plot commands that produce text labels now accept modifiers to specify text color, font, size, and rotation angle. See set label (p. 351). Not all terminal types support these options, however. The enhanced text mode previously available for the postscript and pm terminals has been extended to other terminal types as well. Terminal types currently supported include aqua, dumb, jpeg, pdf, pm, png, postscript, x11, windows, and wxt. See enhanced text (p. 694).

Sample scripts: textcolor.dem textrotate.dem

6.7 More text encodings

Several terminals, including postscript, x11 and pm, support additional text encodings: ISO 8859-1 (Latin 1), ISO 8859-2 (Latin 2), ISO 8859-15 (variant of 8859-1 with Euro sign), KOI8-R and KOI8-U (cyrillic), and miscellaneous codepages. See encoding (p. 317) for more details.

6.8 Arrows

Single- or double-ended arrows can be placed on a plot individually from the command line or from a data file via the plot with vectors style. See set style arrow (p. 441), plotting styles vectors (p. 484).

Sample scripts: arrowstyle.dem vector.dem

6.9 Data file format

The new set datafile command can be used to specify information about the format of input data files, including the characters used to separate fields, to indicate comment lines, and to specify missing data. Gnuplot now attempts to recognize text fields with embedded blanks as single entities based on the datafile format settings. This allows input from csv (comma-separated value) files such as those exported by spreadsheet programs. See set datafile (p. 293). See also the binary (p. 154) option (introduced in 4.2).

6.10 New commands

set view map selects a top-view 2D projection of 3D surface plot.

set term push and set term pop save and restore the current terminal type.

load and save commands accept piped input and output, respectively.

6.11 Other changes and additions

Since gnuplot 4.0, unset <something> is preferred to set no<something>. The older form has been deprecated. Version 4.2 continues to allow the older syntax, but such backwards compatibility may be lost in future versions.

Commands of the form set <something> <style> also are deprecated in favor of the more general form set style <something> <options>. Many more plot elements now have style options of their own, including arrows, filled areas, lines, and points. There are also style settings for input data and formatting. Please see set style (p. 439), set decimalsign (p. 307), and set datafile (p. 293).

The MS Windows package includes an additional executable pgnuplot.exe to support piping command through standard input, which is otherwise not available for graphical applications on this system.

6.12 Accompanying documentation

In directory docs/psdocs/ you may find new information in the gnuplot output postscript file guide, list of postscript symbols in different encodings.

Improved FAQ. Please read it before asking your question in a public forum.

There are plenty of new demos *.dem in the demo/ directory. Please run them, for example by

     load "all.dem"

before asking for help. Plots produced by the demo scripts can also be viewed at

7 Batch/Interactive Operation

gnuplot may be executed in either batch or interactive modes, and the two may even be mixed together on many systems.

Any command-line arguments are assumed to be names of files containing gnuplot commands (with the exception of standard X11 arguments, which are processed first). Each file is loaded with the load command, in the order specified. gnuplot exits after the last file is processed. When no load files are named, gnuplot enters into an interactive mode. The special filename "-" is used to denote standard input.

Both the exit and quit commands terminate the current command file and load the next one, until all have been processed.


To launch an interactive session:


To launch a batch session using two command files "input1" and "input2":

     gnuplot input1 input2

To launch an interactive session after an initialization file "header" and followed by another command file "trailer":

     gnuplot header - trailer

8 Command-line-editing

Command-line editing is supported by the Unix, Atari, VMS, MS-DOS and OS/2 versions of gnuplot. Also, a history mechanism allows previous commands to be edited and re-executed. After the command line has been edited, a newline or carriage return will enter the entire line without regard to where the cursor is positioned.

(The readline function in gnuplot is not the same as the readline used in GNU Bash and GNU Emacs. If the GNU version is desired, it may be selected instead of the gnuplot version at compile time.)

The editing commands are as follows:

Command-line Editing Commands


Line Editing

^B move back a single character.
^F move forward a single character.
^A move to the beginning of the line.
^E move to the end of the line.
^H, DEL delete the previous character.
^D delete the current character.
^K delete from current position to the end of line.
^L, ^R redraw line in case it gets trashed.
^U delete the entire line.
^W delete from the current word to the end of line.


^P move back through history.
^N move forward through history.

On the IBM PC, the use of a TSR program such as DOSEDIT or CED may be desired for line editing. The default makefile assumes that this is the case; by default gnuplot will be compiled with no line-editing capability. If you want to use gnuplot’s line editing, set READLINE in the makefile and add readline.obj to the link file. The following arrow keys may be used on the IBM PC and Atari versions if readline is used:

Arrow keyFunction

Left same as ^B.
Right same as ^F.
Ctrl Left same as ^A.
Ctrl Rightsame as ^E.
Up same as ^P.
Down same as ^N.

The Atari version of readline defines some additional key aliases:

Key Function

Undo same as ^L.
Home same as ^A.
Ctrl Homesame as ^E.
Esc same as ^U.
Help help‘ plus return.
Ctrl Help help‘.


Comments are supported as follows: a # may appear in most places in a line and gnuplot will ignore the rest of the line. It will not have this effect inside quotes, inside numbers (including complex numbers), inside command substitutions, etc. In short, it works anywhere it makes sense to work.

See also set datafile commentschars (p. 300) for specifying comment characters in data files.

10 Coordinates

The commands set arrow, set key, set label and set object allow you to draw something at an arbitrary position on the graph. This position is specified by the syntax:

     {<system>} <x>, {<system>} <y> {,{<system>} <z>}

Each <system> can either be first, second, graph, screen, or character.

first places the x, y, or z coordinate in the system defined by the left and bottom axes; second places it in the system defined by the second axes (top and right); graph specifies the area within the axes — 0,0 is bottom left and 1,1 is top right (for splot, 0,0,0 is bottom left of plotting area; use negative z to get to the base — see set ticslevel (p. 527)); screen specifies the screen area (the entire area — not just the portion selected by set size), with 0,0 at bottom left and 1,1 at top right; and character gives the position in character widths and heights from the bottom left of the screen area (screen 0,0), character coordinates depend on the chosen font size.

If the coordinate system for x is not specified, first is used. If the system for y is not specified, the one used for x is adopted.

In some cases, the given coordinate is not an absolute position but a relative value (e.g., the second position in set arrow ... rto). In most cases, the given value serves as difference to the first position. If the given coordinate resides in a logarithmic axis the value is interpreted as factor. For example,

     set logscale x  
     set arrow 100,5 rto 10,2

plots an arrow from position 100,5 to position 1000,7 since the x axis is logarithmic while the y axis is linear.

If one (or more) axis is timeseries, the appropriate coordinate should be given as a quoted time string according to the timefmt format string. See set xdata (p. 503) and set timefmt (p. 495). gnuplot will also accept an integer expression, which will be interpreted as seconds from 1 January 2000.

11 Datastrings

The configuration option –enable-datastrings allows gnuplot to read and process text fields in datafiles. A text field consists of either an arbitrary string of printable characters containing no whitespace or an arbitrary string of characters, possibly including whitespace, delimited by double quotes. The following sample line from a datafile is interpreted to contain four columns, with a text field in column 3:

 1.000 2.000 "Third column is all of this text" 4.00

Text fields can be positioned within a 2-D or 3-D plot using the commands:

 plot ’datafile’ using 1:2:4 with labels  
 splot ’datafile using 1:2:3:4 with labels

A column of text data can also be used to label the ticmarks along one or more of the plot axes. The example below plots a line through a series of points with (X,Y) coordinates taken from columns 3 and 4 of the input datafile. However, rather than generating regularly spaced tics along the x axis labeled numerically, gnuplot will position a tic mark along the x axis at the X coordinate of each point and label the tic mark with text taken from column 1 of the input datafile.

 set xtics  
 plot ’datafile’ using 3:4:xticlabels(1) with linespoints

There is also an option that will interpret the first entry in a column of input data as a text field, and use it as the key title for data plotted from that column. The example given below will use the first entry in column 2 to generate a title in the key box, while processing the remainder of columns 2 and 4 to draw the required line:

 plot ’datafile’ using 1:(f($2)/$4) title 2 with lines

See set style labels (p. 482), using xticlabels (p. 192), plot title (p. 209), using (p. 185).

12 Environment

A number of shell environment variables are understood by gnuplot. None of these are required, but may be useful.

If GNUTERM is defined, it is used as the name of the terminal type to be used. This overrides any terminal type sensed by gnuplot on start-up, but is itself overridden by the .gnuplot (or equivalent) start-up file (see start-up (p. 85)) and, of course, by later explicit changes.

On Unix, AmigaOS, AtariTOS, MS-DOS and OS/2, GNUHELP may be defined to be the pathname of the HELP file (gnuplot.gih).

On VMS, the logical name GNUPLOT$HELP should be defined as the name of the help library for gnuplot. The gnuplot help can be put inside any system help library, allowing access to help from both within and outside gnuplot if desired.

On Unix, HOME is used as the name of a directory to search for a .gnuplot file if none is found in the current directory. On AmigaOS, AtariTOS, MS-DOS, Windows and OS/2, GNUPLOT is used. On Windows, the NT-specific variable USERPROFILE is tried, too. VMS, SYS$LOGIN: is used. Type help start-up.

On Unix, PAGER is used as an output filter for help messages.

On Unix, AtariTOS and AmigaOS, SHELL is used for the shell command. On MS-DOS and OS/2, COMSPEC is used for the shell command.

On MS-DOS, if the BGI or Watcom interface is used, PCTRM is used to tell the maximum resolution supported by your monitor by setting it to S<max. horizontal resolution>. E.g. if your monitor’s maximum resolution is 800x600, then use:

     set PCTRM=S800

If PCTRM is not set, standard VGA is used.

FIT_SCRIPT may be used to specify a gnuplot command to be executed when a fit is interrupted — see fit (p. 118). FIT_LOG specifies the default filename of the logfile maintained by fit.

GNUPLOT_LIB may be used to define additional search directories for data and command files. The variable may contain a single directory name, or a list of directories separated by a platform-specific path separator, eg. ’:’ on Unix, or ’;’ on DOS/Windows/OS/2/Amiga platforms. The contents of GNUPLOT_LIB are appended to the loadpath variable, but not saved with the save and save set commands.

Several gnuplot terminal drivers access TrueType fonts via the gd library. For these drivers the font search path is controlled by the environmental variable GDFONTPATH. Furthermore, a default font for these drivers may be set via the environmental variable GNUPLOT_DEFAULT_GDFONT.

The postscript terminal uses its own font search path. It is controlled by the environmental variable GNUPLOT_FONTPATH. The format is the same as for GNUPLOT_LIB. The contents of GNUPLOT_FONTPATH are appended to the fontpath variable, but not saved with the save and save set commands.

GNUPLOT_PS_DIR is used by the postscript driver to use external prologue files. Depending on the build process, gnuplot contains either a builtin copy of those files or simply a default hardcoded path. Use this variable to test the postscript terminal with custom prologue files. See postscript prologue (p. 700).

13 Expressions

In general, any mathematical expression accepted by C, FORTRAN, Pascal, or BASIC is valid. The precedence of these operators is determined by the specifications of the C programming language. White space (spaces and tabs) is ignored inside expressions.

Complex constants are expressed as {<real>,<imag>}, where <real> and <imag> must be numerical constants. For example, {3,2} represents 3 + 2i; {0,1} represents ’i’ itself. The curly braces are explicitly required here.

Note that gnuplot uses both "real" and "integer" arithmetic, like FORTRAN and C. Integers are entered as "1", "-10", etc; reals as "1.0", "-10.0", "1e1", 3.5e-1, etc. The most important difference between the two forms is in division: division of integers truncates: 5/2 = 2; division of reals does not: 5.0/2.0 = 2.5. In mixed expressions, integers are "promoted" to reals before evaluation: 5/2e0 = 2.5. The result of division of a negative integer by a positive one may vary among compilers. Try a test like "print -5/2" to determine if your system chooses -2 or -3 as the answer.

The integer expression "1/0" may be used to generate an "undefined" flag, which causes a point to ignored; the ternary operator gives an example. Or you can use the pre-defined variable NaN to achieve the same result.

The real and imaginary parts of complex expressions are always real, whatever the form in which they are entered: in {3,2} the "3" and "2" are reals, not integers.

Gnuplot can also perform simple operations on strings and string variables. For example, the expression ("A" . "B" eq "AB") evaluates as true, illustrating the string concatenation operator and the string equality operator.

A string which contains a numerical value is promoted to the corresponding integer or real value if used in a numerical expression. Thus ("3" + "4" == 7) and (6.78 == "6.78") both evaluate to true. An integer, but not a real or complex value, is promoted to a string if used in string concatenation. A typical case is the use of integers to construct file names or other strings; e.g. ("file" . 4 eq "file4") is true.

Substrings can be specified using a postfixed range descriptor [beg:end]. For example, "ABCDEF"[3:4] == "CD" and "ABCDEF"[4:*] == "DEF" The syntax "string"[beg:end] is exactly equivalent to calling the built-in string-valued function substr("string",beg,end), except that you cannot omit either beg or end from the function call.

13.1 Functions

The functions in gnuplot are the same as the corresponding functions in the Unix math library, except that all functions accept integer, real, and complex arguments, unless otherwise noted.

For those functions that accept or return angles that may be given in either degrees or radians (sin(x), cos(x), tan(x), asin(x), acos(x), atan(x), atan2(x) and arg(z)), the unit may be selected by set angles, which defaults to radians.

Math library functions

Function ArgumentsReturns

abs(x) any absolute value of x, |x|; same type
abs(x) complex length of x, ∘ ------2---------2
  real(x) + imag(x)
acos(x) any cos-1x (inverse cosine)
acosh(x) any cosh-1x (inverse hyperbolic cosine) in radians
arg(x) complex the phase of x
asin(x) any sin-1x (inverse sin)
asinh(x) any sinh-1x (inverse hyperbolic sin) in radians
atan(x) any tan-1x (inverse tangent)
atan2(y,x) int or real tan-1(y∕x) (inverse tangent)
atanh(x) any tanh-1x (inverse hyperbolic tangent) in radians
besj0(x) int or real j0 Bessel function of x, in radians
besj1(x) int or real j1 Bessel function of x, in radians
besy0(x) int or real y0 Bessel function of x, in radians
besy1(x) int or real y1 Bessel function of x, in radians
ceil(x) any x, smallest integer not less than x (real part)
cos(x) any cosx, cosine of x
cosh(x) any coshx, hyperbolic cosine of x in radians
erf(x) any erf(real(x)), error function of real(x)
erfc(x) any erfc(real(x)), 1.0 - error function of real(x)
exp(x) any ex, exponential function of x
floor(x) any x, largest integer not greater than x (real part)
gamma(x) any gamma(real(x)), gamma function of real(x)
ibeta(p,q,x) any ibeta(real(p,q,x)), ibeta function of real(p,q,x)
inverf(x) any inverse error function of real(x)
igamma(a,x) any igamma(real(a,x)), igamma function of real(a,x)
imag(x) complex imaginary part of x as a real number
invnorm(x) any inverse normal distribution function of real(x)
int(x) real integer part of x, truncated toward zero
lambertw(x) real Lambert W function
lgamma(x) any lgamma(real(x)), lgamma function of real(x)
log(x) any log ex, natural logarithm (base e) of x
log10(x) any log 10x, logarithm (base 10) of x
norm(x) any normal distribution (Gaussian) function of real(x)
rand(x) any rand(x), pseudo random number generator
real(x) any real part of x
sgn(x) any 1 if x > 0, -1 if x < 0, 0 if x = 0. imag(x) ignored
sin(x) any sinx, sine of x
sinh(x) any sinhx, hyperbolic sine of x in radians
sqrt(x) any √ --
  x, square root of x
tan(x) any tanx, tangent of x
tanh(x) any tanhx, hyperbolic tangent of x in radians

String functions

Function ArgumentsReturns

gprintf(”format”,x) any string result from applying gnuplot’s format parser
sprintf(”format”,x,...) multiple string result from C-language sprintf
strlen(”string”) string int length of string
strstrt(”string”,”key”) strings int index of first character of substring ”key”
substr(”string”,beg,end) multiple string ”string”[beg:end]
strftime(”timeformat”,t) any string result from applying gnuplot’s time parser
strptime(”timeformat”,s) string seconds since year 2000 as given in string s
system(”command”) string string containing output stream of shell command
word(”string”,n) string, int returns the nth word in ”string”
words(”string”) string returns the number of words in ”string”

other gnuplot functions

Function Arguments Returns

column(x) int column x during datafile manipulation.
defined(X) variable name [DEPRECATED] returns 1 if X is defined, 0 otherwise.
exists(”X”) ”variable name”returns 1 if a variable named X is defined, 0 otherwise.
stringcolumn(x) int content of column x as a string.
timecolumn(x) int timecolumn x during datafile manipulation.
tm_hour(x) int the hour
tm_mday(x) int the day of the month
tm_min(x) int the minute
tm_mon(x) int the month
tm_sec(x) int the second
tm_wday(x) int the day of the week
tm_yday(x) int the day of the year
tm_year(x) int the year
valid(x) int test validity of column(x) during datafile manip.

See also

airfoil.dem: use of functions and complex variables for airfoils demo.

13.1.1 Random number generator

The behavior of the built-in function rand(x) has changed as of version 3.8l. Older scripts that expected rand(x>0) to produce sequential pseudo-random numbers from the same seeded sequence must be changed to call rand(0) instead. The current behavior is as follows:

‘rand(0)‘  returns a pseudo random number in the interval [0:1] generated  
           from the current value of two internal 32-bit seeds.  
‘rand(-1)‘ resets both seeds to a standard value.  
‘rand(x)‘  for x>0 sets both seeds to a value based on the value of x.  
‘rand({x,y})‘ for x>0 sets seed1 to x and seed2 to y.

13.2 Operators

The operators in gnuplot are the same as the corresponding operators in the C programming language, except that all operators accept integer, real, and complex arguments, unless otherwise noted. The ** operator (exponentiation) is supported, as in FORTRAN.

Parentheses may be used to change order of evaluation.

13.2.1 Unary

The following is a list of all the unary operators and their usages:

Unary Operators


- -a unary minus
+ +a unary plus (no-operation)
~ ~a * one’s complement
! !a * logical negation
! a! * factorial
$ $3 * call arg/column during ‘using‘ manipulation

(*) Starred explanations indicate that the operator requires an integer argument.

Operator precedence is the same as in Fortran and C. As in those languages, parentheses may be used to change the order of operation. Thus -2**2 = -4, but (-2)**2 = 4.

The factorial operator returns a real number to allow a greater range.

13.2.2 Binary

The following is a list of all the binary operators and their usages:

Binary Operators


** a**b exponentiation
* a*b multiplication
/ a/b division
% a%b * modulo
+ a+b addition
- a-b subtraction
== a==b equality
!= a!=b inequality
< a<b less than
<= a<=b less than or equal to
> a>b greater than
>= a>=b greater than or equal to
& a&b * bitwise AND
^ a^b * bitwise exclusive OR
| a|b * bitwise inclusive OR
&& a&&b * logical AND
|| a||b * logical OR
. A.B string concatenation
eq A eq B string equality
ne A ne B string inequality

(*) Starred explanations indicate that the operator requires integer arguments. Capital letters A and B indicate that the operator requires string arguments.

Logical AND (&&) and OR (||) short-circuit the way they do in C. That is, the second && operand is not evaluated if the first is false; the second || operand is not evaluated if the first is true.

13.2.3 Ternary

There is a single ternary operator:

Ternary Operator


?: a?b:c ternary operation

The ternary operator behaves as it does in C. The first argument (a), which must be an integer, is evaluated. If it is true (non-zero), the second argument (b) is evaluated and returned; otherwise the third argument (c) is evaluated and returned.

The ternary operator is very useful both in constructing piecewise functions and in plotting points only when certain conditions are met.


Plot a function that is to equal sin(x) for 0 <= x < 1, 1/x for 1 <= x < 2, and undefined elsewhere:

     f(x) = 0<=x && x<1 ? sin(x) : 1<=x && x<2 ? 1/x : 1/0  
     plot f(x)

Note that gnuplot quietly ignores undefined values, so the final branch of the function (1/0) will produce no plottable points. Note also that f(x) will be plotted as a continuous function across the discontinuity if a line style is used. To plot it discontinuously, create separate functions for the two pieces. (Parametric functions are also useful for this purpose.)

For data in a file, plot the average of the data in columns 2 and 3 against the datum in column 1, but only if the datum in column 4 is non-negative:

     plot ’file’ using 1:( $4<0 ? 1/0 : ($2+$3)/2 )

Please see plot datafile using (p. 185) for an explanation of the using (p. 185) syntax.

13.3 Gnuplot-defined variables

The variable pi is defined to be pi, see

     print pi

Additionally, gnuplot may define some variables under various operations.

Working with interactive terminals with mouse functionality defines variables with names that begin "MOUSE_", see mouse variables (p. 83) for details.

Further, there are several "read-only" variables that begin "GPVAL_", like GPVAL_TERM, GPVAL_X_MIN, GPVAL_X_MAX, GPVAL_Y_MIN,... Type show variables all to display their list and values. Values related to axes parameters (ranges, log base) are values used during the last plot, not those currently set.

The fit mechanism uses several variables with names that begin "FIT_". It is safest to avoid using such names. "FIT_LIMIT", however, is one that you may wish to redefine. Under set fit errorvariables, the error for each fitted parameter will be stored in a variable named like the parameter, but with "_err" appended. See the documentation on fit (p. 118) for details.

See user-defined variables (p. 57), mouse variables (p. 83), and fit (p. 118).

13.4 User-defined variables and functions

New user-defined variables and functions of one through five variables may be declared and used anywhere, including on the plot command itself.

User-defined function syntax:

     <func-name>( <dummy1> {,<dummy2>} ... {,<dummy5>} ) = <expression>

where <expression> is defined in terms of <dummy1> through <dummy5>.

User-defined variable syntax:

     <variable-name> = <constant-expression>


     w = 2  
     q = floor(tan(pi/2 - 0.1))  
     f(x) = sin(w*x)  
     sinc(x) = sin(pi*x)/(pi*x)  
     delta(t) = (t == 0)  
     ramp(t) = (t > 0) ? t : 0  
     min(a,b) = (a < b) ? a : b  
     comb(n,k) = n!/(k!*(n-k)!)  
     len3d(x,y,z) = sqrt(x*x+y*y+z*z)  
     plot f(x) = sin(x*a), a = 0.2, f(x), a = 0.4, f(x)

     file = "mydata.inp"  
     file(n) = sprintf("run_%d.dat",n)

The final two examples illustrate a user-defined string variable and a user-defined string function.

Note that the variables pi (3.14159...) and NaN (IEEE "Not a Number") are already defined. You can redefine these to something else if you really need to. The original values can be recovered by setting:

     NaN = GPVAL_NaN  
     pi  = GPVAL_pi

Other variables may be defined under various gnuplot operations like mousing in interactive terminals or fitting; see gnuplot-defined variables (p. 56) for details.

You can check for existence of a given variable V by the exists("V") expression. For example

     a = 10  
     if (exists("a")) print "a is defined"  
     if (!exists("b")) print "b is not defined"

Valid names are the same as in most programming languages: they must begin with a letter, but subsequent characters may be letters, digits, "$", or "_".

See show functions (p. 329), functions (p. 49), gnuplot-defined variables (p. 56), macros (p. 94).

14 Glossary

Throughout this document an attempt has been made to maintain consistency of nomenclature. This cannot be wholly successful because as gnuplot has evolved over time, certain command and keyword names have been adopted that preclude such perfection. This section contains explanations of the way some of these terms are used.

A "page" or "screen" is the entire area addressable by gnuplot. On a monitor, it is the full screen; on a plotter, it is a single sheet of paper.

A screen may contain one or more "plots". A plot is defined by an abscissa and an ordinate, although these need not actually appear on it, as well as the margins and any text written therein.

A plot contains one "graph". A graph is defined by an abscissa and an ordinate, although these need not actually appear on it.

A graph may contain one or more "lines". A line is a single function or data set. "Line" is also a plotting style. The word will also be used in sense "a line of text". Presumably the context will remove any ambiguity.

The lines on a graph may have individual names. These may be listed together with a sample of the plotting style used to represent them in the "key", sometimes also called the "legend".

The word "title" occurs with multiple meanings in gnuplot. In this document, it will always be preceded by the adjective "plot", "line", or "key" to differentiate among them.

A 2-d graph may have up to four labelled axes. The names of the four axes for these usages are "x" for the axis along the bottom border of the plot, "y" for the left border, "x2" for the top border, and "y2" for the right border.

A 3-d graph may have up to three labelled axes – "x", "y" and "z". It is not possible to say where on the graph any particular axis will fall because you can change the direction from which the graph is seen with set view.

When discussing data files, the term "record" will be resurrected and used to denote a single line of text in the file, that is, the characters between newline or end-of-record characters. A "point" is the datum extracted from a single record. A "datablock" is a set of points from consecutive records, delimited by blank records. A line, when referred to in the context of a data file, is a subset of a datablock.

15 Linetype, colors, and styles

Each gnuplot terminal type provides a set of distinct "linetypes". These may differ in color, in thickness, in dot/dash pattern, or in some combination of color and dot/dash. The default linetypes for a particular terminal can be previewed by issuing the test command after setting the terminal type. The pre-defined colors and dot/dash patterns are not guaranteed to be consistent for all terminal types, but all terminals use the special linetype -1 to mean a solid line in the primary foreground color (normally black). By default, successive functions or datafiles plotted by a single command will be assigned successive linetypes. You can override this default by specifying a particular linetype for any function, datafile, or plot element.


    plot "foo", "bar"                 # plot two files using linetypes 1, 2  
    plot sin(x) linetype 4            # terminal-specific linetype color 4  
    plot sin(x) lt -1                 # black

For many terminal types it is also possible to assign user-defined colors using explicit rgb (red, green, blue) values, named colors, or color values that refer to the current PM3D palette.


    plot sin(x) lt rgb "violet"       # one of gnuplot’s named colors  
    plot sin(x) lt rgb "#FF00FF"      # explicit RGB triple in hexadecimal  
    plot sin(x) lt palette cb -45     # whatever color corresponds to -45  
                                      # in the current cbrange of the palette  
    plot sin(x) lt palette frac 0.3   # fractional value along the palette

See show palette colornames (p. 424), set palette (p. 400), cbrange (p. 536).

For terminals that support dot/dash patterns, each default linetype has both a dot-dash pattern and a default color. However, you can override the default color by using the keyword linecolor, abbreviated lc. For example, the postscript terminal provides a dashed blue line as linetype 3. The plot commands below use this same dash pattern for three plots, one in blue (the default), another in red (the default for linetype 1), and a third in gold.


    set term postscript dashed color  
    plot ’foo’ lt 3, ’baz’ lt 3 linecolor 1, ’bar’ lt 3 lc rgb ’gold’

Lines can have additional properties such as linewidth. You can associate these various properties, as well as equivalent properties for point symbols, into user-defined "line styles" using the command set style line. Once you have defined a linestyle, you can use it in a plot command to control the appearance of one or more plot elements.


    # define a new line style with terminal-independent color cyan,  
    # linewidth 3, and associated point type 6 (a circle with a dot in it).  
    set style line 5 lt rgb "cyan" lw 3 pt 6  
    plot sin(x) with linespoints ls 5          # user-defined line style 5

See linestyle (p. 450), set style line (p. 450).

15.1 Colorspec

Many commands allow you to specify a linetype with an explicit color. Note that not all terminals support RGB colors or pm3d palette colors.


     ... {linetype | lt} <colorspec>  
     ... {linecolor | lc} <colorspec>  
     ... {textcolor | tc} <colorspec>

where <colorspec> has one of the following forms:

     rgbcolor "colorname"  
     rgbcolor "#RRGGBB"  
     rgbcolor variable  
     palette frac <val>      # <val> runs from 0 to 1  
     palette cb <value>      # <val> lies within cbrange  
     palette z  
     variable                # color index is read from input file

"colorname" refers to one of the color names built in to gnuplot. For a list of the available names, see show palette colornames (p. 424).

"#RRGGBB" is a hexadecimal constant preceded by the "#" symbol. The RRGGBB represents the red, green, and blue components of the color, each on a scale from 0 - 255. For example, magenta = full-scale red + full-scale blue would be represented by #FF00FF, which is the hexadecimal representation of (255 << 16) + (0 << 8) + (255).

"rgb variable" requires an additional column in the using specifier, and is only available in 3D plotting mode (splot). The extra column is interpreted as a 24-bit packed RGB triple. These are most easily specified in a data file as hexadecimal values (see above).


     rgb(r,g,b) = 65536 * int(r) + 256 * int(g) + int(b)  
     splot "data" using 1:2:3:(rgb($1,$2,$3)) with points lc rgb variable

The color palette is a linear gradient of colors that smoothly maps a single numerical value onto a particular color. Two such mappings are always in effect. palette frac maps a fractional value between 0 and 1 onto the full range of the color palette. palette cb maps the range of the color axis onto the same palette. See set cbrange (p. 536). See also set colorbox (p. 288). You can use either of these to select a constant color from the current palette.

"palette z" maps the z value of each plot segment or plot element into the cbrange mapping of the palette. This allows smoothly-varying color along a 3D line or surface. It also allows coloring 2D plots by palette values read from an extra column of data.

15.1.1 Linecolor variable

Most plot commands assign a single color (linetype) to each element of the plot. If there are multiple plots on a single graph, the default color (linetype) is incremented sequentially. You can instead assign a separate color for each data point, line segment, or label based on additional information in the input data file. This is indicated by the colorspec keyword variable.

lc variable tells the program to use the value read from one column of the input data as a linestyle index, and use the color belonging to that linestyle. This requires a corresponding additional column in the using specifier. Text colors can be set similarly using tc variable.

A single data file may contain multiple sets of data, separated by two blank lines. Each of these separate sets is assigned an index value (see index (p. 168)) that can be retrieved via the using specifier column(-2). All data in the file is drawn with the same color/linestyle/pointtype properties by default. The command lc variable can be used to assign different colors to each data set in the file by using the index value from pseudocolumn -2.


     # Use the third column of data to assign colors to individual points  
     plot ’data’ using 1:2:3 with points lc variable

     # Use the data set index to choose a linestyle color  
     plot ’data’ using 1:2:(column(-2)) with lines lc variable

16 Mouse input

The x11, pm, windows, ggi, and wxt terminals allow interaction with the current plot using the mouse. They also support the definition of hotkeys to activate pre-defined functions by hitting a single key while the mouse focus is in the active plot window. It is even possible to combine mouse input with batch command scripts, by invoking the command pause mouse and then using the mouse variables returned by mouse clicking as parameters for subsequent scripted actions. See bind (p. 73) and mouse variables (p. 83). See also the command set mouse (p. 376).

16.1 Bind

The bind allows defining or redefining a hotkey, i.e. a sequence of gnuplot commands which will be executed when a certain key or key sequence is pressed while the driver’s window has the input focus. Note that bind is only available if gnuplot was compiled with mouse support and it is used by all mouse-capable terminals. Bindings overwrite the builtin bindings (like in every real editor), except <space> and ’q’ which cannot be rebound (unless one exception, see below). Mouse buttons cannot be rebound.

You get the list of all hotkeys by typing bind or by hitting ’h’ in the graph window.

Note that multikey-bindings with modifiers have to be quoted.

Normally hotkeys are only recognized when the currently active plot window has focus. bind allwindows <key> ... (short form: bind all <key> ...) causes the binding for <key> to apply to all gnuplot plot windows, active or not. In this case gnuplot variable MOUSE_KEY_WINDOW is set to the ID of the originating window, and may be used by the bound command.

By default, the <space> hotkey raises gnuplot’s command window. On some terminals (e.g. x11, wx), ’q’ closes the graph window. These defaults can be changed to ctrl-space and ctrl-q by starting gnuplot as ’gnuplot -ctrlq’, see x11 command-line-options (p. 747), or by the X Resource ’gnuplot*ctrlq’. Note: if <space> (or ctrl-space) does not raise the gnuplot window under X11, see discussion in raise (p. 231).


     bind {allwindows} [<key-sequence>] ["<gnuplot commands>"]  


- set bindings:

   bind a "replot"  
   bind "ctrl-a" "plot x*x"  
   bind "ctrl-alt-a" ’print "great"’  
   bind Home "set view 60,30; replot"  
   bind all Home ’print "This is window ",MOUSE_KEY_WINDOW’

- show bindings:

   bind "ctrl-a"          # shows the binding for ctrl-a  
   bind                   # shows all bindings

- remove bindings:

   bind "ctrl-alt-a" ""   # removes binding for ctrl-alt-a  
                            (note that builtins cannot be removed)  
   bind!                  # installs default (builtin) bindings

- bind a key to toggle something:

 bind "ctrl-r" "v=v+1;if(v%2)set term x11 noraise; else set term x11 raise"

Modifiers (ctrl / alt) are case insensitive, keys not:

   ctrl-alt-a == CtRl-alT-a  
   ctrl-alt-a != ctrl-alt-A

List of modifiers (alt == meta):

   ctrl, alt

List of supported special keys:

  "BackSpace", "Tab", "Linefeed", "Clear", "Return", "Pause", "Scroll_Lock",  
  "Sys_Req", "Escape", "Delete", "Home", "Left", "Up", "Right", "Down",  
  "PageUp", "PageDown", "End", "Begin",

  "KP_Space", "KP_Tab", "KP_Enter", "KP_F1", "KP_F2", "KP_F3", "KP_F4",  
  "KP_Home", "KP_Left", "KP_Up", "KP_Right", "KP_Down", "KP_PageUp",  
  "KP_PageDown", "KP_End", "KP_Begin", "KP_Insert", "KP_Delete", "KP_Equal",  
  "KP_Multiply", "KP_Add", "KP_Separator", "KP_Subtract", "KP_Decimal",  

  "KP_1" - "KP_9", "F1" - "F12"

See also help for mouse (p. 376) and if (p. 137).

16.2 Mouse variables

When mousing is active, clicking in the active window will set several user variables that can be accessed from the gnuplot command line. The coordinates of the mouse at the time of the click are stored in MOUSE_X MOUSE_Y MOUSE_X2 and MOUSE_Y2. The mouse button clicked, and any meta-keys active at that time, are stored in MOUSE_BUTTON MOUSE_SHIFT MOUSE_ALT and MOUSE_CTRL. These variables are set to undefined at the start of every plot, and only become defined in the event of a mouse click in the active plot window. To determine from a script if the mouse has been clicked in the active plot window, it is sufficient to test for any one of these variables being defined.

     plot ’something’  
     pause mouse  
     if (defined(MOUSE_BUTTON)) call ’something_else’; \  
     else print "No mouse click."

It is also possible to track keystrokes in the plot window using the mousing code.

     plot ’something’  
     pause mouse keypress  
     print "Keystroke ", MOUSE_KEY, " at ", MOUSE_X, " ", MOUSE_Y

When pause mouse keypress is terminated by a keypress, then MOUSE_KEY will contain the ascii character value of the key that was pressed. MOUSE_CHAR will contain the character itself as a string variable. If the pause command is terminated abnormally (e.g. by ctrl-C or by externally closing the plot window) then MOUSE_KEY will equal -1.

Note that after a zoom by mouse, you can read the new ranges as GPVAL_X_MIN, GPVAL_X_MAX, GPVAL_Y_MIN, and GPVAL_Y_MAX, see gnuplot-defined variables (p. 56).

17 Plotting

There are three gnuplot commands which actually create a plot: plot, splot and replot. plot generates 2-d plots, splot generates 3-d plots (actually 2-d projections, of course), and replot appends its arguments to the previous plot or splot and executes the modified command.

Much of the general information about plotting can be found in the discussion of plot; information specific to 3-d can be found in the splot section.

plot operates in either rectangular or polar coordinates – see set polar (p. 425) for details of the latter. splot operates only in rectangular coordinates, but the set mapping command allows for a few other coordinate systems to be treated. In addition, the using option allows both plot and splot to treat almost any coordinate system you’d care to define.

plot also lets you use each of the four borders – x (bottom), x2 (top), y (left) and y2 (right) – as an independent axis. The axes option lets you choose which pair of axes a given function or data set is plotted against. A full complement of set commands exists to give you complete control over the scales and labelling of each axis. Some commands have the name of an axis built into their names, such as set xlabel. Other commands have one or more axis names as options, such as set logscale xy. Commands and options controlling the z axis have no effect on 2-d graphs.

splot can plot surfaces and contours in addition to points and/or lines. In addition to splot, see set isosamples (p. 335) for information about defining the grid for a 3-d function; splot datafile (p. 540) for information about the requisite file structure for 3-d data values; and set contour (p. 291) and set cntrparam (p. 281) for information about contours.

In splot, control over the scales and labels of the axes are the same as with plot, except that commands and options controlling the x2 and y2 axes have no effect whereas of course those controlling the z axis do take effect.

18 Start-up

When gnuplot is run, it looks for an initialization file to load. This file is called .gnuplot on Unix and AmigaOS systems, and GNUPLOT.INI on other systems. If this file is not found in the current directory, the program will look for it in the HOME directory (under AmigaOS, Atari(single)TOS, MS-DOS, Windows and OS/2, the environment variable GNUPLOT should contain the name of this directory; on Windows NT, it will use USERPROFILE if GNUPLOT isn’t defined). Note: if NOCWDRC is defined during the installation, gnuplot will not read from the current directory.

If the initialization file is found, gnuplot executes the commands in it. These may be any legal gnuplot commands, but typically they are limited to setting the terminal and defining frequently-used functions or variables.

19 String constants and string variables

In addition to string constants, most gnuplot commands also accept a string variable, a string expression, or a function that returns a string. For example, the following four methods of creating a plot all result in the same plot title:

     four = "4"  
     graph4 = "Title for plot #4"  
     graph(n) = sprintf("Title for plot #%d",n)

     plot ’data.4’ title "Title for plot #4"  
     plot ’data.4’ title graph4  
     plot ’data.4’ title "Title for plot #".four  
     plot ’data.4’ title graph(4)

Since integers are promoted to strings when operated on by the string concatenation operator, the following method also works:

     N = 4  
     plot ’data.’.N title "Title for plot #".N

In general, elements on the command line will only be evaluated as possible string variables if they are not otherwise recognizable as part of the normal gnuplot syntax. So the following sequence of commands is legal, although probably should be avoided so as not to cause confusion:

     plot = "my_datafile.dat"  
     title = "My Title"  
     plot plot title title

There are three binary operators that require string operands: the string concatenation operator ".", the string equality operator "eq" and the string inequality operator "ne". The following example will print TRUE.

    if ("A"."B" eq "AB") print "TRUE"

See also the two string formatting functions gprintf (p. 323) and sprintf (p. 51).

Substrings can be specified by appending a range specifier to any string, string variable, or string-valued function. The range specifier has the form [begin:end], where begin is the index of the first character of the substring and end is the index of the last character of the substring. The first character has index 1. The begin or end fields may be empty, or contain ’*’, to indicate the true start or end of the original string. E.g. str[:] and str[*:*] both describe the full string str.

20 Substitution and Command line macros

When a command line to gnuplot is first read, i.e. before it is interpreted or executed, two forms of lexical substitution are performed. These are triggered by the presence of text in backquotes (ascii character 96) or preceded by @ (ascii character 64).

20.1 Substitution of system commands in backquotes

Command-line substitution is specified by a system command enclosed in backquotes. This command is spawned and the output it produces replaces the backquoted text on the command line. Some implementations also support pipes; see plot datafile special-filenames (p. 174).

Command-line substitution can be used anywhere on the gnuplot command line, except inside strings delimited by single quotes.


This will run the program leastsq and replace leastsq (including backquotes) on the command line with its output:

     f(x) = ‘leastsq‘

or, in VMS

     f(x) = ‘run leastsq‘

These will generate labels with the current time and userid:

     set label "generated on ‘date +%Y-%m-%d‘ by ‘whoami‘" at 1,1  
     set timestamp "generated on %Y-%m-%d by ‘whoami‘"

20.2 Substitution of string variables as macros

Substitution of command line macros is disabled by default, but may be enabled using the set macros command. If macro substitution is enabled, the character @ is used to trigger substitution of the current value of a string variable into the command line. The text in the string variable may contain any number of lexical elements. This allows string variables to be used as command line macros. Only string constants may be expanded using this mechanism, not string-valued expressions. For example:

     set macros  
     style1 = "lines lt 4 lw 2"  
     style2 = "points lt 3 pt 5 ps 2"  
     range1 = "using 1:3"  
     range2 = "using 1:5"  
     plot "foo" @range1 with @style1, "bar" @range2 with @style2

The line containing @ symbols is expanded on input, so that by the time it is executed the effect is identical to having typed in full

     plot "foo" using 1:3 with lines lt 4 lw 2, \  
          "bar" using 1:5 with points lt 3 pt 5 ps 2

The function exists() may be useful in connection with macro evaluation. The following example checks that C can safely be expanded as the name of a user-defined variable:

     C = "pi"  
     if (exists(C)) print C," = ", @C

Macro expansion does not occur inside either single or double quotes. However macro expansion does occur inside backquotes.

20.3 String variables, macros, and command line substitution

The interaction of string variables, backquotes and macro substitution is somewhat complicated. Backquotes do not block macro substitution, so

     filename = "mydata.inp"  
     lines = ‘ wc --lines @filename | sed "s/ .*//" ‘

results in the number of lines in mydata.inp being stored in the integer variable lines. And double quotes do not block backquote substitution, so

     mycomputer = "‘uname -n‘"

results in the string returned by the system command uname -n being stored in the string variable mycomputer.

However, macro substitution is not performed inside double quotes, so you cannot define a system command as a macro and then use both macro and backquote substitution at the same time.

      machine_id = "uname -n"  
      mycomputer = "‘@machine_id‘"  # doesn’t work!!

This fails because the double quotes prevent @machine_id from being interpreted as a macro. To store a system command as a macro and execute it later you must instead include the backquotes as part of the macro itself. This is accomplished by defining the macro as shown below. Notice that the sprintf format nests all three types of quotes.

     machine_id = sprintf(’"‘uname -n‘"’)  
     mycomputer = @machine_id

21 Syntax

Version 4 of gnuplot is much less sensitive than earlier versions to the order of keywords and suboptions. However, if you get error messages from specifying options that you think should work, please try rearranging them into the exact order listed by the documentation.

Options and any accompanying parameters are separated by spaces whereas lists and coordinates are separated by commas. Ranges are separated by colons and enclosed in brackets [], text and file names are enclosed in quotes, and a few miscellaneous things are enclosed in parentheses. Braces {} are used for a few special purposes.

Commas are used to separate coordinates on the set commands arrow, key, and label; the list of variables being fitted (the list after the via keyword on the fit command); lists of discrete contours or the loop parameters which specify them on the set cntrparam command; the arguments of the set commands dgrid3d, dummy, isosamples, offsets, origin, samples, size, time, and view; lists of tics or the loop parameters which specify them; the offsets for titles and axis labels; parametric functions to be used to calculate the x, y, and z coordinates on the plot, replot and splot commands; and the complete sets of keywords specifying individual plots (data sets or functions) on the plot, replot and splot commands.

Parentheses are used to delimit sets of explicit tics (as opposed to loop parameters) and to indicate computations in the using filter of the fit, plot, replot and splot commands.

(Parentheses and commas are also used as usual in function notation.)

Square brackets are used to delimit ranges given in set, plot or splot commands.

Colons are used to separate extrema in range specifications (whether they are given on set, plot or splot commands) and to separate entries in the using filter of the plot, replot, splot and fit commands.

Semicolons are used to separate commands given on a single command line.

Braces are used in text to be specially processed by some terminals, like postscript. They are also used to denote complex numbers: {3,2} = 3 + 2i.

At present you should not embed \n inside {} when using the PostScript terminal in enhanced text mode.

The EEPIC, Imagen, Uniplex, LaTeX, and TPIC drivers allow a newline to be specified by \\ in a single-quoted string or \\\\ in a double-quoted string.

21.1 Quote Marks

Gnuplot uses three forms of quote marks for delimiting text strings, double-quote (ascii 34), single-quote (ascii 39), and backquote (ascii 96).

Filenames may be entered with either single- or double-quotes. In this manual the command examples generally single-quote filenames and double-quote other string tokens for clarity.

String constants and text strings used for labels, titles, or other plot elements may be enclosed in either single quotes or double quotes. Further processing of the quoted text depends on the choice of quote marks.

Backslash processing of special characters like \n (newline) and \345 (octal character code) is performed for double-quoted strings. In single-quoted strings, backslashes are just ordinary characters. To get a single-quote (ascii 39) in a single-quoted string, it has to be doubled. Thus the strings "d\" s’ b\\" and ’d" s’ ’ b\’ are completely equivalent.

Text justification is the same for each line of a multi-line string. Thus the center-justified string

     "This is the first line of text.\nThis is the second line."

will produce

                      This is the first line of text.  
                         This is the second line.


     ’This is the first line of text.\nThis is the second line.’

will produce

         This is the first line of text.\nThis is the second line.

Enhanced text processing is performed for both double-quoted text and single-quoted text, but only by terminals supporting this mode. See enhanced text (p. 694).

Back-quotes are used to enclose system commands for substitution into the command line. See substitution (p. 91).

22 Time/Date data

gnuplot supports the use of time and/or date information as input data. This feature is activated by the commands set xdata time, set ydata time, etc.

Internally all times and dates are converted to the number of seconds from the year 2000. The command set timefmt defines the format for all inputs: data files, ranges, tics, label positions — in short, anything that accepts a data value must receive it in this format. Since only one input format can be in force at a given time, all time/date quantities being input at the same time must be presented in the same format. Thus if both x and y data in a file are time/date, they must be in the same format.

The conversion to and from seconds assumes Universal Time (which is the same as Greenwich Standard Time). There is no provision for changing the time zone or for daylight savings. If all your data refer to the same time zone (and are all either daylight or standard) you don’t need to worry about these things. But if the absolute time is crucial for your application, you’ll need to convert to UT yourself.

Commands like show xrange will re-interpret the integer according to timefmt. If you change timefmt, and then show the quantity again, it will be displayed in the new timefmt. For that matter, if you give the deactivation command (like set xdata), the quantity will be shown in its numerical form.

The commands set format or set tics format define the format that will be used for tic labels, whether or not the specified axis is time/date.

If time/date information is to be plotted from a file, the using option must be used on the plot or splot command. These commands simply use white space to separate columns, but white space may be embedded within the time/date string. If you use tabs as a separator, some trial-and-error may be necessary to discover how your system treats them.

The following example demonstrates time/date plotting.

Suppose the file "data" contains records like

     03/21/95 10:00  6.02e23

This file can be plotted by

     set xdata time  
     set timefmt "%m/%d/%y"  
     set xrange ["03/21/95":"03/22/95"]  
     set format x "%m/%d"  
     set timefmt "%m/%d/%y %H:%M"  
     plot "data" using 1:3

which will produce xtic labels that look like "03/21".

See the descriptions of each command for more details.

Part II

This section lists the commands acceptable to gnuplot in alphabetical order. Printed versions of this document contain all commands; on-line versions may not be complete. Indeed, on some systems there may be no commands at all listed under this heading.

Note that in most cases unambiguous abbreviations for command names and their options are permissible, i.e., "p f(x) w li" instead of "plot f(x) with lines".

In the syntax descriptions, braces ({}) denote optional arguments and a vertical bar (|) separates mutually exclusive choices.

23 Cd

The cd command changes the working directory.


     cd ’<directory-name>’

The directory name must be enclosed in quotes.


     cd ’subdir’  
     cd ".."

It is recommended for DOS and Windows users to use single-quotes — backslash [\] has special significance inside double-quotes and has to be escaped. For example,

     cd "c:\newdata"

fails, but

     cd ’c:\newdata’  
     cd "c:\\newdata"

works as expected.

24 Call

The call command is identical to the load command with one exception: you can have up to ten additional parameters to the command (delimited according to the standard parser rules) which can be substituted into the lines read from the file. As each line is read from the called input file, it is scanned for the sequence $ (dollar-sign) followed by a digit (0–9). If found, the sequence is replaced by the corresponding parameter from the call command line. If the parameter was specified as a string in the call line, it is substituted without its enclosing quotes. Sequence $# is replaced by the number of passed parameters. $ followed by any character will be that character; e.g. use $$ to get a single $. Providing more than ten parameters on the call command line will cause an error. A parameter that was not provided substitutes as nothing. Files being called may themselves contain call or load commands.

The call command must be the last command on a multi-command line.


     call "<input-file>" <parameter-0> <parm-1> ... <parm-9>

The name of the input file must be enclosed in quotes, and it is recommended that parameters are similarly enclosed in quotes (future versions of gnuplot may treat quoted and unquoted arguments differently).


If the file ’’ contains the line:

     print "argc=$# p0=$0 p1=$1 p2=$2 p3=$3 p4=$4 p5=$5 p6=$6 p7=x$7x"

entering the command:

     call ’’ "abcd" 1.2 + "’quoted’" -- "$2"

will display:

     argc=7 p0=abcd p1=1.2 p2=+ p3=’quoted’ p4=- p5=- p6=$2 p7=xx

NOTE: there is a clash in syntax with the datafile using callback operator. Use $$n or column(n) to access column n from a datafile inside a called datafile plot.

25 Clear

The clear command erases the current screen or output device as specified by set output. This usually generates a formfeed on hardcopy devices. Use set terminal to set the device type.

For some terminals clear erases only the portion of the plotting surface defined by set size, so for these it can be used in conjunction with set multiplot to create an inset.


     set multiplot  
     plot sin(x)  
     set origin 0.5,0.5  
     set size 0.4,0.4  
     plot cos(x)  
     unset multiplot

Please see set multiplot (p. 378), set size (p. 434), and set origin (p. 386) for details of these commands.

26 Exit

The commands exit and quit, as well as the END-OF-FILE character (usually Ctrl-D) terminate input from the current input stream: terminal session, pipe, and file input (pipe).

If input streams are nested (inherited load scripts), then reading will continue in the parent stream. When the top level stream is closed, the program itself will exit.

The command exit gnuplot will immediately and unconditionally cause gnuplot to exit even if the input stream is multiply nested. In this case any open output files may not be completed cleanly. Example of use:

     bind "ctrl-x" "unset output; exit gnuplot"

See help for batch/interactive (p. 37) for more details.

27 Fit

The fit command can fit a user-defined function to a set of data points (x,y) or (x,y,z), using an implementation of the nonlinear least-squares (NLLS) Marquardt-Levenberg algorithm. Any user-defined variable occurring in the function body may serve as a fit parameter, but the return type of the function must be real.


     fit {[xrange] {[yrange]}} <function> ’<datafile>’  
         via ’<parameter file>’ | <var1>{,<var2>,...}

Ranges may be specified to temporarily limit the data which is to be fitted; any out-of-range data points are ignored. The syntax is


analogous to plot; see plot ranges (p. 200).

<function> is any valid gnuplot expression, although it is usual to use a previously user-defined function of the form f(x) or f(x,y).

<datafile> is treated as in the plot command. All the plot datafile modifiers (using, every,...) except smooth and the deprecated thru are applicable to fit. See plot datafile (p. 151).

The default data formats for fitting functions with a single independent variable, y=f(x), are {x:}y or x:y:s; those formats can be changed with the datafile using qualifier. The third item (a column number or an expression), if present, is interpreted as the standard deviation of the corresponding y value and is used to compute a weight for the datum, 1/s**2. Otherwise, all data points are weighted equally, with a weight of one. Note that if you don’t specify a using option at all, no y deviations are read from the datafile even if it does have a third column, so you’ll always get unit weights.

To fit a function with two independent variables, z=f(x,y), the required format is using with four items, x:y:z:s. The complete format must be given — no default columns are assumed for a missing token. Weights for each data point are evaluated from ’s’ as above. If error estimates are not available, a constant value can be specified as a constant expression (see plot datafile using (p. 185)), e.g., using 1:2:3:(1).

Multiple datasets may be simultaneously fit with functions of one independent variable by making y a ’pseudo-variable’, e.g., the dataline number, and fitting as two independent variables. See fit multi-branch (p. 133).

The via qualifier specifies which parameters are to be adjusted, either directly, or by referencing a parameter file.


     f(x) = a*x**2 + b*x + c  
     g(x,y) = a*x**2 + b*y**2 + c*x*y  
     FIT_LIMIT = 1e-6  
     fit f(x) ’measured.dat’ via ’start.par’  
     fit f(x) ’measured.dat’ using 3:($7-5) via ’start.par’  
     fit f(x) ’./data/trash.dat’ using 1:2:3 via a, b, c  
     fit g(x,y) ’surface.dat’ using 1:2:3:(1) via a, b, c

After each iteration step, detailed information about the current state of the fit is written to the display. The same information about the initial and final states is written to a log file, "fit.log". This file is always appended to, so as to not lose any previous fit history; it should be deleted or renamed as desired. By using the command set fit logfile, the name of the log file can be changed.

If gnuplot was built with this option, and you activated it using set fit errorvariables, the error for each fitted parameter will be stored in a variable named like the parameter, but with "_err" appended. Thus the errors can be used as input for further computations.

The fit may be interrupted by pressing Ctrl-C (any key but Ctrl-C under MSDOS and Atari Multitasking Systems). After the current iteration completes, you have the option to (1) stop the fit and accept the current parameter values, (2) continue the fit, (3) execute a gnuplot command as specified by the environment variable FIT_SCRIPT. The default for FIT_SCRIPT is replot, so if you had previously plotted both the data and the fitting function in one graph, you can display the current state of the fit.

Once fit has finished, the update command may be used to store final values in a file for subsequent use as a parameter file. See update (p. 563) for details.

27.1 Adjustable parameters

There are two ways that via can specify the parameters to be adjusted, either directly on the command line or indirectly, by referencing a parameter file. The two use different means to set initial values.

Adjustable parameters can be specified by a comma-separated list of variable names after the via keyword. Any variable that is not already defined is created with an initial value of 1.0. However, the fit is more likely to converge rapidly if the variables have been previously declared with more appropriate starting values.

In a parameter file, each parameter to be varied and a corresponding initial value are specified, one per line, in the form

     varname = value

Comments, marked by ’#’, and blank lines are permissible. The special form

     varname = value       # FIXED

means that the variable is treated as a ’fixed parameter’, initialized by the parameter file, but not adjusted by fit. For clarity, it may be useful to designate variables as fixed parameters so that their values are reported by fit. The keyword # FIXED has to appear in exactly this form.

27.2 Short introduction

fit is used to find a set of parameters that ’best’ fits your data to your user-defined function. The fit is judged on the basis of the sum of the squared differences or ’residuals’ (SSR) between the input data points and the function values, evaluated at the same places. This quantity is often called ’chisquare’ (i.e., the Greek letter chi, to the power of 2). The algorithm attempts to minimize SSR, or more precisely, WSSR, as the residuals are ’weighted’ by the input data errors (or 1.0) before being squared; see fit error_estimates (p. 124) for details.

That’s why it is called ’least-squares fitting’. Let’s look at an example to see what is meant by ’non-linear’, but first we had better go over some terms. Here it is convenient to use z as the dependent variable for user-defined functions of either one independent variable, z=f(x), or two independent variables, z=f(x,y). A parameter is a user-defined variable that fit will adjust, i.e., an unknown quantity in the function declaration. Linearity/non-linearity refers to the relationship of the dependent variable, z, to the parameters which fit is adjusting, not of z to the independent variables, x and/or y. (To be technical, the second {and higher} derivatives of the fitting function with respect to the parameters are zero for a linear least-squares problem).

For linear least-squares (LLS), the user-defined function will be a sum of simple functions, not involving any parameters, each multiplied by one parameter. NLLS handles more complicated functions in which parameters can be used in a large number of ways. An example that illustrates the difference between linear and nonlinear least-squares is the Fourier series. One member may be written as

    z=a*sin(c*x) + b*cos(c*x).

If a and b are the unknown parameters and c is constant, then estimating values of the parameters is a linear least-squares problem. However, if c is an unknown parameter, the problem is nonlinear.

In the linear case, parameter values can be determined by comparatively simple linear algebra, in one direct step. However LLS is a special case which is also solved along with more general NLLS problems by the iterative procedure that gnuplot uses. fit attempts to find the minimum by doing a search. Each step (iteration) calculates WSSR with a new set of parameter values. The Marquardt-Levenberg algorithm selects the parameter values for the next iteration. The process continues until a preset criterion is met, either (1) the fit has "converged" (the relative change in WSSR is less than FIT_LIMIT), or (2) it reaches a preset iteration count limit, FIT_MAXITER (see fit control variables (p. 127)). The fit may also be interrupted and subsequently halted from the keyboard (see fit (p. 118)). The user variable FIT_CONVERGED contains 1 if the previous fit command terminated due to convergence; it contains 0 if the previous fit terminated for any other reason.

Often the function to be fitted will be based on a model (or theory) that attempts to describe or predict the behaviour of the data. Then fit can be used to find values for the free parameters of the model, to determine how well the data fits the model, and to estimate an error range for each parameter. See fit error_estimates (p. 124).

Alternatively, in curve-fitting, functions are selected independent of a model (on the basis of experience as to which are likely to describe the trend of the data with the desired resolution and a minimum number of parameters*functions.) The fit solution then provides an analytic representation of the curve.

However, if all you really want is a smooth curve through your data points, the smooth option to plot may be what you’ve been looking for rather than fit.

27.3 Error estimates

In fit, the term "error" is used in two different contexts, data error estimates and parameter error estimates.

Data error estimates are used to calculate the relative weight of each data point when determining the weighted sum of squared residuals, WSSR or chisquare. They can affect the parameter estimates, since they determine how much influence the deviation of each data point from the fitted function has on the final values. Some of the fit output information, including the parameter error estimates, is more meaningful if accurate data error estimates have been provided.

The ’statistical overview’ describes some of the fit output and gives some background for the ’practical guidelines’.

27.3.1 Statistical overview

The theory of non-linear least-squares (NLLS) is generally described in terms of a normal distribution of errors, that is, the input data is assumed to be a sample from a population having a given mean and a Gaussian (normal) distribution about the mean with a given standard deviation. For a sample of sufficiently large size, and knowing the population standard deviation, one can use the statistics of the chisquare distribution to describe a "goodness of fit" by looking at the variable often called "chisquare". Here, it is sufficient to say that a reduced chisquare (chisquare/degrees of freedom, where degrees of freedom is the number of datapoints less the number of parameters being fitted) of 1.0 is an indication that the weighted sum of squared deviations between the fitted function and the data points is the same as that expected for a random sample from a population characterized by the function with the current value of the parameters and the given standard deviations.

If the standard deviation for the population is not constant, as in counting statistics where variance = counts, then each point should be individually weighted when comparing the observed sum of deviations and the expected sum of deviations.

At the conclusion fit reports ’stdfit’, the standard deviation of the fit, which is the rms of the residuals, and the variance of the residuals, also called ’reduced chisquare’ when the data points are weighted. The number of degrees of freedom (the number of data points minus the number of fitted parameters) is used in these estimates because the parameters used in calculating the residuals of the datapoints were obtained from the same data. These values are exported to the variables

     FIT_NDF = Number of degrees of freedom  
     FIT_WSSR = Weighted sum-of-squares residual  
     FIT_STDFIT = sqrt(WSSR/NDF)

To estimate confidence levels for the parameters, one can use the minimum chisquare obtained from the fit and chisquare statistics to determine the value of chisquare corresponding to the desired confidence level, but considerably more calculation is required to determine the combinations of parameters which produce such values.

Rather than determine confidence intervals, fit reports parameter error estimates which are readily obtained from the variance-covariance matrix after the final iteration. By convention, these estimates are called "standard errors" or "asymptotic standard errors", since they are calculated in the same way as the standard errors (standard deviation of each parameter) of a linear least-squares problem, even though the statistical conditions for designating the quantity calculated to be a standard deviation are not generally valid for the NLLS problem. The asymptotic standard errors are generally over-optimistic and should not be used for determining confidence levels, but are useful for qualitative purposes.

The final solution also produces a correlation matrix, which gives an indication of the correlation of parameters in the region of the solution; if one parameter is changed, increasing chisquare, does changing another compensate? The main diagonal elements, autocorrelation, are all 1; if all parameters were independent, all other elements would be nearly 0. Two variables which completely compensate each other would have an off-diagonal element of unit magnitude, with a sign depending on whether the relation is proportional or inversely proportional. The smaller the magnitudes of the off-diagonal elements, the closer the estimates of the standard deviation of each parameter would be to the asymptotic standard error.

27.3.2 Practical guidelines

If you have a basis for assigning weights to each data point, doing so lets you make use of additional knowledge about your measurements, e.g., take into account that some points may be more reliable than others. That may affect the final values of the parameters.

Weighting the data provides a basis for interpreting the additional fit output after the last iteration. Even if you weight each point equally, estimating an average standard deviation rather than using a weight of 1 makes WSSR a dimensionless variable, as chisquare is by definition.

Each fit iteration will display information which can be used to evaluate the progress of the fit. (An ’*’ indicates that it did not find a smaller WSSR and is trying again.) The ’sum of squares of residuals’, also called ’chisquare’, is the WSSR between the data and your fitted function; fit has minimized that. At this stage, with weighted data, chisquare is expected to approach the number of degrees of freedom (data points minus parameters). The WSSR can be used to calculate the reduced chisquare (WSSR/ndf) or stdfit, the standard deviation of the fit, sqrt(WSSR/ndf). Both of these are reported for the final WSSR.

If the data are unweighted, stdfit is the rms value of the deviation of the data from the fitted function, in user units.

If you supplied valid data errors, the number of data points is large enough, and the model is correct, the reduced chisquare should be about unity. (For details, look up the ’chi-squared distribution’ in your favourite statistics reference.) If so, there are additional tests, beyond the scope of this overview, for determining how well the model fits the data.

A reduced chisquare much larger than 1.0 may be due to incorrect data error estimates, data errors not normally distributed, systematic measurement errors, ’outliers’, or an incorrect model function. A plot of the residuals, e.g., plot ’datafile’ using 1:($2-f($1)), may help to show any systematic trends. Plotting both the data points and the function may help to suggest another model.

Similarly, a reduced chisquare less than 1.0 indicates WSSR is less than that expected for a random sample from the function with normally distributed errors. The data error estimates may be too large, the statistical assumptions may not be justified, or the model function may be too general, fitting fluctuations in a particular sample in addition to the underlying trends. In the latter case, a simpler function may be more appropriate.

You’ll have to get used to both fit and the kind of problems you apply it to before you can relate the standard errors to some more practical estimates of parameter uncertainties or evaluate the significance of the correlation matrix.

Note that fit, in common with most NLLS implementations, minimizes the weighted sum of squared distances (y-f(x))**2. It does not provide any means to account for "errors" in the values of x, only in y. Also, any "outliers" (data points outside the normal distribution of the model) will have an exaggerated effect on the solution.

27.4 Control

There are a number of gnuplot variables that can be defined to affect fit. Those which can be defined once gnuplot is running are listed under ’control_variables’ while those defined before starting gnuplot are listed under ’environment_variables’.

27.4.1 Control variables

The default epsilon limit (1e-5) may be changed by declaring a value for


When the sum of squared residuals changes between two iteration steps by a factor less than this number (epsilon), the fit is considered to have ’converged’.

The maximum number of iterations may be limited by declaring a value for


A value of 0 (or not defining it at all) means that there is no limit.

If you need even more control about the algorithm, and know the Marquardt-Levenberg algorithm well, there are some more variables to influence it. The startup value of lambda is normally calculated automatically from the ML-matrix, but if you want to, you may provide your own one with


Specifying FIT_START_LAMBDA as zero or less will re-enable the automatic selection. The variable


gives the factor by which lambda is increased or decreased whenever the chi-squared target function increased or decreased significantly. Setting FIT_LAMBDA_FACTOR to zero re-enables the default factor of 10.0.

Other variables with the FIT_ prefix may be added to fit, so it is safer not to use that prefix for user-defined variables.

The variables FIT_SKIP and FIT_INDEX were used by earlier releases of gnuplot with a ’fit’ patch called gnufit and are no longer available. The datafile every modifier provides the functionality of FIT_SKIP. FIT_INDEX was used for multi-branch fitting, but multi-branch fitting of one independent variable is now done as a pseudo-3D fit in which the second independent variable and using are used to specify the branch. See fit multi-branch (p. 133).

27.4.2 Environment variables

The environment variables must be defined before gnuplot is executed; how to do so depends on your operating system.


changes the name (and/or path) of the file to which the fit log will be written from the default of "fit.log" in the working directory. The default value can be overwritten using the command set fit logfile.


specifies a command that may be executed after an user interrupt. The default is replot, but a plot or load command may be useful to display a plot customized to highlight the progress of the fit.

27.5 Multi-branch

In multi-branch fitting, multiple data sets can be simultaneously fit with functions of one independent variable having common parameters by minimizing the total WSSR. The function and parameters (branch) for each data set are selected by using a ’pseudo-variable’, e.g., either the dataline number (a ’column’ index of -1) or the datafile index (-2), as the second independent variable.

Example: Given two exponential decays of the form, z=f(x), each describing a different data set but having a common decay time, estimate the values of the parameters. If the datafile has the format x:z:s, then

    f(x,y) = (y==0) ? a*exp(-x/tau) : b*exp(-x/tau)  
    fit f(x,y) ’datafile’ using  1:-2:2:3  via a, b, tau

For a more complicated example, see the file "hexa.fnc" used by the "fit.dem" demo.

Appropriate weighting may be required since unit weights may cause one branch to predominate if there is a difference in the scale of the dependent variable. Fitting each branch separately, using the multi-branch solution as initial values, may give an indication as to the relative effect of each branch on the joint solution.

27.6 Starting values

Nonlinear fitting is not guaranteed to converge to the global optimum (the solution with the smallest sum of squared residuals, SSR), and can get stuck at a local minimum. The routine has no way to determine that; it is up to you to judge whether this has happened.

fit may, and often will get "lost" if started far from a solution, where SSR is large and changing slowly as the parameters are varied, or it may reach a numerically unstable region (e.g., too large a number causing a floating point overflow) which results in an "undefined value" message or gnuplot halting.

To improve the chances of finding the global optimum, you should set the starting values at least roughly in the vicinity of the solution, e.g., within an order of magnitude, if possible. The closer your starting values are to the solution, the less chance of stopping at another minimum. One way to find starting values is to plot data and the fitting function on the same graph and change parameter values and replot until reasonable similarity is reached. The same plot is also useful to check whether the fit stopped at a minimum with a poor fit.

Of course, a reasonably good fit is not proof there is not a "better" fit (in either a statistical sense, characterized by an improved goodness-of-fit criterion, or a physical sense, with a solution more consistent with the model.) Depending on the problem, it may be desirable to fit with various sets of starting values, covering a reasonable range for each parameter.

27.7 Tips

Here are some tips to keep in mind to get the most out of fit. They’re not very organized, so you’ll have to read them several times until their essence has sunk in.

The two forms of the via argument to fit serve two largely distinct purposes. The via "file" form is best used for (possibly unattended) batch operation, where you just supply the startup values in a file and can later use update to copy the results back into another (or the same) parameter file.

The via var1, var2, ... form is best used interactively, where the command history mechanism may be used to edit the list of parameters to be fitted or to supply new startup values for the next try. This is particularly useful for hard problems, where a direct fit to all parameters at once won’t work without good starting values. To find such, you can iterate several times, fitting only some of the parameters, until the values are close enough to the goal that the final fit to all parameters at once will work.

Make sure that there is no mutual dependency among parameters of the function you are fitting. For example, don’t try to fit a*exp(x+b), because a*exp(x+b)=a*exp(b)*exp(x). Instead, fit either a*exp(x) or exp(x+b).

A technical issue: the parameters must not be too different in magnitude. The larger the ratio of the largest and the smallest absolute parameter values, the slower the fit will converge. If the ratio is close to or above the inverse of the machine floating point precision, it may take next to forever to converge, or refuse to converge at all. You will have to adapt your function to avoid this, e.g., replace ’parameter’ by ’1e9*parameter’ in the function definition, and divide the starting value by 1e9.

If you can write your function as a linear combination of simple functions weighted by the parameters to be fitted, by all means do so. That helps a lot, because the problem is no longer nonlinear and should converge with only a small number of iterations, perhaps just one.

Some prescriptions for analysing data, given in practical experimentation courses, may have you first fit some functions to your data, perhaps in a multi-step process of accounting for several aspects of the underlying theory one by one, and then extract the information you really wanted from the fitting parameters of those functions. With fit, this may often be done in one step by writing the model function directly in terms of the desired parameters. Transforming data can also quite often be avoided, though sometimes at the cost of a more difficult fit problem. If you think this contradicts the previous paragraph about simplifying the fit function, you are correct.

A "singular matrix" message indicates that this implementation of the Marquardt-Levenberg algorithm can’t calculate parameter values for the next iteration. Try different starting values, writing the function in another form, or a simpler function.

Finally, a nice quote from the manual of another fitting package (fudgit), that kind of summarizes all these issues: "Nonlinear fitting is an art!"

28 Help

The help command displays on-line help. To specify information on a particular topic use the syntax:

     help {<topic>}

If <topic> is not specified, a short message is printed about gnuplot. After help for the requested topic is given, a menu of subtopics is given; help for a subtopic may be requested by typing its name, extending the help request. After that subtopic has been printed, the request may be extended again or you may go back one level to the previous topic. Eventually, the gnuplot command line will return.

If a question mark (?) is given as the topic, the list of topics currently available is printed on the screen.

29 History

history command lists or saves previous entries in the history of the command line editing, or executes an entry.

Here you find ’usage by examples’:

     history               # show the complete history  
     history 5             # show last 5 entries in the history  
     history quiet 5       # show last 5 entries without entry numbers  
     history ""     # write the complete history to file  
     history "" append # append the complete history to file  
     history 10 ""  # write last 10 commands to file  
     history 10 "|head -5 >>" # write 5 history commands using pipe  
     history ?load         # show all history entries starting with "load"  
     history ?"set c"      # like above, several words enclosed in quotes  
     hi !reread            # execute last entry starting with "reread"  
     hist !"set xr"        # like above, several words enclosed in quotes  
     hi !hi                # guess yourself :-))

On systems which support a popen function (Unix), the output of history can be piped through an external program by starting the file name with a ’|’, as one of the above examples demonstrates.

30 If

The if command allows commands to be executed conditionally.


     if (<condition>) <command-line> [; else if (<condition>) ...; else ...]

<condition> will be evaluated. If it is true (non-zero), then the command(s) of the <command-line> will be executed. If <condition> is false (zero), then the entire <command-line> is ignored until the next occurrence of else. Note that use of ; to allow multiple commands on the same line will not end the conditionalized commands.


     if (pi!=acos(-1)) print "?Fixing pi!"; pi=acos(-1); print pi

will display:

     ?Fixing pi!  


     if (1==2) print "Never see this"; print "Or this either"

will not display anything.


     v=v+1; if (v%2) print "2" ; else if (v%3) print "3"; else print "fred"

(repeat the last line repeatedly!)

See reread (p. 234) for an example of how if (p. 137) and reread (p. 234) can be used together to perform a loop.

31 Load

The load command executes each line of the specified input file as if it had been typed in interactively. Files created by the save command can later be loaded. Any text file containing valid commands can be created and then executed by the load command. Files being loaded may themselves contain load or call commands. See comments (p. 41) for information about comments in commands. To load with arguments, see call (p. 112).

The load command must be the last command on a multi-command line.


     load "<input-file>"

The name of the input file must be enclosed in quotes.

The special filename "-" may be used to load commands from standard input. This allows a gnuplot command file to accept some commands from standard input. Please see help for batch/interactive (p. 37) for more details.

On some systems which support a popen function (Unix), the load file can be read from a pipe by starting the file name with a ’<’.


     load ’work.gnu’  
     load "func.dat"  
     load "<"

The load command is performed implicitly on any file names given as arguments to gnuplot. These are loaded in the order specified, and then gnuplot exits.

32 Lower


     lower {plot_window_nb}

The lower command lowers (opposite to raise) plot window(s) associated with the interactive terminal of your gnuplot session, i.e. pm, win, wxt or x11. It puts the plot window to bottom in the z-order windows stack of the window manager of your desktop.

As x11 and wxt support multiple plot windows, then by default they lower these windows in descending order of most recently created on top to the least recently created on bottom. If a plot number is supplied as an optional parameter, only the associated plot window will be lowered if it exists.

The optional parameter is ignored for single plot-window terminals, i.e. pm and win.

33 Pause

The pause command displays any text associated with the command and then waits a specified amount of time or until the carriage return is pressed. pause is especially useful in conjunction with load files.


     pause <time> {"<string>"}  
     pause mouse {<endcondition>}{, <endcondition>} {"<string>"}

<time> may be any constant or expression. Choosing -1 will wait until a carriage return is hit, zero (0) won’t pause at all, and a positive number will wait the specified number of seconds. The time is rounded to an integer number of seconds if subsecond time resolution is not supported by the given platform. pause 0 is synonymous with print.

If the current terminal supports mousing, then pause mouse will terminate on either a mouse click or on ctrl-C. For all other terminals, or if mousing is not active, pause mouse is equivalent to pause -1.

If one or more end conditions are given after pause mouse, then any one of the conditions will terminate the pause. The possible end conditions are keypress, button1, button2, button3, close, and any. If the pause terminates on a keypress, then the ascii value of the key pressed is returned in MOUSE_KEY. The character itself is returned as a one character string in MOUSE_CHAR. Hotkeys (bind command) are disabled if keypress is one of the end conditions. Zooming is disabled if button3 is one of the end conditions.

In all cases the coordinates of the mouse are returned in variables MOUSE_X, MOUSE_Y, MOUSE_X2, MOUSE_Y2. See mouse variables (p. 83).

Note: Since pause communicates with the operating system rather than the graphics, it may behave differently with different device drivers (depending upon how text and graphics are mixed).


     pause -1    # Wait until a carriage return is hit  
     pause 3     # Wait three seconds  
     pause -1  "Hit return to continue"  
     pause 10  "Isn’t this pretty?  It’s a cubic spline."  
     pause mouse "Click any mouse button on selected data point"  
     pause mouse keypress "Type a letter from A-F in the active window"  
     pause mouse button1,keypress  
     pause mouse any "Any key or button will terminate"

The variant "pause mouse key" will resume after any keypress in the active plot window. If you want to wait for a particular key to be pressed, you can use a reread loop such as:

     printf "I will resume after you hit the Tab key in the plot window"  
     load "wait_for_tab"

File "wait_for_tab" contains the lines

     pause mouse key  
     if (MOUSE_KEY != 9) reread

34 Plot

plot is the primary command for drawing plots with gnuplot. It creates plots of functions and data in many, many ways. plot is used to draw 2-d functions and data; splot draws 2-d projections of 3-d surfaces and data. plot and splot contain many common features; see splot (p. 539) for differences. Note specifically that although the binary <binary list> variation does work for both plot and splot, there are small differences between these modes. Furthermore, plot’s axes option does not exist for splot.


     plot {<ranges>}  
          {<function> | {"<datafile>" {datafile-modifiers}}}  
          {axes <axes>} {<title-spec>} {with <style>}  
          {, {definitions,} <function> ...}

where either a <function> or the name of a data file enclosed in quotes is supplied. A function is a mathematical expression or a pair of mathematical expressions in parametric mode. The expressions may be defined completely or in part earlier in the stream of gnuplot commands (see user-defined (p. 57)).

It is also possible to define functions and parameters on the plot command itself. This is done merely by isolating them from other items with commas.

There are four possible sets of axes available; the keyword <axes> is used to select the axes for which a particular line should be scaled. x1y1 refers to the axes on the bottom and left; x2y2 to those on the top and right; x1y2 to those on the bottom and right; and x2y1 to those on the top and left. Ranges specified on the plot command apply only to the first set of axes (bottom left).


     plot sin(x)  
     plot f(x) = sin(x*a), a = .2, f(x), a = .4, f(x)  
     plot [t=1:10] [-pi:pi*2] tan(t), \  
          "data.1" using (tan($2)):($3/$4) smooth csplines \  
                   axes x1y2 notitle with lines 5

See also show plot (p. 390).

34.1 Data

Discrete data contained in a file can be displayed by specifying the name of the data file (enclosed in single or double quotes) on the plot command line.


     plot ’<file_name>’ {binary <binary list>}  
                        {index <index list>}  
                        {every <every list>}  
                        {thru <thru expression>}  
                        {using <using list>}  
                        {smooth <option>}

The modifiers binary, index, every, thru, using, and smooth are discussed separately. In brief, binary allows data entry from a binary file (default is ASCII), index selects which data sets in a multi-data-set file are to be plotted, every specifies which points within a single data set are to be plotted, using determines how the columns within a single record are to be interpreted (thru is a special case of using), and smooth allows for simple interpolation and approximation. (splot has a similar syntax, but does not support the smooth and thru options.)


Data files should contain at least one data point per record (using can select one data point from the record). Records beginning with # (and also with ! on VMS) will be treated as comments and ignored. Each data point represents an (x,y) pair. For plots with error bars or error bars with lines (see set style errorbars (p. 486) or set style errorlines (p. 487)), each data point is (x,y,ydelta), (x,y,ylow,yhigh), (x,y,xdelta), (x,y,xlow,xhigh), or (x,y,xlow,xhigh,ylow,yhigh).

In all cases, the numbers of each record of a data file must be separated by white space (one or more blanks or tabs) unless a format specifier is provided by the using option. This white space divides each record into columns. However, whitespace inside a pair of double quotes is ignored when counting columns, so the following datafile line has three columns:

     1.0 "second column" 3.0

Data may be written in exponential format with the exponent preceded by the letter e or E. The fortran exponential specificiers d, D, q, and Q may also be used if the command set datafile fortran is in effect.

Only one column (the y value) need be provided. If x is omitted, gnuplot provides integer values starting at 0.

In datafiles, blank records (records with no characters other than blanks and a newline and/or carriage return) are significant.

Single blank records designate discontinuities in a plot; no line will join points separated by a blank records (if they are plotted with a line style).

Two blank records in a row indicate a break between separate data sets. See index (p. 168).

If autoscaling has been enabled (set autoscale), the axes are automatically extended to include all datapoints, with a whole number of tic marks if tics are being drawn. This has two consequences: i) For splot, the corner of the surface may not coincide with the corner of the base. In this case, no vertical line is drawn. ii) When plotting data with the same x range on a dual-axis graph, the x coordinates may not coincide if the x2tics are not being drawn. This is because the x axis has been autoextended to a whole number of tics, but the x2 axis has not. The following example illustrates the problem:

     reset; plot ’-’, ’-’ axes x2y1  
     1 1  
     19 19  
     1 1  
     19 19  

To avoid this, you can use the fixmin/fixmax feature of the set autoscale command, which turns off the automatic extension of the axis range upto the next tic mark.


Gnuplot can read binary data files. However, adequate information about details of the file format must be given on the command line or extracted from the file itself for a supported binary filetype. In particular, there are two structures for binary files, a matrix binary format and a general binary format.

The matrix binary format contains a two dimensional array of 32 bit IEEE float values with an additional column and row of coordinate values. As with ASCII matrix, in the using list, repetition of the coordinate row constitutes column 1, repetition of the coordinate column constitutes column 2, and the array of values constitutes column 3.

The general binary format contains an arbitrary number of columns for which information must be specified at the command line. For example, array, record, format and using can indicate the size, format and dimension of data. There are a variety of useful commands for skipping file headers and changing endianess. There are a set of commands for positioning and translating data since often coordinates are not part of the file when uniform sampling is inherent in the data. Different from matrix binary or ASCII, general binary does not treat the generated columns as 1, 2 or 3 in the using list. Rather, column 1 begins with column 1 of the file, or as specified in the format list.

There are global default settings for the various binary options which may be set using the same syntax as the options when used as part of the (s)plot <filename> binary ... command. This syntax is set datafile binary .... The general rule is that common command-line specified parameters override file-extracted parameters which override default parameters.

Matrix binary is the default binary format when no keywords specific to general binary are given, i.e., array, record, format, filetype.

General binary data can be entered at the command line via the special file name ’-’. However, this is intended for use through a pipe where programs can exchange binary data, not for keyboards. There is no "end of record" character for binary data. Gnuplot continues reading from a pipe until it has read the number of points declared in the array qualifier.

See datafile binary (p. 154) for more details.

34.1.1 Binary

The binary keyword allows a data file to be binary as opposed to ASCII. There are two formats for binary–matrix binary and general binary. Matrix binary is a fixed format in which data appears in a 2D array with an extra row and column for coordinate values. General binary is a flexible format for which details about the file must be given at the command line.

See binary matrix (p. 541) or binary general (p. 155) for more details.

34.1.2 Binary general

General binary data in which format information is not necessarily part of the file can be read by giving further details about the file format at the command line. Although the syntax is slightly arcane to the casual user, general binary is particularly useful for application programs using gnuplot and sending large amounts of data.


     plot ’<file_name>’ {binary <binary list>} ...  
     splot ’<file_name>’ {binary <binary list>} ...

General binary format is activated by keywords in <binary list> pertaining to information about file structure, i.e., array, record, format or filetype. Otherwise, matrix binary format is assumed. (See binary matrix (p. 541) for more details.)

There are some standard file types that may be read for which details about the binary format may be extracted automatically. (Type show datafile binary at the command line for a list.) Otherwise, details must be specified at the command line or set in the defaults. Keywords are described below.

The keyword filetype in <binary list> controls the routine used to read the file, i.e., the format of the data. For a list of the supported file types, type show datafile binary filetypes. If no file type is given, the rule is that traditional gnuplot binary is assumed for splot if the binary keyword stands alone. In all other circumstances, for plot or when one of the <binary list> keywords appears, a raw binary file is assumed whereby the keywords specify the binary format.

General binary data files fall into two basic classes, and some files may be of both classes depending upon how they are treated. There is that class for which uniform sampling is assumed and point coordinates must be generated. This is the class for which full control via the <binary list> keywords applies. For this class, the settings precedence is that command line parameters override in-file parameters, which override default settings. The other class is that set of files for which coordinate information is contained within the file or there is possibly a non-uniform sampling such as gnuplot binary.

Other than for the unique data files such as gnuplot binary, one should think of binary data as conceptually the same as ASCII data. Each point has columns of information which are selected via the <using list> associated with using. When no format string is specified, gnuplot will retrieve a number of binary variables equal to the largest column given in the <using list>. For example, using 1:3 will result in three columns being read, of which the second will be ignored. There are default using lists based upon the typical number of parameters associated with a certain plot type. For example, with image has a default of using 1, while with rgbimage has a default of using 1:2:3. Note that the special characters for using representing point/line/index generally should not be used for binary data. There are keywords in <binary list> that control this.

Array Describes the sampling array dimensions associated with the binary file. The coordinates will be generated by gnuplot. A number must be specified for each dimension, thereby calling out the size of the array. For example, array=10x20 means the underlying sampling structure is two-dimensional with 10 points along the first (x) dimension and 20 points along the second (y) dimension. A special "number", Inf, can be used to indicate that data should be read until the end of file. A colon can be used to separate the dimensions for multiple records. For example, array=25:35 indicates there are two one-dimensional records within the file. The colon behavior applies to the remaining keywords in this list for which it makes sense to be associated with individual records.

Currently, syntax allows for up to three-dimensional arrays. However, no conventions have yet been made for handling three-dimensional coordinates.

Record This keyword serves the same function as array, having the same syntax. However, record causes gnuplot to not generate coordinate information. This is for the case where such information may be included in one of the columns of the binary data file.

Format The default binary format is a float. For more flexibility, the format can include details about variable sizes. For example, format="%uchar%int%float" associates an unsigned character with the first using column, an int with the second column and a float with the third column. If the number of size specifications is less than the greatest column number, the size is implicitly taken to be similar to the last given variable size.

Furthermore, the format specification can include "discarded" terms via the * character. For example, to skip the middle column of the previous example, one could write format="%uchar%*int%float" and gnuplot will discard the middle integer. To list variable sizes, type show datafile binary datasizes. There are a group of names that are machine dependent along with their sizes in bytes for the particular compilation. There is also a group of names which attempt to be machine independent.

Endian Often the endianess of binary data in the file does not agree with the endianess used by the platform on which gnuplot is running. Several words can direct gnuplot how to arrange bytes. For example endian=little means treat the binary file as having byte significance from least to greatest. The options are

             little:  least significant to greatest significance  
                big:  greatest significance to least significance  
            default:  assume file endianess is the same as compiler  
        swap (swab):  Interchange the significance.  (If things  
                      don’t look right, try this.)

Gnuplot can support "middle" ("pdp") endian if it is compiled with that option.

Filetype For some standard binary file formats gnuplot can extract all the necessary information from the file in question. As an example, "format=edf" will read ESRF Header File format files. For a list of the currently supported file formats, type show datafile binary filetypes.

There is a special file type called auto for which gnuplot will check if the binary file’s extension is a quasi-standard extension for a supported format.

Command line keywords may be used to override settings extracted from the file. The settings from the file override any defaults. (See set datafile binary (p. 305) for details.)

Avs avs is one of the automatically recognized binary file types for images. AVS is an extremely simple format, suitable mostly for streaming between applications. It consists of 2 longs (xwidth, ywidth) followed by a stream of pixels, each with four bytes of information alpha/red/green/blue.

Edf edf is one of the automatically recognized binary file types for images. EDF stands for ESRF Data Format, and it supports both edf and ehf formats (the latter means ESRF Header Format). More information on specifications can be found at

See also binary (p. 154).

Keywords The following keywords apply only when generating coordinates. That is, when the keyword array is used.

Scan A great deal of confusion can arise concerning the relationship between how gnuplot scans a binary file and the dimensions seen on the plot. To lessen the confusion, conceptually think of gnuplot always scanning the binary file point/line/plane or fast/medium/slow. Then this keyword is used to tell gnuplot how to map this scanning convention to the Cartesian convention shown in plots, i.e., x/y/z. The qualifier for scan is a two or three letter code representing where point is assigned (first letter), line is assigned (second letter), and plane is assigned (third letter). For example, scan=yx means the fastest, point-by-point, increment should be mapped along the Cartesian y dimension and the middle, line-by-line, increment should be mapped along the x dimension. When the plotting mode is plot, the qualifier code can include the two letters x and y. For splot, it can include the three letters x, y and z.

There is nothing restricting the inherent mapping from point/line/plane to apply only to Cartesian coordinates. For this reason there are cylindrical coordinate synonyms for the qualifier codes where t (theta), r and z are analogous to the x, y and z of Cartesian coordinates.

Transpose Shorthand notation for scan=yx or scan=yxz.

Dx, dy, dz When gnuplot generates coordinates, it uses the spacing described by these keywords. For example dx=10 dy=20 would mean space samples along the x dimension by 10 and space samples along the y dimension by 20. dy cannot appear if dx does not appear. Similarly, dz cannot appear if dy does not appear. If the underlying dimensions are greater than the keywords specified, the spacing of the highest dimension given is extended to the other dimensions. For example, if an image is being read from a file and only dx=3.5 is given gnuplot uses a delta x and delta y of 3.5. The following keywords also apply only when generating coordinates. However they may also be used with matrix binary files.

Flipx, flipy, flipz Sometimes the scanning directions in a binary datafile are not consistent with that assumed by gnuplot. These keywords can flip the scanning direction along dimensions x, y, z.

Origin When gnuplot generates coordinates based upon transposition and flip, it attempts to always position the lower left point in the array at the origin, i.e., the data lies in the first quadrant of a Cartesian system after transpose and flip. To position the array somewhere else on the graph, the origin keyword directs gnuplot to position the lower left point of the array at a point specified by a tuple. The tuple should be a double for plot and a triple for splot. For example, origin=(100,100):(100,200) is for two records in the file and intended for plotting in two dimensions. A second example, origin=(0,0,3.5), is for plotting in three dimensions.

Center Similar to origin, this keyword will position the array such that its center lies at the point given by the tuple. For example, center=(0,0). Center does not apply when the size of the array is Inf.

Rotate The transpose and flip commands provide some flexibility in generating and orienting coordinates. However, for full degrees of freedom, it is possible to apply a rotational vector described by a rotational angle in two dimensions. The rotate keyword applies to the two-dimensional plane, whether it be plot or splot. The rotation is done with respect to the positive angle of the Cartesian plane.

The angle can be expressed in radians, radians as a multiple of pi, or degrees. For example, rotate=1.5708, rotate=0.5pi and rotate=90deg are equivalent.

If origin is specified, the rotation is done about the lower left sample point before translation. Otherwise, the rotation is done about the array center.

Perpendicular For splot, the concept of a rotational vector is implemented by a triple representing the vector to be oriented normal to the two-dimensional x-y plane. Naturally, the default is (0,0,1). Thus specifying both rotate and perpendicular together can orient data myriad ways in three-space. The two-dimensional rotation is done first, followed by the three-dimensional rotation. That is, if R’ is the rotational 2 x 2 matrix described by an angle, and P is the 3 x 3 matrix projecting (0,0,1) to (xp,yp,zp), let R be constructed from R’ at the upper left sub-matrix, 1 at element 3,3 and zeros elsewhere. Then the matrix formula for translating data is v’ = P R v, where v is the 3 x 1 vector of data extracted from the data file. In cases where the data of the file is inherently not three-dimensional, logical rules are used to place the data in three-space. (E.g., usually setting the z-dimension value to zero and placing 2D data in the x-y plane.)

Binary examples Examples:

     # Selects two float values (second one implicit) with a float value  
     # discarded between them for an indefinite length of 1D data.  
     plot ’<file_name>’ binary format="%float%*float" using 1:2 with lines

     # The data file header contains all details necessary for creating  
     # coordinates from an EDF file.  
     plot ’<file_name>’ binary filetype=edf with image  
     plot ’<file_name>.edf’ binary filetype=auto with image

     # Selects three unsigned characters for components of a raw RGB image  
     # and flips the y-dimension so that typical image orientation (start  
     # at top left corner) translates to the Cartesian plane.  Pixel  
     # spacing is given and there are two images in the file.  One of them  
     # is translated via origin.  
     plot ’<file_name>’ binary array=512x1024:1024x512 format=’%uchar’ \  
          dx=2:1 dy=1:2 origin=(0,0):(1024,1024) flipy u 1:2:3 w rgbimage

     # Four separate records in which the coordinates are part of the  
     # data file.  The file was created with a endianess different from  
     # the system on which gnuplot is running.  
     splot ’<file_name>’ binary record=30:30:29:26 endian=swap u 1:2:3

See also binary matrix (p. 541).

34.1.3 Every

The every keyword allows a periodic sampling of a data set to be plotted.

In the discussion a "point" is a datum defined by a single record in the file; "block" here will mean the same thing as "datablock" (see glossary (p. 63)).


     plot ’file’ every {<point_incr>}  

The data points to be plotted are selected according to a loop from <start_point> to <end_point> with increment <point_incr> and the blocks according to a loop from <start_block> to <end_block> with increment <block_incr>.

The first datum in each block is numbered ’0’, as is the first block in the file.

Note that records containing unplottable information are counted.

Any of the numbers can be omitted; the increments default to unity, the start values to the first point or block, and the end values to the last point or block. If every is not specified, all points in all lines are plotted.


     every :::3::3    # selects just the fourth block (’0’ is first)  
     every :::::9     # selects the first 10 blocks  
     every 2:2        # selects every other point in every other block  
     every ::5::15    # selects points 5 through 15 in each block


simple plot demos (simple.dem)


Non-parametric splot demos

, and

Parametric splot demos


34.1.4 Example datafile

This example plots the data in the file "population.dat" and a theoretical curve:

     pop(x) = 103*exp((1965-x)/10)  
     plot [1960:1990] ’population.dat’, pop(x)

The file "population.dat" might contain:

     # Gnu population in Antarctica since 1965  
        1965   103  
        1970   55  
        1975   34  
        1980   24  
        1985   10

34.1.5 Index

The index keyword allows you to select specific data sets in a multi-data-set file for plotting.


     plot ’file’ index <m>{{:<n>}:<p>}

Data sets are separated by pairs of blank records. index <m> selects only set <m>; index <m>:<n> selects sets in the range <m> to <n>; and index <m>:<n>:<p> selects indices <m>, <m>+<p>, <m>+2<p>, etc., but stopping at <n>. Following C indexing, the index 0 is assigned to the first data set in the file. Specifying too large an index results in an error message. If index is not specified, all sets are plotted as a single data set.


     plot ’file’ index 4:5

For each point in the file, the index value of the data set it appears in is available via the pseudo-column column(-2). This leads to an alternative way of distinguishing individual data sets within a file as shown below. This is more awkward that the index command if all you are doing is selecting one data set for plotting, but is very useful if you want to assign different properties to each data set. See lc variable (p. 71).


     plot ’file’ using 1:(column(-2)==4 ? $2 : NaN)        # very awkward  
     plot ’file’ using 1:2:(column(-2)) linecolor variable # very useful!

34.1.6 Smooth

gnuplot includes a few general-purpose routines for interpolation and approximation of data; these are grouped under the smooth option. More sophisticated data processing may be performed by preprocessing the data externally or by using fit with an appropriate model.


     smooth {unique | frequency | csplines | acsplines | bezier | sbezier}

unique and frequency plot the data after making them monotonic. Each of the other routines uses the data to determine the coefficients of a continuous curve between the endpoints of the data. This curve is then plotted in the same manner as a function, that is, by finding its value at uniform intervals along the abscissa (see set samples (p. 433)) and connecting these points with straight line segments (if a line style is chosen).

If autoscale is in effect, the ranges will be computed such that the plotted curve lies within the borders of the graph.

If autoscale is not in effect, and the smooth option is either acspline or cspline, the sampling of the generated curve is done across the intersection of the x range covered by the input data and the fixed abscissa range as defined by set xrange.

If too few points are available to allow the selected option to be applied, an error message is produced. The minimum number is one for unique and frequency, four for acsplines, and three for the others.

The smooth options have no effect on function plots.

Acsplines The acsplines option approximates the data with a "natural smoothing spline". After the data are made monotonic in x (see smooth unique (p. 174)), a curve is piecewise constructed from segments of cubic polynomials whose coefficients are found by the weighting the data points; the weights are taken from the third column in the data file. That default can be modified by the third entry in the using list, e.g.,

     plot ’data-file’ using 1:2:(1.0) smooth acsplines

Qualitatively, the absolute magnitude of the weights determines the number of segments used to construct the curve. If the weights are large, the effect of each datum is large and the curve approaches that produced by connecting consecutive points with natural cubic splines. If the weights are small, the curve is composed of fewer segments and thus is smoother; the limiting case is the single segment produced by a weighted linear least squares fit to all the data. The smoothing weight can be expressed in terms of errors as a statistical weight for a point divided by a "smoothing factor" for the curve so that (standard) errors in the file can be used as smoothing weights.


     plot ’data_file’ using 1:2:(sw($3,100)) smooth acsplines

Bezier The bezier option approximates the data with a Bezier curve of degree n (the number of data points) that connects the endpoints.

Csplines The csplines option connects consecutive points by natural cubic splines after rendering the data monotonic (see smooth unique (p. 174)).

Sbezier The sbezier option first renders the data monotonic (unique) and then applies the bezier algorithm.

Unique The unique option makes the data monotonic in x; points with the same x-value are replaced by a single point having the average y-value. The resulting points are then connected by straight line segments.


Frequency The frequency option makes the data monotonic in x; points with the same x-value are replaced by a single point having the summed y-values. The resulting points are then connected by straight line segments.

34.1.7 Special-filenames

A special filename of ’-’ specifies that the data are inline; i.e., they follow the command. Only the data follow the command; plot options like filters, titles, and line styles remain on the plot command line. This is similar to << in unix shell script, and $DECK in VMS DCL. The data are entered as though they are being read from a file, one data point per record. The letter "e" at the start of the first column terminates data entry. The using option can be applied to these data — using it to filter them through a function might make sense, but selecting columns probably doesn’t!

’-’ is intended for situations where it is useful to have data and commands together, e.g., when gnuplot is run as a sub-process of some front-end application. Some of the demos, for example, might use this feature. While plot options such as index and every are recognized, their use forces you to enter data that won’t be used. For example, while

     plot ’-’ index 0, ’-’ index 1  



does indeed work,

     plot ’-’, ’-’  

is a lot easier to type.

If you use ’-’ with replot, you may need to enter the data more than once (see replot (p. 233)).

A blank filename (’ ’) specifies that the previous filename should be reused. This can be useful with things like

     plot ’a/very/long/filename’ using 1:2, ’’ using 1:3, ’’ using 1:4

(If you use both ’-’ and ’ ’ on the same plot command, you’ll need to have two sets of inline data, as in the example above.)

On some computer systems with a popen function (Unix), the datafile can be piped through a shell command by starting the file name with a ’<’. For example,

     pop(x) = 103*exp(-x/10)  
     plot "< awk ’{print $1-1965, $2}’ population.dat", pop(x)

would plot the same information as the first population example but with years since 1965 as the x axis. If you want to execute this example, you have to delete all comments from the data file above or substitute the following command for the first part of the command above (the part up to the comma):

     plot "< awk ’$0 !~ /^#/ {print $1-1965, $2}’ population.dat"

While this approach is most flexible, it is possible to achieve simple filtering with the using or thru keywords.

34.1.8 Thru

The thru function is provided for backward compatibility.


     plot ’file’ thru f(x)

It is equivalent to:

     plot ’file’ using 1:(f($2))

While the latter appears more complex, it is much more flexible. The more natural

     plot ’file’ thru f(y)

also works (i.e. you can use y as the dummy variable).

thru is parsed for splot and fit but has no effect.

34.1.9 Using

The most common datafile modifier is using.


     plot ’file’ using {<entry> {:<entry> {:<entry> ...}}} {’format’}

If a format is specified, each datafile record is read using the C library’s ’scanf’ function, with the specified format string. Otherwise the record is read and broken into columns. By default the separation between columns is whitespace (spaces and/or tabs), but see datafile separator (p. 297).

Each <entry> may be a simple column number that selects the value from one field of the input fit, an expression enclosed in parentheses, or empty.

If the entry is an expression in parentheses, then the function column(N) may be used to indicate the value in column N. That is, column(1) refers to the first item read, column(2) to the second, and so on. The special symbols $1, $2, ... are shorthand for column(1), column(2) ... The function valid(N) tests whether the value in the Nth column is a valid number.

In addition to the actual columns 1...N in the input data file, gnuplot presents data from several "pseudo-columns" that hold bookkeeping information. E.g. $0 or column(0) returns the sequence number of this data record within a dataset.

An empty <entry> will default to its order in the list of entries. For example, using ::4 is interpreted as using 1:2:4.

If the using list has but a single entry, that <entry> will be used for y and the data point number (pseudo-column $0) is used for x; for example, "plot ’file’ using 1" is identical to "plot ’file’ using 0:1". If the using list has two entries, these will be used for x and y. See set style (p. 439) and fit (p. 118) for details about plotting styles that make use of data from additional columns of input.

’scanf’ accepts several numerical specifications but gnuplot requires all inputs to be double-precision floating-point variables, so "%lf" is essentially the only permissible specifier. A format string given by the user must contain at least one such input specifier, and no more than seven of them. ’scanf’ expects to see white space — a blank, tab ("\t"), newline ("\n"), or formfeed ("\f") — between numbers; anything else in the input stream must be explicitly skipped.

Note that the use of "\t", "\n", or "\f" requires use of double-quotes rather than single-quotes.

Using_examples This creates a plot of the sum of the 2nd and 3rd data against the first: The format string specifies comma- rather than space-separated columns. The same result could be achieved by specifying set datafile separator ",".

     plot ’file’ using 1:($2+$3) ’%lf,%lf,%lf’

In this example the data are read from the file "MyData" using a more complicated format:

     plot ’MyData’ using "%*lf%lf%*20[^\n]%lf"

The meaning of this format is:

     %*lf        ignore a number  
     %lf         read a double-precision number (x by default)  
     %*20[^\n]   ignore 20 non-newline characters  
     %lf         read a double-precision number (y by default)

One trick is to use the ternary ?: operator to filter data:

     plot ’file’ using 1:($3>10 ? $2 : 1/0)

which plots the datum in column two against that in column one provided the datum in column three exceeds ten. 1/0 is undefined; gnuplot quietly ignores undefined points, so unsuitable points are suppressed. Or you can use the pre-defined variable NaN to achieve the same result.

In fact, you can use a constant expression for the column number, provided it doesn’t start with an opening parenthesis; constructs like using 0+(complicated expression) can be used. The crucial point is that the expression is evaluated once if it doesn’t start with a left parenthesis, or once for each data point read if it does.

If timeseries data are being used, the time can span multiple columns. The starting column should be specified. Note that the spaces within the time must be included when calculating starting columns for other data. E.g., if the first element on a line is a time with an embedded space, the y value should be specified as column three.

It should be noted that plot ’file’, plot ’file’ using 1:2, and plot ’file’ using ($1):($2) can be subtly different: 1) if file has some lines with one column and some with two, the first will invent x values when they are missing, the second will quietly ignore the lines with one column, and the third will store an undefined value for lines with one point (so that in a plot with lines, no line joins points across the bad point); 2) if a line contains text at the first column, the first will abort the plot on an error, but the second and third should quietly skip the garbage.

In fact, it is often possible to plot a file with lots of lines of garbage at the top simply by specifying

     plot ’file’ using 1:2

However, if you want to leave text in your data files, it is safer to put the comment character (#) in the first column of the text lines.

Feeble using demos.

If gnuplot is built with configuration option –enable-datastrings, then additional modifiers to using can specify handling of text fields in the datafile. See datastrings (p. 43), using xticlabels (p. 192), using title (p. 191).

Using title If gnuplot is built with configuration option –enable-datastrings, then the first entry of a column of the input data file can be used as a string to provide the plot title in the key box. The column containing specified is independent of the column[s] used for the plot itself.

  plot ’data’ using 1:($2/$3) title column(N)

In this case the entry in the first row of column N will be used for the key entry of the plot constructed from dividing column 2 by column 3. The entry in the first row of columns 2 and 3 will be ignored.

Xticlabels If gnuplot is built with configuration option –enable-datastrings, then a column of the input data file can be used to label axis tic marks. The format of such a plot command is

 plot ’datafile’ using <xcol>:<ycol>:xticlabels(<labelcol>) with <plotstyle>

Tic labels may be read for any of the plot axes: x x2 y y2 z. The ticlabels(<labelcol>) specifiers must come after all of the data coordinate specifiers in the using portion of the command. For each data point which has a valid set of X,Y[,Z] coordinates, the text field found in column <labelcol> is added to the list of xtic labels at the same X coordinate as the point it belongs to. xticlabels(<labelcol>) may be shortened to xtic(<labelcol>).


     splot "data" using 2:4:6:xtic(1):ytic(3):ztic(6)

In this example the x and y axis tic labels are taken from different columns than the x and y coordinate values. The z axis tics, however, are generated from the z coordinate of the corresponding point.

X2ticlabels See plot using xticlabels (p. 192).

Yticlabels See plot using xticlabels (p. 192).

Y2ticlabels See plot using xticlabels (p. 192).

Zticlabels See plot using xticlabels (p. 192).

34.2 Errorbars

Error bars are supported for 2-d data file plots by reading one to four additional columns (or using entries); these additional values are used in different ways by the various errorbar styles.

In the default situation, gnuplot expects to see three, four, or six numbers on each line of the data file — either

     (x, y, ydelta),  
     (x, y, ylow, yhigh),  
     (x, y, xdelta),  
     (x, y, xlow, xhigh),  
     (x, y, xdelta, ydelta), or  
     (x, y, xlow, xhigh, ylow, yhigh).

The x coordinate must be specified. The order of the numbers must be exactly as given above, though the using qualifier can manipulate the order and provide values for missing columns. For example,

     plot ’file’ with errorbars  
     plot ’file’ using 1:2:(sqrt($1)) with xerrorbars  
     plot ’file’ using 1:2:($1-$3):($1+$3):4:5 with xyerrorbars

The last example is for a file containing an unsupported combination of relative x and absolute y errors. The using entry generates absolute x min and max from the relative error.

The y error bar is a vertical line plotted from (x, ylow) to (x, yhigh). If ydelta is specified instead of ylow and yhigh, ylow = y - ydelta and yhigh = y + ydelta are derived. If there are only two numbers on the record, yhigh and ylow are both set to y. The x error bar is a horizontal line computed in the same fashion. To get lines plotted between the data points, plot the data file twice, once with errorbars and once with lines (but remember to use the notitle option on one to avoid two entries in the key). Alternately, use the errorlines command (see errorlines (p. 196)).

The error bars have crossbars at each end unless set bars is used (see set bars (p. 263) for details).

If autoscaling is on, the ranges will be adjusted to include the error bars.

See also

errorbar demos.

See plot using (p. 185), plot with (p. 213), and set style (p. 439) for more information.

34.3 Errorlines

Lines with error bars are supported for 2-d data file plots by reading one to four additional columns (or using entries); these additional values are used in different ways by the various errorlines styles.

In the default situation, gnuplot expects to see three, four, or six numbers on each line of the data file — either

     (x, y, ydelta),  
     (x, y, ylow, yhigh),  
     (x, y, xdelta),  
     (x, y, xlow, xhigh),  
     (x, y, xdelta, ydelta), or  
     (x, y, xlow, xhigh, ylow, yhigh).

The x coordinate must be specified. The order of the numbers must be exactly as given above, though the using qualifier can manipulate the order and provide values for missing columns. For example,

     plot ’file’ with errorlines  
     plot ’file’ using 1:2:(sqrt($1)) with xerrorlines  
     plot ’file’ using 1:2:($1-$3):($1+$3):4:5 with xyerrorlines

The last example is for a file containing an unsupported combination of relative x and absolute y errors. The using entry generates absolute x min and max from the relative error.

The y error bar is a vertical line plotted from (x, ylow) to (x, yhigh). If ydelta is specified instead of ylow and yhigh, ylow = y - ydelta and yhigh = y + ydelta are derived. If there are only two numbers on the record, yhigh and ylow are both set to y. The x error bar is a horizontal line computed in the same fashion.

The error bars have crossbars at each end unless set bars is used (see set bars (p. 263) for details).

If autoscaling is on, the ranges will be adjusted to include the error bars.

See plot using (p. 185), plot with (p. 213), and set style (p. 439) for more information.

34.4 Parametric

When in parametric mode (set parametric) mathematical expressions must be given in pairs for plot and in triplets for splot.


     plot sin(t),t**2  
     splot cos(u)*cos(v),cos(u)*sin(v),sin(u)

Data files are plotted as before, except any preceding parametric function must be fully specified before a data file is given as a plot. In other words, the x parametric function (sin(t) above) and the y parametric function (t**2 above) must not be interrupted with any modifiers or data functions; doing so will generate a syntax error stating that the parametric function is not fully specified.

Other modifiers, such as with and title, may be specified only after the parametric function has been completed:

     plot sin(t),t**2 title ’Parametric example’ with linespoints

See also

Parametric Mode Demos.

34.5 Ranges

The optional ranges specify the region of the graph that will be displayed.



The first form applies to the independent variable (xrange or trange, if in parametric mode). The second form applies to the dependent variable yrange (and xrange, too, if in parametric mode). <dummy-var> is a new name for the independent variable. (The defaults may be changed with set dummy.) The optional <min> and <max> terms can be constant expressions or *.

In non-parametric mode, the order in which ranges must be given is xrange and yrange.

In parametric mode, the order for the plot command is trange, xrange, and yrange. The following plot command shows setting the trange to [-pi:pi], the xrange to [-1.3:1.3] and the yrange to [-1:1] for the duration of the graph:

     plot [-pi:pi] [-1.3:1.3] [-1:1] sin(t),t**2

Note that the x2range and y2range cannot be specified here — set x2range and set y2range must be used.

Ranges are interpreted in the order listed above for the appropriate mode. Once all those needed are specified, no further ones must be listed, but unneeded ones cannot be skipped — use an empty range [] as a placeholder.

* can be used to allow autoscaling of either of min and max. See also set autoscale (p. 255).

Ranges specified on the plot or splot command line affect only that graph; use the set xrange, set yrange, etc., commands to change the default ranges for future graphs.

With time data, you must provide the range (in the same manner as the time appears in the datafile) within quotes. gnuplot uses the timefmt string to read the value — see set timefmt (p. 495).


This uses the current ranges:

     plot cos(x)

This sets the x range only:

     plot [-10:30] sin(pi*x)/(pi*x)

This is the same, but uses t as the dummy-variable:

     plot [t = -10 :30]  sin(pi*t)/(pi*t)

This sets both the x and y ranges:

     plot [-pi:pi] [-3:3]  tan(x), 1/x

This sets only the y range, and turns off autoscaling on both axes:

     plot [ ] [-2:sin(5)*-8] sin(x)**besj0(x)

This sets xmax and ymin only:

     plot [:200] [-pi:]  exp(sin(x))

This sets the x range for a timeseries:

     set timefmt "%d/%m/%y %H:%M"  
     plot ["1/6/93 12:00":"5/6/93 12:00"] ’timedata.dat’

34.6 Title

A line title for each function and data set appears in the key, accompanied by a sample of the line and/or symbol used to represent it. It can be changed by using the title option.


     title "<title>" | notitle ["<ignored title>"]

where <title> is the new title of the line and must be enclosed in quotes. The quotes will not be shown in the key. A special character may be given as a backslash followed by its octal value ("\345"). The tab character "\t" is understood. Note that backslash processing occurs only for strings enclosed in double quotes — use single quotes to prevent such processing. The newline character "\n" is not processed in key entries in either type of string.

The line title and sample can be omitted from the key by using the keyword notitle. A null title (title ’ ’) is equivalent to notitle. If only the sample is wanted, use one or more blanks (title ’ ’). If notitle is followed by a string this string is ignored.

If key autotitles is set (which is the default) and neither title nor notitle are specified the line title is the function name or the file name as it appears on the plot command. If it is a file name, any datafile modifiers specified will be included in the default title.

The layout of the key itself (position, title justification, etc.) can be controlled by set key. Please see set key (p. 336) for details.


This plots y=x with the title ’x’:

     plot x

This plots x squared with title "x^2" and file "data.1" with title "measured data":

     plot x**2 title "x^2", ’data.1’ t "measured data"

This puts an untitled circular border around a polar graph:

     set polar; plot my_function(t), 1 notitle

34.7 With

Functions and data may be displayed in one of a large number of styles. The with keyword provides the means of selection.


     with <style> { {linestyle | ls <line_style>}  
                    | {{linetype  | lt <line_type>}  
                       {linewidth | lw <line_width>}  
                       {linecolor | lc <colorspec>}  
                       {pointtype | pt <point_type>}  
                       {pointsize | ps <point_size>}  
                       {fill | fs <fillstyle>}  
                       {nohidden3d | nocontours}  

where <style> is either lines, points, linespoints, impulses, dots, steps, fsteps, histeps, errorbars, labels, xerrorbars, yerrorbars, xyerrorbars, errorlines, xerrorlines, yerrorlines, xyerrorlines, boxes, histograms, filledcurves, boxerrorbars, boxxyerrorbars, financebars, candlesticks, vectors, image, rgbimage or pm3d. Some of these styles require additional information. See plotting styles (p. 456) for details of each style. fill is relevant only to certain 2D plots (currently boxes boxxyerrorbars and candlesticks). Note that filledcurves and pm3d can take an additional option not listed above (the latter only when used in the splot command) — see their help or examples below for more details.

Default styles are chosen with the set style function and set style data commands.

By default, each function and data file will use a different line type and point type, up to the maximum number of available types. All terminal drivers support at least six different point types, and re-use them, in order, if more are required. The LaTeX driver supplies an additional six point types (all variants of a circle), and thus will only repeat after 12 curves are plotted with points. The PostScript drivers (postscript) supplies a total of 64.

If you wish to choose the line or point type for a single plot, <line_type> and <point_type> may be specified. These are positive integer constants (or expressions) that specify the line type and point type to be used for the plot. Use test to display the types available for your terminal.

You may also scale the line width and point size for a plot by using <line_width> and <point_size>, which are specified relative to the default values for each terminal. The pointsize may also be altered globally — see set pointsize (p. 424) for details. But note that both <point_size> as set here and as set by set pointsize multiply the default point size — their effects are not cumulative. That is, set pointsize 2; plot x w p ps 3 will use points three times default size, not six.

It is also possible to specify pointsize variable either as part of a line style or for an individual plot. In this case one extra column of input is required, i.e. 3 columns for a 2D plot and 4 columns for a 3D splot. The size of each individual point is determined by multiplying the global pointsize by the value read from the data file.

If you have defined specific line type/width and point type/size combinations with set style line, one of these may be selected by setting <line_style> to the index of the desired style.

If gnuplot was built with pm3d support, the special keyword palette is allowed for smooth color change of lines, points and dots in splots. The color is chosen from a smooth palette which was set previously with the command set palette. The color value corresponds to the z-value of the point coordinates or to the color coordinate if specified by the 4th parameter in using. Both 2d and 3d plots (plot and splot commands) can use palette colors as specified by either their fractional value or the corresponding value mapped to the colorbox range. A palette color value can also be read from an explicitly specified column in the using specifier. Z value. See colors (p. 64), set palette (p. 400), linetype (p. 63).

The keyword nohidden3d applies only to plots made with the splot command. Normally the global option set hidden3d applies to all plots in the graph. You can attach the nohidden3d option to any individual plots that you want to exclude from the hidden3d processing. The individual elements other than surfaces (i.e. lines, dots, labels, ...) of a plot marked nohidden3d will all be drawn, even if they would normally be obscured by other plot elements.

Similarly, the keyword nocontours will turn off contouring for an individual plot even if the global property "set contour" is active.

The keywords may be abbreviated as indicated.

Note that the linewidth, pointsize and palette options are not supported by all terminals.


This plots sin(x) with impulses:

     plot sin(x) with impulses

This plots x with points, x**2 with the default:

     plot x w points, x**2

This plots tan(x) with the default function style, file "data.1" with lines:

     plot [ ] [-2:5] tan(x), ’data.1’ with l

This plots "leastsq.dat" with impulses:

     plot ’leastsq.dat’ w i

This plots the data file "population" with boxes:

     plot ’population’ with boxes

This plots "exper.dat" with errorbars and lines connecting the points (errorbars require three or four columns):

     plot ’exper.dat’ w lines, ’exper.dat’ notitle w errorbars

Another way to plot "exper.dat" with errorlines (errorbars require three or four columns):

     plot ’exper.dat’ w errorlines

This plots sin(x) and cos(x) with linespoints, using the same line type but different point types:

     plot sin(x) with linesp lt 1 pt 3, cos(x) with linesp lt 1 pt 4

This plots file "data" with points of type 3 and twice usual size:

     plot ’data’ with points pointtype 3 pointsize 2

This plots file "data" with variable pointsize read from column 4

     plot ’data’ using 1:2:4 with points pt 5 pointsize variable

This plots two data sets with lines differing only by weight:

     plot ’d1’ t "good" w l lt 2 lw 3, ’d2’ t "bad" w l lt 2 lw 1

This plots filled curve of x*x and a color stripe:

     plot x*x with filledcurve closed, 40 with filledcurve y1=10

This plots x*x and a color box:

     plot x*x, (x>=-5 && x<=5 ? 40 : 1/0) with filledcurve y1=10 lt 8

This plots a surface with color lines:

     splot x*x-y*y with line palette

This plots two color surfaces at different altitudes:

     splot x*x-y*y with pm3d, x*x+y*y with pm3d at t

35 Print

The print command prints the value of <expression> to the screen. It is synonymous with pause 0. <expression> may be anything that gnuplot can evaluate that produces a number, or it can be a string.


     print <expression> {, <expression>, ...}

See expressions (p. 48). The output file can be set with set print.

36 Pwd

The pwd command prints the name of the working directory to the screen.

37 Quit

The exit and quit commands and END-OF-FILE character will exit gnuplot. Each of these commands will clear the output device (as does the clear command) before exiting.

38 Raise


     raise {plot_window_nb}

The raise command raises (opposite to lower) plot window(s) associated with the interactive terminal of your gnuplot session, i.e. pm, win, wxt or x11. It puts the plot window to front (top) in the z-order windows stack of the window manager of your desktop.

As x11 and wxt support multiple plot windows, then by default they raise these windows in descending order of most recently created on top to the least recently created on bottom. If a plot number is supplied as an optional parameter, only the associated plot window will be raised if it exists.

The optional parameter is ignored for single plot-windows terminal, i.e. pm and win.

If the window is not raised under X11, then (1) they don’t run in the same  
X11 session (telnet or ssh session, for example), or (2) raising is blocked  
by your window manager. On KDE, you may like to go to the KDE Control Center  
=> Desktop => Window Behaviour => Advanced and set the "Focus stealing  
prevention level" to None (default is Low).

39 Replot

The replot command without arguments repeats the last plot or splot command. This can be useful for viewing a plot with different set options, or when generating the same plot for several devices.

Arguments specified after a replot command will be added onto the last plot or splot command (with an implied ’,’ separator) before it is repeated. replot accepts the same arguments as the plot and splot commands except that ranges cannot be specified. Thus you can use replot to plot a function against the second axes if the previous command was plot but not if it was splot.

N.B. — use of

     plot ’-’ ; ... ; replot

is not recommended. gnuplot does not store the inline data internally, so since replot appends new information to the previous plot and then executes the modified command, the ’-’ from the initial plot will expect to read inline data again.

Note that replot does not work in multiplot mode, since it reproduces only the last plot rather than the entire screen.

See also command-line-editing (p. 40) for ways to edit the last plot (p. 149) (splot (p. 539)) command.

See also show plot (p. 390) to show the whole current plotting command, and the possibility to copy it into the history (p. 136).

40 Reread

The reread command causes the current gnuplot command file, as specified by a load command or on the command line, to be reset to its starting point before further commands are read from it. This essentially implements an endless loop of the commands from the beginning of the command file to the reread command. (But this is not necessarily a disaster — reread can be very useful when used in conjunction with if. See if (p. 137) for details.) The reread command has no effect if input from standard input.


Suppose the file "looper" contains the commands

     plot sin(x*a)  
     pause -1  
     if(a<5) reread

and from within gnuplot you submit the commands

     load ’looper’

The result will be five plots (separated by the pause message).

Suppose the file "data" contains six columns of numbers with a total yrange from 0 to 10; the first is x and the next are five different functions of x. Suppose also that the file "plotter" contains the commands

     c_p = c_p+1  
     plot "$0" using 1:c_p with lines linetype c_p  
     if(c_p <  n_p) reread

and from within gnuplot you submit the commands

     unset key  
     set yrange [0:10]  
     set multiplot  
     call ’plotter’ ’data’  
     unset multiplot

The result is a single graph consisting of five plots. The yrange must be set explicitly to guarantee that the five separate graphs (drawn on top of each other in multiplot mode) will have exactly the same axes. The linetype must be specified; otherwise all the plots would be drawn with the same type. See animate.dem in demo directory for an animated example.

41 Reset

The reset command causes all graph-related options that can be set with the set command to take on their default values. This command is useful, e.g., to restore the default graph settings at the end of a command file, or to return to a defined state after lots of settings have been changed within a command file. Please refer to the set command to see the default values that the various options take.

The following set commands do not change the graph status and are thus left unchanged: the terminal set with set term, the output file set with set output and directory paths set with set loadpath and set fontpath.

42 Save

The save command saves user-defined functions, variables, the set term status, all set options, or all of these, plus the last plot (splot) command to the specified file.


     save  {<option>} ’<filename>’

where <option> is functions, variables, terminal or set. If no option is used, gnuplot saves functions, variables, set options and the last plot (splot) command.

saved files are written in text format and may be read by the load command. For save with the set option or without any option, the terminal choice and the output filename are written out as a comment, to get an output file that works in other installations of gnuplot, without changes and without risk of unwillingly overwriting files.

save terminal will write out just the terminal status, without the comment marker in front of it. This is mainly useful for switching the terminal setting for a short while, and getting back to the previously set terminal, afterwards, by loading the saved terminal status. Note that for a single gnuplot session you may rather use the other method of saving and restoring current terminal by the commands set term push and set term pop, see set term (p. 489).

The filename must be enclosed in quotes.

The special filename "-" may be used to save commands to standard output. On systems which support a popen function (Unix), the output of save can be piped through an external program by starting the file name with a ’|’. This provides a consistent interface to gnuplot’s internal settings to programs which communicate with gnuplot through a pipe. Please see help for batch/interactive (p. 37) for more details.


     save ’work.gnu’  
     save functions ’func.dat’  
     save var ’var.dat’  
     save set ’options.dat’  
     save term ’myterm.gnu’  
     save ’-’  
     save ’|grep title >’

43 Set-show

The set command can be used to set lots of options. No screen is drawn, however, until a plot, splot, or replot command is given.

The show command shows their settings; show all shows all the settings.

Options changed using set can be returned to the default state by giving the corresponding unset command. See also the reset (p. 238) command, which returns all settable parameters to default values.

If a variable contains time/date data, show will display it according to the format currently defined by set timefmt, even if that was not in effect when the variable was initially defined.

43.1 Angles

By default, gnuplot assumes the independent variable in polar graphs is in units of radians. If set angles degrees is specified before set polar, then the default range is [0:360] and the independent variable has units of degrees. This is particularly useful for plots of data files. The angle setting also applies to 3-d mapping as set via the set mapping command.


     set angles {degrees | radians}  
     show angles

The angle specified in set grid polar is also read and displayed in the units specified by set angles.

set angles also affects the arguments of the machine-defined functions sin(x), cos(x) and tan(x), and the outputs of asin(x), acos(x), atan(x), atan2(x), and arg(x). It has no effect on the arguments of hyperbolic functions or Bessel functions. However, the output arguments of inverse hyperbolic functions of complex arguments are affected; if these functions are used, set angles radians must be in effect to maintain consistency between input and output arguments.

     set angles radians  
     print y         #prints {1.16933, 0.154051}  
     print asinh(y)  #prints {1.0, 0.1}


     set angles degrees  
     print y         #prints {1.16933, 0.154051}  
     print asinh(y)  #prints {57.29578, 5.729578}

See also

poldat.dem: polar plot using set angles demo.

43.2 Arrow

Arbitrary arrows can be placed on a plot using the set arrow command.


     set arrow {<tag>} {from <position>} {to|rto <position>}  
               { {arrowstyle | as <arrow_style>}  
                 | { {nohead | head | backhead | heads}  
                     {size <length>,<angle>{,<backangle>}}  
                     {filled | empty | nofilled}  
                     {front | back}  
                     { {linestyle | ls <line_style>}  
                       | {linetype | lt <line_type>}  
                         {linewidth | lw <line_width} } } }

     unset arrow {<tag>}  
     show arrow {<tag>}

<tag> is an integer that identifies the arrow. If no tag is given, the lowest unused tag value is assigned automatically. The tag can be used to delete or change a specific arrow. To change any attribute of an existing arrow, use the set arrow command with the appropriate tag and specify the parts of the arrow to be changed.

The <position>s are specified by either x,y or x,y,z, and may be preceded by first, second, graph, screen, or character to select the coordinate system. Unspecified coordinates default to 0. The end points can be specified in one of five coordinate systems — first or second axes, graph, screen, or character. See coordinates (p. 41) for details. A coordinate system specifier does not carry over from the "from" position to the "to" position. Arrows outside the screen boundaries are permitted but may cause device errors. If the end point is specified by "rto" instead of "to" it is drawn relatively to the start point. For linear axes, graph and screen coordinates, the distance between the start and the end point corresponds to the given relative coordinate. For logarithmic axes, the relative given coordinate corresponds to the factor of the coordinate between start and end point. Thus, a negative relative value or zero are not allowed for logarithmic axes.

Specifying nohead produces an arrow drawn without a head — a line segment. This gives you yet another way to draw a line segment on the plot. By default, an arrow has a head at its end. Specifying backhead draws an arrow head at the start point of the arrow while heads draws arrow heads on both ends of the line. Not all terminal types support double-ended arrows.

Head size can be controlled by size <length>,<angle> or size <length>,<angle>,<backangle>, where <length> defines length of each branch of the arrow head and <angle> the angle (in degrees) they make with the arrow. <Length> is in x-axis units; this can be changed by first, second, graph, screen, or character before the <length>; see coordinates (p. 41) for details. <Backangle> only takes effect when filled or empty is also used. Then, <backangle> is the angle (in degrees) the back branches make with the arrow (in the same direction as <angle>). The fig terminal has a restricted backangle function. It supports three different angles. There are two thresholds: Below 70 degrees, the arrow head gets an indented back angle. Above 110 degrees, the arrow head has an acute back angle. Between these thresholds, the back line is straight.

Specifying filled produces filled arrow heads (if heads are used). Filling is supported on filled-polygon capable terminals, see help of pm3d (p. 390) for their list, otherwise the arrow heads are closed but not filled. The same result (closed but not filled arrow head) is reached by specifying empty. Further, filling and outline is obviously not supported on terminals drawing arrows by their own specific routines, like metafont, metapost, latex or tgif.

The line style may be selected from a user-defined list of line styles (see set style line (p. 450)) or may be defined here by providing values for <line_type> (an index from the default list of styles) and/or <line_width> (which is a multiplier for the default width).

Note, however, that if a user-defined line style has been selected, its properties (type and width) cannot be altered merely by issuing another set arrow command with the appropriate index and lt or lw.

If front is given, the arrow is written on top of the graphed data. If back is given (the default), the arrow is written underneath the graphed data. Using front will prevent an arrow from being obscured by dense data.


To set an arrow pointing from the origin to (1,2) with user-defined style 5, use:

     set arrow to 1,2 ls 5

To set an arrow from bottom left of plotting area to (-5,5,3), and tag the arrow number 3, use:

     set arrow 3 from graph 0,0 to -5,5,3

To change the preceding arrow to end at 1,1,1, without an arrow head and double its width, use:

     set arrow 3 to 1,1,1 nohead lw 2

To draw a vertical line from the bottom to the top of the graph at x=3, use:

     set arrow from 3, graph 0 to 3, graph 1 nohead

To draw a vertical arrow with T-shape ends, use:

     set arrow 3 from 0,-5 to 0,5 heads size screen 0.1,90

To draw an arrow relatively to the start point, where the relative distances are given in graph coordinates, use:

     set arrow from 0,-5 rto graph 0.1,0.1

To draw an arrow with relative end point in logarithmic x axis, use:

     set logscale x  
     set arrow from 100,-5 rto 10,10

This draws an arrow from 100,-5 to 1000,5. For the logarithmic x axis, the relative coordinate 10 means "factor 10" while for the linear y axis, the relative coordinate 10 means "difference 10".

To delete arrow number 2, use:

     unset arrow 2

To delete all arrows, use:

     unset arrow

To show all arrows (in tag order), use:

     show arrow

arrows demos.

43.3 Autoscale

Autoscaling may be set individually on the x, y or z axis or globally on all axes. The default is to autoscale all axes.


     set autoscale {<axes>{|min|max|fixmin|fixmax|fix} | fix | keepfix}  
     unset autoscale {<axes>}  
     show autoscale

where <axes> is either x, y, z, cb, x2, y2 or xy. A keyword with min or max appended (this cannot be done with xy) tells gnuplot to autoscale just the minimum or maximum of that axis. If no keyword is given, all axes are autoscaled.

A keyword with fixmin, fixmax or fix appended tells gnuplot to disable extension of the axis range to the next tic mark position, for autoscaled axes using equidistant tics; set autoscale fix sets this for all axes. Command set autoscale keepfix autoscales all axes while keeping the fix settings.

When autoscaling, the axis range is automatically computed and the dependent axis (y for a plot and z for splot) is scaled to include the range of the function or data being plotted.

If autoscaling of the dependent axis (y or z) is not set, the current y or z range is used.

Autoscaling the independent variables (x for plot and x,y for splot) is a request to set the domain to match any data file being plotted. If there are no data files, autoscaling an independent variable has no effect. In other words, in the absence of a data file, functions alone do not affect the x range (or the y range if plotting z = f(x,y)).

Please see set xrange (p. 508) for additional information about ranges.

The behavior of autoscaling remains consistent in parametric mode, (see set parametric (p. 389)). However, there are more dependent variables and hence more control over x, y, and z axis scales. In parametric mode, the independent or dummy variable is t for plots and u,v for splots. autoscale in parametric mode, then, controls all ranges (t, u, v, x, y, and z) and allows x, y, and z to be fully autoscaled.

Autoscaling works the same way for polar mode as it does for parametric mode for plot, with the extension that in polar mode set dummy can be used to change the independent variable from t (see set dummy (p. 314)).

When tics are displayed on second axes but no plot has been specified for those axes, x2range and y2range are inherited from xrange and yrange. This is done before xrange and yrange are autoextended to a whole number of tics, which can cause unexpected results. You can use the fixmin or fixmax options to avoid this.


This sets autoscaling of the y axis (other axes are not affected):

     set autoscale y

This sets autoscaling only for the minimum of the y axis (the maximum of the y axis and the other axes are not affected):

     set autoscale ymin

This disables extension of the x2 axis tics to the next tic mark, thus keeping the exact range as found in the plotted data and functions:

     set autoscale x2fixmin  
     set autoscale x2fixmax

This sets autoscaling of the x and y axes:

     set autoscale xy

This sets autoscaling of the x, y, z, x2 and y2 axes:

     set autoscale

This disables autoscaling of the x, y, z, x2 and y2 axes:

     unset autoscale

This disables autoscaling of the z axis only:

     unset autoscale z

43.3.1 Parametric mode

When in parametric mode (set parametric), the xrange is as fully scalable as the y range. In other words, in parametric mode the x axis can be automatically scaled to fit the range of the parametric function that is being plotted. Of course, the y axis can also be automatically scaled just as in the non-parametric case. If autoscaling on the x axis is not set, the current x range is used.

Data files are plotted the same in parametric and non-parametric mode. However, there is a difference in mixed function and data plots: in non-parametric mode with autoscaled x, the x range of the datafile controls the x range of the functions; in parametric mode it has no influence.

For completeness a last command set autoscale t is accepted. However, the effect of this "scaling" is very minor. When gnuplot determines that the t range would be empty, it makes a small adjustment if autoscaling is true. Otherwise, gnuplot gives an error. Such behavior may, in fact, not be very useful and the command set autoscale t is certainly questionable.

splot extends the above ideas as you would expect. If autoscaling is set, then x, y, and z ranges are computed and each axis scaled to fit the resulting data.

43.3.2 Polar mode

When in polar mode (set polar), the xrange and the yrange are both found from the polar coordinates, and thus they can both be automatically scaled. In other words, in polar mode both the x and y axes can be automatically scaled to fit the ranges of the polar function that is being plotted.

When plotting functions in polar mode, the rrange may be autoscaled. When plotting data files in polar mode, the trange may also be autoscaled. Note that if the trange is contained within one quadrant, autoscaling will produce a polar plot of only that single quadrant.

Explicitly setting one or two ranges but not others may lead to unexpected results. See also

polar demos.

43.4 Bars

The set bars command controls the tics at the ends of error bars, and also the width of the boxes in plot styles candlesticks and financebars.


     set bars {small | large | fullwidth | <size>}  
     unset bars  
     show bars

small is a synonym for 0.0, and large for 1.0. The default is 1.0 if no size is given.

The keyword fullwidth is relevant only to histograms with errorbars. It sets the width of the errorbar ends to be the same as the width of the associated box in the histogram. It does not change the width of the box itself.

43.5 Bmargin

The command set bmargin sets the size of the bottom margin. Please see set margin (p. 375) for details.

43.6 Border

The set border and unset border commands control the display of the graph borders for the plot and splot commands. Note that the borders do not necessarily coincide with the axes; with plot they often do, but with splot they usually do not.


     set border {<integer>} {front | back} {linewidth | lw <line_width>}  
                {{linestyle | ls <line_style>} | {linetype | lt <line_type>}}  
     unset border  
     show border

With a splot displayed in an arbitrary orientation, like set view 56,103, the four corners of the x-y plane can be referred to as "front", "back", "left" and "right". A similar set of four corners exist for the top surface, of course. Thus the border connecting, say, the back and right corners of the x-y plane is the "bottom right back" border, and the border connecting the top and bottom front corners is the "front vertical". (This nomenclature is defined solely to allow the reader to figure out the table that follows.)

The borders are encoded in a 12-bit integer: the bottom four bits control the border for plot and the sides of the base for splot; the next four bits control the verticals in splot; the top four bits control the edges on top of the splot. In detail, <integer> should be the sum of the appropriate entries from the following table:

Graph Border Encoding

Bit plot splot

1 bottom bottom left front
2 left bottom left back
4 top bottom right front
8 right bottom right back
16 no effect left vertical
32 no effect back vertical
64 no effect right vertical
128 no effect front vertical
256 no effect top left back
512 no effect top right back
1024no effect top left front
2048no effect top right front

Various bits or combinations of bits may be added together in the command.

The default is 31, which is all four sides for plot, and base and z axis for splot.

In 2D plots the border is normally drawn on top of all plots elements (front). If you want the border to be drawn behind the plot elements, use set border back.

Using the optional <line_style>, <line_type> and <line_width> specifiers, the way the border lines are drawn can be influenced (limited by what the current terminal driver supports).

For plot, tics may be drawn on edges other than bottom and left by enabling the second axes – see set xtics (p. 514) for details.

If a splot draws only on the base, as is the case with "unset surface; set contour base", then the verticals and the top are not drawn even if they are specified.

The set grid options ’back’, ’front’ and ’layerdefault’ also control the order in which the border lines are drawn with respect to the output of the plotted data.


Draw default borders:

     set border

Draw only the left and bottom (plot) or both front and back bottom left (splot) borders:

     set border 3

Draw a complete box around a splot:

     set border 4095

Draw a topless box around a splot, omitting the front vertical:

     set border 127+256+512 # or set border 1023-128

Draw only the top and right borders for a plot and label them as axes:

     unset xtics; unset ytics; set x2tics; set y2tics; set border 12

43.7 Boxwidth

The set boxwidth command is used to set the default width of boxes in the boxes, boxerrorbars, candlesticks and histograms styles.


     set boxwidth {<width>} {absolute|relative}  
     show boxwidth

By default, adjacent boxes are extended in width until they touch each other. A different default width may be specified using the set boxwidth command. Relative widths are interpreted as being a fraction of this default width.

An explicit value for the boxwidth is interpreted as being a number of units along the current x axis (absolute) unless the modifier relative is given. If the x axis is a log-scale (see set log (p. 366)) then the value of boxwidth is truly "absolute" only at x=1; this physical width is maintained everywhere along the axis (i.e. the boxes do not become narrower the value of x increases). If the range spanned by a log scale x axis is far from x=1, some experimentation may be required to find a useful value of boxwidth.

The default is superseded by explicit width information taken from an extra data column in styles boxes or boxerrorbars. In a four-column data set, the fourth column will be interpreted as the box width unless the width is set to -2.0, in which case the width will be calculated automatically. See style boxes (p. 458) and style boxerrorbars (p. 458) for more details.

To set the box width to automatic use the command

     set boxwidth

or, for four-column data,

     set boxwidth -2

The same effect can be achieved with the using keyword in plot:

     plot ’file’ using 1:2:3:4:(-2)

To set the box width to half of the automatic size use

     set boxwidth 0.5 relative

To set the box width to an absolute value of 2 use

     set boxwidth 2 absolute

43.8 Clabel

gnuplot will vary the linetype used for each contour level when clabel is set. When this option on (the default), a legend labels each linestyle with the z level it represents. It is not possible at present to separate the contour labels from the surface key.


     set clabel {’<format>’}  
     unset clabel  
     show clabel

The default for the format string is %8.3g, which gives three decimal places. This may produce poor label alignment if the key is altered from its default configuration.

The first contour linetype, or only contour linetype when clabel is off, is the surface linetype +1; contour points are the same style as surface points.

See also set contour (p. 291).

43.9 Clip

gnuplot can clip data points and lines that are near the boundaries of a graph.


     set clip <clip-type>  
     unset clip <clip-type>  
     show clip

Three clip types for points and lines are supported by gnuplot: points, one, and two. One, two, or all three clip types may be active for a single graph. Note that clipping of color filled quadrangles drawn by pm3d maps and surfaces is not controlled by this command, but by set pm3d clip1in and set pm3d clip4in.

The points clip type forces gnuplot to clip (actually, not plot at all) data points that fall within but too close to the boundaries. This is done so that large symbols used for points will not extend outside the boundary lines. Without clipping points near the boundaries, the plot may look bad. Adjusting the x and y ranges may give similar results.

Setting the one clip type causes gnuplot to draw a line segment which has only one of its two endpoints within the graph. Only the in-range portion of the line is drawn. The alternative is to not draw any portion of the line segment.

Some lines may have both endpoints out of range, but pass through the graph. Setting the two clip-type allows the visible portion of these lines to be drawn.

In no case is a line drawn outside the graph.

The defaults are noclip points, clip one, and noclip two.

To check the state of all forms of clipping, use

     show clip

For backward compatibility with older versions, the following forms are also permitted:

     set clip  
     unset clip

set clip is synonymous with set clip points; unset clip turns off all three types of clipping.

43.10 Cntrparam

set cntrparam controls the generation of contours and their smoothness for a contour plot. show contour displays current settings of cntrparam as well as contour.


     set cntrparam { { linear  
                     | cubicspline  
                     | bspline  
                     | points <n>  
                     | order <n>  
                     | levels { auto {<n>} | <n>  
                                | discrete <z1> {,<z2>{,<z3>...}}  
                                | incremental <start>, <incr> {,<end>}  
     show contour

This command has two functions. First, it sets the values of z for which contour points are to be determined (by linear interpolation between data points or function isosamples.) Second, it controls the way contours are drawn between the points determined to be of equal z. <n> should be an integral constant expression and <z1>, <z2> ... any constant expressions. The parameters are:

linear, cubicspline, bspline — Controls type of approximation or interpolation. If linear, then straight line segments connect points of equal z magnitude. If cubicspline, then piecewise-linear contours are interpolated between the same equal z points to form somewhat smoother contours, but which may undulate. If bspline, a guaranteed-smoother curve is drawn, which only approximates the position of the points of equal-z.

points — Eventually all drawings are done with piecewise-linear strokes. This number controls the number of line segments used to approximate the bspline or cubicspline curve. Number of cubicspline or bspline segments (strokes) = points * number of linear segments.

order — Order of the bspline approximation to be used. The bigger this order is, the smoother the resulting contour. (Of course, higher order bspline curves will move further away from the original piecewise linear data.) This option is relevant for bspline mode only. Allowed values are integers in the range from 2 (linear) to 10.

levels — Selection of contour levels, controlled by auto (default), discrete, incremental, and <n>, number of contour levels.

For auto, <n> specifies a nominal number of levels; the actual number will be adjusted to give simple labels. If the surface is bounded by zmin and zmax, contours will be generated at integer multiples of dz between zmin and zmax, where dz is 1, 2, or 5 times some power of ten (like the step between two tic marks).

For levels discrete, contours will be generated at z = <z1>, <z2> ... as specified; the number of discrete levels sets the number of contour levels. In discrete mode, any set cntrparam levels <n> are ignored.

For incremental, contours are generated at values of z beginning at <start> and increasing by <increment>, until the number of contours is reached. <end> is used to determine the number of contour levels, which will be changed by any subsequent set cntrparam levels <n>. If the z axis is logarithmic, <increment> will be interpreted as a factor, just like in set ztics.

If the command set cntrparam is given without any arguments specified, the defaults are used: linear, 5 points, order 4, 5 auto levels.


     set cntrparam bspline  
     set cntrparam points 7  
     set cntrparam order 10

To select levels automatically, 5 if the level increment criteria are met:

     set cntrparam levels auto 5

To specify discrete levels at .1, .37, and .9:

     set cntrparam levels discrete .1,1/exp(1),.9

To specify levels from 0 to 4 with increment 1:

     set cntrparam levels incremental  0,1,4

To set the number of levels to 10 (changing an incremental end or possibly the number of auto levels):

     set cntrparam levels 10

To set the start and increment while retaining the number of levels:

     set cntrparam levels incremental 100,50

See also set contour (p. 291) for control of where the contours are drawn, and set clabel (p. 277) for control of the format of the contour labels and linetypes.

See also

contours demo (contours.dem)


contours with user defined levels demo (discrete.dem).

43.11 Color box

The color scheme, i.e. the gradient of the smooth color with min_z and max_z values of pm3d’s palette, is drawn in a color box unless unset colorbox.

     set colorbox  
     set colorbox {  
                { vertical | horizontal }  
                { default | user }  
                { origin x, y }  
                { size x, y }  
                { front | back }  
                { noborder | bdefault | border [line style] }  
     show colorbox  
     unset colorbox

Color box position can be default or user. If the latter is specified the values as given with the origin and size subcommands are used. The box can be drawn after (front) or before (back) the graph or the surface.

The orientation of the color gradient can be switched by options vertical and horizontal.

origin x, y and size x, y are used only in combination with the user option. The x and y values are interpreted as screen coordinates by default, and this is the only legal option for 3D plots. 2D plots, including splot with set view map, allow any coordinate system to be specified. Try for example:

   set colorbox horiz user origin .1,.02 size .8,.04

which will draw a horizontal gradient somewhere at the bottom of the graph.

border turns the border on (this is the default). noborder turns the border off. If an positive integer argument is given after border, it is used as a line style tag which is used for drawing the border, e.g.:

   set style line 2604 linetype -1 linewidth .4  
   set colorbox border 2604

will use line style 2604, a thin line with the default border color (-1) for drawing the border. bdefault (which is the default) will use the default border line style for drawing the border of the color box.

The axis of the color box is called cb and it is controlled by means of the usual axes commands, i.e. set/unset/show with cbrange, [m]cbtics, format cb, grid [m]cb, cblabel, and perhaps even cbdata, [no]cbdtics, [no]cbmtics.

set colorbox without any parameter switches the position to default. unset colorbox resets the default parameters for the colorbox and switches the colorbox off.

See also help for set pm3d (p. 390), set palette (p. 400), x11 pm3d (p. 752), and set style line (p. 450).

43.12 Contour

set contour enables contour drawing for surfaces. This option is available for splot only. It requires grid data, see grid_data (p. 558) for more details. If contours are desired from non-grid data, set dgrid3d can be used to create an appropriate grid.


     set contour {base | surface | both}  
     unset contour  
     show contour

The three options specify where to draw the contours: base draws the contours on the grid base where the x/ytics are placed, surface draws the contours on the surfaces themselves, and both draws the contours on both the base and the surface. If no option is provided, the default is base.

See also set cntrparam (p. 281) for the parameters that affect the drawing of contours, and set clabel (p. 277) for control of labelling of the contours.

The surface can be switched off (see set surface (p. 487)), giving a contour-only graph. Though it is possible to use set size to enlarge the plot to fill the screen, more control over the output format can be obtained by writing the contour information to a file, and rereading it as a 2-d datafile plot:

     unset surface  
     set contour  
     set cntrparam ...  
     set table ’filename’  
     splot ...  
     unset table  
     # contour info now in filename  
     set term <whatever>  
     plot ’filename’

In order to draw contours, the data should be organized as "grid data". In such a file all the points for a single y-isoline are listed, then all the points for the next y-isoline, and so on. A single blank line (a line containing no characters other than blank spaces and a carriage return and/or a line feed) separates one y-isoline from the next. See also splot datafile (p. 540).

See also

contours demo (contours.dem)


contours with user defined levels demo (discrete.dem).

43.13 Data style

This form of the command is deprecated. Please see set style data (p. 444).

43.14 Datafile

The set datafile command options control interpretation of fields read from input data files by the plot, splot, and fit commands. Four such options are currently implemented.

43.14.1 Set datafile fortran

The set datafile fortran command enables a special check for values in the input file expressed as Fortran D or Q constants. This extra check slows down the input process, and should only be selected if you do in fact have datafiles containing Fortran D or Q constants. The option can be disabled again using unset datafile fortran.

43.14.2 Set datafile missing

The set datafile missing command allows you to tell gnuplot what character string is used in a data file to denote missing data. Exactly how this missing value will be treated depends on the using specifier of the plot or splot command.


     set datafile missing {"<string>"}  
     show datafile missing  
     unset datafile


     # Ignore entries containing IEEE NaN ("Not a Number") code  
     set datafile missing "NaN"


     set datafile missing "?"  
     set style data lines  
     plot ’-’  
        1 10  
        2 20  
        3 ?  
        4 40  
        5 50  
     plot ’-’ using 1:2  
        1 10  
        2 20  
        3 ?  
        4 40  
        5 50  
     plot ’-’ using 1:($2)  
        1 10  
        2 20  
        3 ?  
        4 40  
        5 50  

The first plot will recognize only the first datum in the "3 ?" line. It will use the single-datum-on-a-line convention that the line number is "x" and the datum is "y", so the point will be plotted (in this case erroneously) at (2,3).

The second plot will correctly ignore the middle line. The plotted line will connect the points at (2,20) and (4,40).

The third plot will also correctly ignore the middle line, but the plotted line will not connect the points at (2,20) and (4,40).

There is no default character for missing, but in many cases any non-parsible string of characters found where a numerical value is expected will be treated as missing data.

43.14.3 Set datafile separator

The command set datafile separator "<char>" tells gnuplot that data fields in subsequent input files are separated by <char> rather than by whitespace. The most common use is to read in csv (comma-separated value) files written by spreadsheet or database programs. By default data fields are separated by whitespace.


     set datafile separator {"<char>" | whitespace}


     # Input file contains tab-separated fields  
     set datafile separator "\t"

     # Input file contains comma-separated values fields  
     set datafile separator ","

43.14.4 Set datafile commentschars

The set datafile commentschars command allows you to tell gnuplot what characters are used in a data file to denote comments. Gnuplot will ignore rest of the line behind the specified characters if either of them is the first non-blank character on the line.


     set datafile commentschars {"<string>"}  
     show datafile commentschars  
     unset commentschars

Default value of the string is "#!" on VMS and "#" otherwise.

Then, the following line in a data file is completely ignored

   # 1 2 3 4

but the following

   1 # 3 4

produces rather unexpected plot unless

   set datafile missing ’#’

is specified as well.


     set datafile commentschars "#!%"

43.14.5 Set datafile binary

The set datafile binary command is used to set the defaults when reading binary data files. The syntax matches precisely that used for commands plot and splot. See binary (p. 154) for details about <binary list>.


     set datafile binary <binary list>  
     show datafile binary  
     show datafile  
     unset datafile


     set datafile binary filetype=auto  
     set datafile binary array=512x512 format="%uchar"

43.15 Decimalsign

The set decimalsign command selects a decimal sign for numbers printed into tic labels or set label strings.


     set decimalsign {<value> | locale {"<locale>"}}  
     unset decimalsign  
     show decimalsign

The argument <value> is a string to be used in place of the usual decimal point. Typical choices include the period, ’.’, and the comma, ’,’, but others may be useful, too. If you omit the <value> argument, the decimal separator is not modified from the usual default, which is a period. Unsetting decimalsign has the same effect as omitting <value>.


Correct typesetting in most European countries requires:

     set decimalsign ’,’

Please note: If you set an explicit string, this affects only numbers that are printed using gnuplot’s gprintf() formatting routine, include axis tics. It does not affect the format expected for input data, and it does not affect numbers printed with the sprintf() formatting routine. To change the behavior of both input and output formatting, instead use the form

     set decimalsign locale

This instructs the program to use both input and output formats in accordance with the current setting of the LC_ALL, LC_NUMERIC, or LANG environmental variables.

     set decimalsign locale "foo"

This instructs the program to format all input and output in accordance with locale "foo", which must be installed. If locale "foo" is not found then an error message is printed and the decimal sign setting is unchanged. On linux systems you can get a list of the locales installed on your machine by typing "locale -a". A typical linux locale string is of the form "sl_SI.UTF-8". A typical Windows locale string is of the form "Slovenian_Slovenia.1250" or "slovenian". Please note that interpretation of the locale settings is done by the C library at runtime. Older C libraries may offer only partial support for locale settings such as the thousands grouping separator character.

     set decimalsign locale; set decimalsign "."

This sets all input and output to use whatever decimal sign is correct for the current locale, but over-rides this with an explicit ’.’ in numbers formatted using gnuplot’s internal gprintf() function.

43.16 Dgrid3d

The set dgrid3d command enables, and can set parameters for, non-grid to grid data mapping. See splot grid_data (p. 558) for more details about the grid data structure.


     set dgrid3d {<row_size>} {,{<col_size>} {,<norm>}}  
     unset dgrid3d  
     show dgrid3d

By default dgrid3d is disabled. When enabled, 3-d data read from a file are always treated as a scattered data set. A grid with dimensions derived from a bounding box of the scattered data and size as specified by the row/col_size parameters is created for plotting and contouring. The grid is equally spaced in x (rows) and in y (columns); the z values are computed as weighted averages of the scattered points’ z values.

The third parameter, norm, controls the weighting: Each data point is weighted inversely by its distance from the grid point raised to the norm power. (Actually, the weights are given by the inverse of dx^norm + dy^norm, where dx and dy are the components of the separation of the grid point from each data point. For some norms that are powers of two, specifically 4, 8, and 16, the computation is optimized by using the Euclidean distance in the weight calculation, (dx^2+dy^2)^norm/2. However, any non-negative integer can be used.)

The closer the data point is to a grid point, the more effect it has on that grid point and the larger the value of norm the less effect more distant data points have on that grid point.

The dgrid3d option is a simple low pass filter that converts scattered data to a grid data set. More sophisticated approaches to this problem exist and should be used to preprocess the data outside gnuplot if this simple solution is found inadequate.

(The z values are found by weighting all data points, not by interpolating between nearby data points; also edge effects may produce unexpected and/or undesired results. In some cases, small norm values produce a grid point reflecting the average of distant data points rather than a local average, while large values of norm may produce "steps" with several grid points having the same value as the closest data point, rather than making a smooth transition between adjacent data points. Some areas of a grid may be filled by extrapolation, to an arbitrary boundary condition. The variables are not normalized; consequently the units used for x and y will affect the relative weights of points in the x and y directions.)


     set dgrid3d 10,10,1     # defaults  
     set dgrid3d ,,4

The first specifies that a grid of size 10 by 10 is to be constructed using a norm value of 1 in the weight computation. The second only modifies the norm, changing it to 4. See also

scatter.dem: dgrid3d demo.

43.17 Dummy

The set dummy command changes the default dummy variable names.


     set dummy {<dummy-var>} {,<dummy-var>}  
     show dummy

By default, gnuplot assumes that the independent, or "dummy", variable for the plot command is "t" if in parametric or polar mode, or "x" otherwise. Similarly the independent variables for the splot command are "u" and "v" in parametric mode (splot cannot be used in polar mode), or "x" and "y" otherwise.

It may be more convenient to call a dummy variable by a more physically meaningful or conventional name. For example, when plotting time functions:

     set dummy t  
     plot sin(t), cos(t)

At least one dummy variable must be set on the command; set dummy by itself will generate an error message.


     set dummy u,v  
     set dummy ,s

The second example sets the second variable to s.

43.18 Encoding

The set encoding command selects a character encoding. Syntax:

     set encoding {<value>}  
     show encoding

Valid values are

  default     - tells a terminal to use its default encoding  
  iso_8859_1  - the most common Western European font used by many  
                Unix workstations and by MS-Windows. This encoding is  
                known in the PostScript world as ’ISO-Latin1’.  
  iso_8859_2  - used in Central and Eastern Europe  
  iso_8859_15 - a variant of iso_8859_1 that includes the Euro symbol  
  koi8r       - popular Unix cyrillic encoding  
  koi8u       - ukrainian Unix cyrillic encoding  
  cp437       - codepage for MS-DOS  
  cp850       - codepage for OS/2, Western Europe  
  cp852       - codepage for OS/2, Central and Eastern Europe  
  cp1250      - codepage for MS Windows, Central and Eastern Europe

Generally you must set the encoding before setting the terminal type. Note that encoding is not supported by all terminal drivers and that the device must be able to produce the desired non-standard characters. The PostScript, X11 and wxt terminals support all encodings. OS/2 Presentation Manager switches automatically to codepage 912 for iso_8859_2.

43.19 Fit

The fit setting defines where the fit command writes its output. If this option was built into your version of gnuplot, it also controls whether parameter errors from the fit will be written into variables.


     set fit {logfile {"<filename>"}} {{no}errorvariables}  
     unset fit  
     show fit

The <filename> argument must be enclosed in single or double quotes.

If no filename is given or unset fit is used the log file is reset to its default value "fit.log" or the value of the environmental variable FIT_LOG.

Users of DOS-like platforms should note that the \ character has special significance in double-quoted strings, so single-quotes should be used for filenames in different directories, or you have to write \\ for each \. Or you can just use forward slashes, even though this is DOS.

If the given logfile name ends with a / or \, it is interpreted to be a directory name, and the actual filename will be "fit.log" in that directory.

If the errorvariables option is turned on, the error of each fitted parameter computed by fit will be copied to a user-defined variable whose name is formed by appending "_err" to the name of the parameter itself. This is useful mainly to put the parameter and its error onto a plot of the data and the fitted function, for reference, as in:

      set fit errorvariables  
      fit f(x) ’datafile’ using 1:2 via a, b  
      print "error of a is:", a_err  
      set label ’a=%6.2f’, a, ’+/- %6.2f’, a_err  
      plot ’datafile’ using 1:2, f(x)

43.20 Fontpath

The fontpath setting defines additional locations for font files searched when including font files. Currently only the postscript terminal supports fontpath. If a file cannot be found in the current directory, the directories in fontpath are tried. Further documentation concerning the supported file formats is included in the terminal postscript section of the documentation.


     set fontpath {"pathlist1" {"pathlist2"...}}  
     show fontpath

Path names may be entered as single directory names, or as a list of path names separated by a platform-specific path separator, eg. colon (’:’) on Unix, semicolon (’;’) on DOS/Windows/OS/2/Amiga platforms. The show fontpath, save and save set commands replace the platform-specific separator with a space character (’ ’) for maximum portability. If a directory name ends with an exclamation mark (’!’) also the subdirectories of this directory are searched for font files.

If the environmental variable GNUPLOT_FONTPATH is set, its contents are appended to fontpath. If it is not set, a system dependent default value is used. It is set by testing several directories for existence when using the fontpath the first time. Thus, the first call of set fontpath, show fontpath, save fontpath, plot, or splot with embedded font files takes a little more time. If you want to save this time you may set the environmental variable GNUPLOT_FONTPATH since probing is switched off, then. You can find out which is the default fontpath by using show fontpath.

However, show fontpath prints the contents of user defined fontpath and system fontpath separately. Also, the save and save set commands save only the user specified parts of fontpath, for portability reasons.

Many other terminal drivers access TrueType fonts via the gd library. For these drivers the font search path is controlled by the environmental variable GDFONTPATH.

43.21 Format

The format of the tic-mark labels can be set with the set format command or with the set tics format or individual set {axis}tics format commands.


     set format {<axes>} {"<format-string>"}  
     set format {<axes>} {’<format-string>’}  
     show format

where <axes> is either x, y, xy, x2, y2, z, cb or nothing (which refers to all axes at once). The length of the string representing a tic mark (after formatting with ’printf’) is restricted to 100 characters. If the format string is omitted, the format will be returned to the default "% g". For LaTeX users, the format "$%g$" is often desirable. If the empty string "" is used, no label will be plotted with each tic, though the tic mark will still be plotted. To eliminate all tic marks, use unset xtics or unset ytics.

Newline (\n) is accepted in the format string. Use double-quotes rather than single-quotes to enable such interpretation. See also syntax (p. 101).

The default format for both axes is "% g", but other formats such as "%.2f" or "%3.0em" are often desirable. Anything accepted by ’printf’ when given a double precision number, and accepted by the terminal, will work. Some other options have been added. If the format string looks like a floating point format, then gnuplot tries to construct a reasonable format.

Characters not preceded by "%" are printed verbatim. Thus you can include spaces and labels in your format string, such as "%g m", which will put " m" after each number. If you want "%" itself, double it: "%g %%".

See also set xtics (p. 514) for more information about tic labels, and set decimalsign (p. 307) for how to use non-default decimal separators in numbers printed this way. See also

electron demo (electron.dem).

43.21.1 Gprintf

The string function gprintf("format",x) uses gnuplot’s own format specifiers, as do the gnuplot commands set format, set timestamp, and others. These format specifiers are not the same as those used by the standard C-language routine sprintf(). Gnuplot also provides an sprintf("format",x,...) routine if you prefer. For a list of gnuplot’s format options, see format specifiers (p. 323).

43.21.2 Format specifiers

The acceptable formats (if not in time/date mode) are:

Tic-mark label numerical format specifiers

Format Explanation

%f floating point notation
%e or %Eexponential notation; an ”e” or ”E” before the power
%g or %Gthe shorter of %e (or %E) and %f
%x or %Xhex
%o or %Ooctal
%t mantissa to base 10
%l mantissa to base of current logscale
%s mantissa to base of current logscale; scientific power
%T power to base 10
%L power to base of current logscale
%S scientific power
%c character replacement for scientific power
%P multiple of pi

A ’scientific’ power is one such that the exponent is a multiple of three. Character replacement of scientific powers ("%c") has been implemented for powers in the range -18 to +18. For numbers outside of this range the format reverts to exponential.

Other acceptable modifiers (which come after the "%" but before the format specifier) are "-", which left-justifies the number; "+", which forces all numbers to be explicitly signed; " " (a space), which makes positive numbers have a space in front of them where negative numbers have "-"; "#", which places a decimal point after floats that have only zeroes following the decimal point; a positive integer, which defines the field width; "0" (the digit, not the letter) immediately preceding the field width, which indicates that leading zeroes are to be used instead of leading blanks; and a decimal point followed by a non-negative integer, which defines the precision (the minimum number of digits of an integer, or the number of digits following the decimal point of a float).

Some systems may not support all of these modifiers but may also support others; in case of doubt, check the appropriate documentation and then experiment.


     set format y "%t"; set ytics (5,10)          # "5.0" and "1.0"  
     set format y "%s"; set ytics (500,1000)      # "500" and "1.0"  
     set format y "%+-12.3f"; set ytics(12345)    # "+12345.000  "  
     set format y "%.2t*10^%+03T"; set ytic(12345)# "1.23*10^+04"  
     set format y "%s*10^{%S}"; set ytic(12345)   # "12.345*10^{3}"  
     set format y "%s %cg"; set ytic(12345)       # "12.345 kg"  
     set format y "%.0P pi"; set ytic(6.283185)   # "2 pi"  
     set format y "%.0f%%"; set ytic(50)          # "50%"

     set log y 2; set format y ’%l’; set ytics (1,2,3)  
     #displays "1.0", "1.0" and "1.5" (since 3 is 1.5 * 2^1)

There are some problem cases that arise when numbers like 9.999 are printed with a format that requires both rounding and a power.

If the data type for the axis is time/date, the format string must contain valid codes for the ’strftime’ function (outside of gnuplot, type "man strftime"). See set timefmt (p. 495) for a list of the allowed input format codes.

43.21.3 Time/date specifiers

In time/date mode, the acceptable formats are:

Tic-mark label Date/Time Format Specifiers

Format Explanation

%a abbreviated name of day of the week
%A full name of day of the week
%b or %habbreviated name of the month
%B full name of the month
%d day of the month, 1–31
%D shorthand for "%m/%d/%y" (only output)
%F shorthand for "%Y-%m-%d" (only output)
%k hour, 0–23 (one or two digits)
%H hour, 00–23 (always two digits)
%l hour, 1–12 (one or two digits)
%I hour, 01–12 (always two digits)
%j day of the year, 1–366
%m month, 1–12
%M minute, 0–60
%p ”am” or ”pm”
%r shorthand for "%I:%M:%S %p" (only output)
%R shorthand for %H:%M" (only output)
%S second, 0–60
%T shorthand for "%H:%M:%S" (only output)
%U week of the year (week starts on Sunday)
%w day of the week, 0–6 (Sunday = 0)
%W week of the year (week starts on Monday)
%y year, 0-99
%Y year, 4-digit

Except for the non-numerical formats, these may be preceded by a "0" ("zero", not "oh") to pad the field length with leading zeroes, and a positive digit, to define the minimum field width (which will be overridden if the specified width is not large enough to contain the number). There is a 24-character limit to the length of the printed text; longer strings will be truncated.


Suppose the text is "76/12/25 23:11:11". Then

     set format x                 # defaults to "12/25/76" \n "23:11"  
     set format x "%A, %d %b %Y"  # "Saturday, 25 Dec 1976"  
     set format x "%r %D"         # "11:11:11 pm 12/25/76"

Suppose the text is "98/07/06 05:04:03". Then

     set format x "%1y/%2m/%3d %01H:%02M:%03S"  # "98/ 7/  6 5:04:003"

43.22 Function style

This form of the command is deprecated. Please see set style function (p. 446).

43.23 Functions

The show functions command lists all user-defined functions and their definitions.


     show functions

For information about the definition and usage of functions in gnuplot, please see expressions (p. 48). See also

splines as user defined functions (spline.dem)


use of functions and complex variables for airfoils (airfoil.dem).

43.24 Grid

The set grid command allows grid lines to be drawn on the plot.


     set grid {{no}{m}xtics} {{no}{m}ytics} {{no}{m}ztics}  
              {{no}{m}x2tics} {{no}{m}y2tics}  
              {polar {<angle>}}  
              {layerdefault | front | back}  
              { {linestyle <major_linestyle>}  
                | {linetype | lt <major_linetype>}  
                  {linewidth | lw <major_linewidth>}  
                { , {linestyle | ls <minor_linestyle>}  
                    | {linetype | lt <minor_linetype>}  
                      {linewidth | lw <minor_linewidth>} } }  
     unset grid  
     show grid

The grid can be enabled and disabled for the major and/or minor tic marks on any axis, and the linetype and linewidth can be specified for major and minor grid lines, also via a predefined linestyle, as far as the active terminal driver supports this.

Additionally, a polar grid can be selected for 2-d plots — circles are drawn to intersect the selected tics, and radial lines are drawn at definable intervals. (The interval is given in degrees or radians, depending on the set angles setting.) Note that a polar grid is no longer automatically generated in polar mode.

The pertinent tics must be enabled before set grid can draw them; gnuplot will quietly ignore instructions to draw grid lines at non-existent tics, but they will appear if the tics are subsequently enabled.

If no linetype is specified for the minor gridlines, the same linetype as the major gridlines is used. The default polar angle is 30 degrees.

If front is given, the grid is drawn on top of the graphed data. If back is given, the grid is drawn underneath the graphed data. Using front will prevent the grid from being obscured by dense data. The default setup, layerdefault, is equivalent to back for 2d plots. In 3D plots the default is to split up the grid and the graph box into two layers: one behind, the other in front of the plotted data and functions. Since hidden3d mode does its own sorting, it ignores all grid drawing order options and passes the grid lines through the hidden line removal machinery instead. These options actually affect not only the grid, but also the lines output by set border and the various ticmarks (see set xtics (p. 514)).

Z grid lines are drawn on the bottom of the plot. This looks better if a partial box is drawn around the plot — see set border (p. 264).

43.25 Hidden3d

The set hidden3d command enables hidden line removal for surface plotting (see splot (p. 539)). Some optional features of the underlying algorithm can also be controlled using this command.


     set hidden3d {defaults} |  
                  { {{offset <offset>} | {nooffset}}  
                    {trianglepattern <bitpattern>}  
                    {{undefined <level>} | {noundefined}}  
                    {{no}bentover} }  
     unset hidden3d  
     show hidden3d

In contrast to the usual display in gnuplot, hidden line removal actually treats the given function or data grids as real surfaces that can’t be seen through, so parts behind the surface will be hidden by it. For this to be possible, the surface needs to have ’grid structure’ (see splot datafile (p. 540) about this), and it has to be drawn with lines or with linespoints.

When hidden3d is set, both the hidden portion of the surface and possibly its contours drawn on the base (see set contour (p. 291)) as well as the grid will be hidden. Each surface has its hidden parts removed with respect to itself and to other surfaces, if more than one surface is plotted. Contours drawn on the surface (set contour surface) don’t work.

Labels and arrows are always visible and are unaffected. The key box is never hidden by the surface. As of gnuplot version 4.2, set hidden3d also affects 3D plotting styles with points, with labels, and with vectors, even if no surface is present in the graph. Individual plots within the graph may be explicitly excluded from this processing by appending the extra option nohidden3d to the with specifier.

Hidden3d does not affect solid surfaces drawn using the pm3d mode. To achieve a similar effect for pm3d surfaces, use instead set pm3d depthorder.

Functions are evaluated at isoline intersections. The algorithm interpolates linearly between function points or data points when determining the visible line segments. This means that the appearance of a function may be different when plotted with hidden3d than when plotted with nohidden3d because in the latter case functions are evaluated at each sample. Please see set samples (p. 433) and set isosamples (p. 335) for discussion of the difference.

The algorithm used to remove the hidden parts of the surfaces has some additional features controllable by this command. Specifying defaults will set them all to their default settings, as detailed below. If defaults is not given, only explicitly specified options will be influenced: all others will keep their previous values, so you can turn on/off hidden line removal via set {no}hidden3d, without modifying the set of options you chose.

The first option, offset, influences the linestyle used for lines on the ’back’ side. Normally, they are drawn in a linestyle one index number higher than the one used for the front, to make the two sides of the surface distinguishable. You can specify a different line style offset to add instead of the default 1, by offset <offset>. Option nooffset stands for offset 0, making the two sides of the surface use the same linestyle.

Next comes the option trianglepattern <bitpattern>. <bitpattern> must be a number between 0 and 7, interpreted as a bit pattern. Each bit determines the visibility of one edge of the triangles each surface is split up into. Bit 0 is for the ’horizontal’ edges of the grid, Bit 1 for the ’vertical’ ones, and Bit 2 for the diagonals that split each cell of the original grid into two triangles. The default pattern is 3, making all horizontal and vertical lines visible, but not the diagonals. You may want to choose 7 to see those diagonals as well.

The undefined <level> option lets you decide what the algorithm is to do with data points that are undefined (missing data, or undefined function values), or exceed the given x-, y- or z-ranges. Such points can either be plotted nevertheless, or taken out of the input data set. All surface elements touching a point that is taken out will be taken out as well, thus creating a hole in the surface. If <level> = 3, equivalent to option noundefined, no points will be thrown away at all. This may produce all kinds of problems elsewhere, so you should avoid this. <level> = 2 will throw away undefined points, but keep the out-of-range ones. <level> = 1, the default, will get rid of out-of-range points as well.

By specifying noaltdiagonal, you can override the default handling of a special case can occur if undefined is active (i.e. <level> is not 3). Each cell of the grid-structured input surface will be divided in two triangles along one of its diagonals. Normally, all these diagonals have the same orientation relative to the grid. If exactly one of the four cell corners is excluded by the undefined handler, and this is on the usual diagonal, both triangles will be excluded. However if the default setting of altdiagonal is active, the other diagonal will be chosen for this cell instead, minimizing the size of the hole in the surface.

The bentover option controls what happens to another special case, this time in conjunction with the trianglepattern. For rather crumply surfaces, it can happen that the two triangles a surface cell is divided into are seen from opposite sides (i.e. the original quadrangle is ’bent over’), as illustrated in the following ASCII art:

   original quadrangle:  A--B      displayed quadrangle:     |\   |  
     ("set view 0,0")    | /|    ("set view 75,75" perhaps)  | \  |  
                         |/ |                                |  \ |  
                         C--D                                |   \|  
                                                             A    D

If the diagonal edges of the surface cells aren’t generally made visible by bit 2 of the <bitpattern> there, the edge CB above wouldn’t be drawn at all, normally, making the resulting display hard to understand. Therefore, the default option of bentover will turn it visible in this case. If you don’t want that, you may choose nobentover instead. See also

hidden line removal demo (hidden.dem)


complex hidden line demo (singulr.dem).

43.26 Historysize

Note: the command set historysize is only available when gnuplot has been configured with the GNU readline.


     set historysize <int>  
     unset historysize

When leaving gnuplot, the value of historysize is used for truncating the history to at most that much lines. The default is 500. unset historysize will disable history truncation and thus allow an infinite number of lines to be written to the history file.

43.27 Isosamples

The isoline density (grid) for plotting functions as surfaces may be changed by the set isosamples command.


     set isosamples <iso_1> {,<iso_2>}  
     show isosamples

Each function surface plot will have <iso_1> iso-u lines and <iso_2> iso-v lines. If you only specify <iso_1>, <iso_2> will be set to the same value as <iso_1>. By default, sampling is set to 10 isolines per u or v axis. A higher sampling rate will produce more accurate plots, but will take longer. These parameters have no effect on data file plotting.

An isoline is a curve parameterized by one of the surface parameters while the other surface parameter is fixed. Isolines provide a simple means to display a surface. By fixing the u parameter of surface s(u,v), the iso-u lines of the form c(v) = s(u0,v) are produced, and by fixing the v parameter, the iso-v lines of the form c(u) = s(u,v0) are produced.

When a function surface plot is being done without the removal of hidden lines, set samples controls the number of points sampled along each isoline; see set samples (p. 433) and set hidden3d (p. 331). The contour algorithm assumes that a function sample occurs at each isoline intersection, so change in samples as well as isosamples may be desired when changing the resolution of a function surface/contour.

43.28 Key

The set key command enables a key (or legend) describing plots on a plot.

The contents of the key, i.e., the names given to each plotted data set and function and samples of the lines and/or symbols used to represent them, are determined by the title and with options of the {s}plot command. Please see plot title (p. 209) and plot with (p. 213) for more information.


     set key {on|off} {default}  
             {{inside | outside} | {lmargin | rmargin | tmargin | bmargin}  
               | {at <position>}}  
             {left | right | center} {top | bottom | center}  
             {vertical | horizontal} {Left | Right}  
             {{no}reverse} {{no}invert}  
             {samplen <sample_length>} {spacing <vertical_spacing>}  
             {width <width_increment>}  
             {height <height_increment>}  
             {{no}autotitle {columnheader}}  
             {title "<text>"} {{no}enhanced}  
             {{no}box { {linestyle | ls <line_style>}  
                        | {linetype | lt <line_type>}  
                          {linewidth | lw <line_width>}}}  
     unset key  
     show key

Plots may be drawn with no visible key by requesting set key off or unset key.

Elements within the key are stacked according to vertical or horizontal. In the case of vertical, the key occupies as few columns as possible. That is, elements are aligned in a column until running out of vertical space at which point a new column is started. In the case of horizontal, the key occupies as few rows as possible.

By default the key is placed in the upper right inside corner of the graph. The keywords left, right, top, bottom, center, inside, outside, lmargin, rmargin, tmargin, bmargin (, above, over, below and under) may be used to automatically place the key in other positions of the graph. Also an at <position> may be given to indicate precisely where the plot should be placed. In this case, the keywords left, right, top, bottom and center serve an analogous purpose for alignment.

To understand positioning, the best concept is to think of a region, i.e., inside/outside, or one of the margins. Along with the region, keywords left/center/right (l/c/r) and top/center/bottom (t/c/b) control where within the particular region the key should be placed.

When in inside mode, the keywords left (l), right (r), top (t), bottom (b), and center (c) push the key out toward the plot boundary as illustrated:

    t/l   t/c   t/r

    c/l    c    c/r

    b/l   b/c   b/r

When in outside mode, automatic placement is similar to the above illustration, but with respect to the view, rather than the graph boundary. That is, a border is moved inward to make room for the key outside of the plotting area, although this may interfere with other labels and may cause an error on some devices. The particular plot border that is moved depends upon the position described above and the stacking direction. For options centered in one of the dimensions, there is no ambiguity about which border to move. For the corners, when the stack direction is vertical, the left or right border is moved inward appropriately. When the stack direction is horizontal, the top or bottom border is moved inward appropriately.

The margin syntax allows automatic placement of key regardless of stack direction. When one of the margins lmargin (lm), rmargin (rm), tmargin (tm), and bmargin (bm) is combined with a single, non-conflicting direction keyword, the following illustrated positions may contain the key:

         l/tm  c/tm  r/tm

    t/lm                  t/rm

    c/lm                  c/rm

    b/lm                  b/rm

         l/bm  c/bm  r/bm

Keywords above and over are synonymous with tmargin. For version compatibility, above or over without an additional l/c/r or stack direction keyword uses center and horizontal. Keywords below and under are synonymous with bmargin. For compatibility, below or under without an additional l/c/r or stack direction keyword uses center and horizontal. A further compatibility issue is that outside appearing without an additional t/b/c or stack direction keyword uses top, right and vertical (i.e., the same as t/rm above).

The <position> can be a simple x,y,z as in previous versions, but these can be preceded by one of five keywords (first, second, graph, screen, character) which selects the coordinate system in which the position of the first sample line is specified. See coordinates (p. 41) for more details. The effect of left, right, top, bottom, and center when <position> is given is to align the key as though it were text positioned using the label command, i.e., left means left align with key to the right of <position>, etc.

Justification of the labels within the key is controlled by Left or Right (default is Right). The text and sample can be reversed (reverse) and a box can be drawn around the key (box {...}) in a specified linetype and linewidth, or a user-defined linestyle. Note that not all terminal drivers support linewidth selection, though.

By default the first plot label is at the top of the key and successive labels are entered below it. The invert option causes the first label to be placed at the bottom of the key, with successive labels entered above it. This option is useful to force the vertical ordering of labels in the key to match the order of box types in a stacked histogram.

The length of the sample line can be controlled by samplen. The sample length is computed as the sum of the tic length and <sample_length> times the character width. samplen also affects the positions of point samples in the key since these are drawn at the midpoint of the sample line, even if the sample line itself is not drawn.

The vertical spacing between lines is controlled by spacing. The spacing is set equal to the product of the pointsize, the vertical tic size, and <vertical_spacing>. The program will guarantee that the vertical spacing is no smaller than the character height.

The <width_increment> is a number of character widths to be added to or subtracted from the length of the string. This is useful only when you are putting a box around the key and you are using control characters in the text. gnuplot simply counts the number of characters in the string when computing the box width; this allows you to correct it.

The <height_increment> is a number of character heights to be added to or subtracted from the height of the key box. This is useful mainly when you are putting a box around the key, otherwise it can be used to adjust the vertical shift of automatically chosen key position by <height_increment>/2.

All plotted curves of plots and splots are titled according to the default option autotitles. The automatic generation of titles can be suppressed by noautotitles; then only those titles explicitly defined by (s)plot ... title ... will be drawn.

The set key autotitle columnheader option is available if gnuplot was built with –enable-datastrings. This command causes the first entry in each column of plotted data to be interpreted as a text string and used as a title for the corresponding plot. If the quantity being plotted is a function of data from several columns, gnuplot may be confused as to which column to draw the title from. In this case it is necessary to specify the column explicitly in the plot command, e.g. plot "datafile" using (($2+$3)/$4) title 3 with lines.

A title can be put on the key (title "<text>") — see also syntax (p. 101) for the distinction between text in single- or double-quotes. The key title uses the same justification as do the plot titles.

An explicitly given title is typeset using enhanced text properties on terminals supporting this, see enhanced text (p. 694) for more details. This default behavior can be switched off by the noenhanced option.

The defaults for set key are on, right, top, vertical, Right, noreverse, noinvert, samplen 4, spacing 1.25, title "", and nobox. The default <linetype> is the same as that used for the plot borders. Entering set key default returns the key to its default configuration.

The key is drawn as a sequence of lines, with one plot described on each line. On the right-hand side (or the left-hand side, if reverse is selected) of each line is a representation that attempts to mimic the way the curve is plotted. On the other side of each line is the text description (the line title), obtained from the plot command. The lines are vertically arranged so that an imaginary straight line divides the left- and right-hand sides of the key. It is the coordinates of the top of this line that are specified with the set key command. In a plot, only the x and y coordinates are used to specify the line position. For a splot, x, y and z are all used as a 3-d location mapped using the same mapping as the graph itself to form the required 2-d screen position of the imaginary line.

When using the TeX or PostScript drivers, or similar drivers where formatting information is embedded in the string, gnuplot is unable to calculate correctly the width of the string for key positioning. If the key is to be positioned at the left, it may be convenient to use the combination set key left Left reverse. The box and gap in the grid will be the width of the literal string.

If splot is being used to draw contours, the contour labels will be listed in the key. If the alignment of these labels is poor or a different number of decimal places is desired, the label format can be specified. See set clabel (p. 277) for details.


This places the key at the default location:

     set key default

This disables the key:

     unset key

This places a key at coordinates 2,3.5,2 in the default (first) coordinate system:

     set key at 2,3.5,2

This places the key below the graph:

     set key below

This places the key in the bottom left corner, left-justifies the text, gives it a title, and draws a box around it in linetype 3:

     set key left bottom Left title ’Legend’ box 3

43.29 Label

Arbitrary labels can be placed on the plot using the set label command.


     set label {<tag>} {"<label text>"} {at <position>}  
               {left | center | right}  
               {norotate | rotate {by <degrees>}}  
               {font "<name>{,<size>}"}  
               {front | back}  
               {textcolor <colorspec>}  
               {point <pointstyle> | nopoint}  
               {offset <offset>}  
     unset label {<tag>}  
     show label

The <position> is specified by either x,y or x,y,z, and may be preceded by first, second, graph, screen, or character to select the coordinate system. See coordinates (p. 41) for details.

The tag is an integer that is used to identify the label. If no <tag> is given, the lowest unused tag value is assigned automatically. The tag can be used to delete or modify a specific label. To change any attribute of an existing label, use the set label command with the appropriate tag, and specify the parts of the label to be changed.

The <label text> can be a string constant, a string variable, or a string- valued expression. See strings (p. 86), sprintf (p. 51), and gprintf (p. 323).

By default, the text is placed flush left against the point x,y,z. To adjust the way the label is positioned with respect to the point x,y,z, add the justification parameter, which may be left, right or center, indicating that the point is to be at the left, right or center of the text. Labels outside the plotted boundaries are permitted but may interfere with axis labels or other text.

If rotate is given, the label is written vertically (if the terminal can do so, of course). If rotate by <degrees> is given, conforming terminals will try to write the text at the specified angle; non-conforming terminals will treat this as vertical text.

Font and its size can be chosen explicitly by font "<name>{,<size>}" if the terminal supports font settings. Otherwise the default font of the terminal will be used.

Normally the enhanced text mode string interpretation, if enabled for the current terminal, is applied to all text strings including label text. The noenhanced property can be used to exempt a specific label from the enhanced text mode processing. The can be useful if the label contains underscores, for example. See enhanced text (p. 694).

If front is given, the label is written on top of the graphed data. If back is given (the default), the label is written underneath the graphed data. Using front will prevent a label from being obscured by dense data.

textcolor <colorspec> changes the color of the label text. <colorspec> can be a linetype, an rgb color, or a palette mapping. See help for colorspec (p. 68) and palette (p. 400). textcolor may be abbreviated tc.

  ‘tc default‘ resets the text color to its default state.  
  ‘tc lt <n>‘ sets the text color to that of line type <n>.  
  ‘tc ls <n>‘ sets the text color to that of line style <n>.  
  ‘tc palette z‘ selects a palette color corresponding to the label z position.  
  ‘tc palette cb <val>‘ selects a color corresponding to <val> on the colorbar.  
  ‘tc palette fraction <val>‘, with 0<=val<=1, selects a color corresponding to  
      the mapping [0:1] to grays/colors of the ‘palette‘.  
  ‘tc rgb "#RRGGBB"‘ selects an arbitrary 24-bit RGB color.

If a <pointstyle> is given, using keywords lt, pt and ps, see style (p. 213), a point with the given style and color of the given line type is plotted at the label position and the text of the label is displaced slightly. This option is used by default for placing labels in mouse enhanced terminals. Use nopoint to turn off the drawing of a point near the label (this is the default).

The displacement defaults to 1,1 in pointsize units if a <pointstyle> is given, 0,0 if no <pointstyle> is given. The displacement can be controlled by the optional offset <offset> where <offset> is specified by either x,y or x,y,z, and may be preceded by first, second, graph, screen, or character to select the coordinate system. See coordinates (p. 41) for details.

If one (or more) axis is timeseries, the appropriate coordinate should be given as a quoted time string according to the timefmt format string. See set xdata (p. 503) and set timefmt (p. 495).

The EEPIC, Imagen, LaTeX, and TPIC drivers allow \\ in a string to specify a newline.


To set a label at (1,2) to "y=x", use:

     set label "y=x" at 1,2

To set a Sigma of size 24, from the Symbol font set, at the center of the graph, use:

     set label "S" at graph 0.5,0.5 center font "Symbol,24"

To set a label "y=x^2" with the right of the text at (2,3,4), and tag the label as number 3, use:

     set label 3 "y=x^2" at 2,3,4 right

To change the preceding label to center justification, use:

     set label 3 center

To delete label number 2, use:

     unset label 2

To delete all labels, use:

     unset label

To show all labels (in tag order), use:

     show label

To set a label on a graph with a timeseries on the x axis, use, for example:

     set timefmt "%d/%m/%y,%H:%M"  
     set label "Harvest" at "25/8/93",1

To display a freshly fitted parameter on the plot with the data and the fitted function, do this after the fit, but before the plot:

     set label sprintf("a = %3.5g",par_a) at 30,15  
     bfit = gprintf("b = %s*10^%S",par_b)  
     set label bfit at 30,20

To set a label displaced a little bit from a small point:

     set label ’origin’ at 0,0 point lt 1 pt 2 ps 3 offset 1,-1

To set a label whose color matches the z value (in this case 5.5) of some point on a 3D splot colored using pm3d:

     set label ’text’ at 0,0,5.5 tc palette z

43.30 Lmargin

The command set lmargin sets the size of the left margin. Please see set margin (p. 375) for details.

43.31 Loadpath

The loadpath setting defines additional locations for data and command files searched by the call, load, plot and splot commands. If a file cannot be found in the current directory, the directories in loadpath are tried.


     set loadpath {"pathlist1" {"pathlist2"...}}  
     show loadpath

Path names may be entered as single directory names, or as a list of path names separated by a platform-specific path separator, eg. colon (’:’) on Unix, semicolon (’;’) on DOS/Windows/OS/2/Amiga platforms. The show loadpath, save and save set commands replace the platform-specific separator with a space character (’ ’) for maximum portability.

If the environment variable GNUPLOT_LIB is set, its contents are appended to loadpath. However, show loadpath prints the contents of user defined loadpath and system loadpath separately. Also, the save and save set commands save only the user specified parts of loadpath, for portability reasons.

43.32 Locale

The locale setting determines the language with which {x,y,z}{d,m}tics will write the days and months.


     set locale {"<locale>"}

<locale> may be any language designation acceptable to your installation. See your system documentation for the available options. The default value is determined from the LC_TIME, LC_ALL, or LANG environment variables.

To change the decimal point locale, see set decimalsign (p. 307).

43.33 Logscale


     set logscale <axes> <base>  
     unset logscale <axes>  
     show logscale

where <axes> may be any combination of x, x2, y, y2, z, and cb in any order, and where <base> is the base of the log scaling. If <base> is not given, then 10 is assumed. If <axes> is not given, then all axes are assumed. unset logscale turns off log scaling for the specified axes.


To enable log scaling in both x and z axes:

     set logscale xz

To enable scaling log base 2 of the y axis:

     set logscale y 2

To enable z and color log axes for a pm3d plot:

     set logscale zcb

To disable z axis log scaling:

     unset logscale z

43.34 Macros

If command line macro substitution is enabled, then tokens in the command line of the form @<stringvariablename> will be replaced by the text string contained in <stringvariablename>. See substitution (p. 91).


    set macros

43.35 Mapping

If data are provided to splot in spherical or cylindrical coordinates, the set mapping command should be used to instruct gnuplot how to interpret them.


     set mapping {cartesian | spherical | cylindrical}

A cartesian coordinate system is used by default.

For a spherical coordinate system, the data occupy two or three columns (or using entries). The first two are interpreted as the azimuthal and polar angles theta and phi (or "longitude" and "latitude"), in the units specified by set angles. The radius r is taken from the third column if there is one, or is set to unity if there is no third column. The mapping is:

     x = r * cos(theta) * cos(phi)  
     y = r * sin(theta) * cos(phi)  
     z = r * sin(phi)

Note that this is a "geographic" spherical system, rather than a "polar" one (that is, phi is measured from the equator, rather than the pole).

For a cylindrical coordinate system, the data again occupy two or three columns. The first two are interpreted as theta (in the units specified by set angles) and z. The radius is either taken from the third column or set to unity, as in the spherical case. The mapping is:

     x = r * cos(theta)  
     y = r * sin(theta)  
     z = z

The effects of mapping can be duplicated with the using filter on the splot command, but mapping may be more convenient if many data files are to be processed. However even if mapping is used, using may still be necessary if the data in the file are not in the required order.

mapping has no effect on plot.

world.dem: mapping demos.

43.36 Margin

The computed margins can be overridden by the set margin commands. show margin shows the current settings.


     set bmargin {{at screen} <margin>}  
     set lmargin {{at screen} <margin>}  
     set rmargin {{at screen} <margin>}  
     set tmargin {{at screen} <margin>}  
     show margin

The default units of <margin> are character heights or widths, as appropriate. A positive value defines the absolute size of the margin. A negative value (or none) causes gnuplot to revert to the computed value. For 3D plots, only the left margin can be set using character units.

The keywords at screen indicates that the margin is specified as a fraction of the full drawing area. This can be used to precisely line up the corners of individual 2D and 3D graphs in a multiplot. This placement ignores the current values of set origin and set size, and is intended as an alternative method for positioning graphs within a multiplot.

Normally the margins of a plot are automatically calculated based on tics, tic labels, axis labels, the plot title, the timestamp and the size of the key if it is outside the borders. If, however, tics are attached to the axes (set xtics axis, for example), neither the tics themselves nor their labels will be included in either the margin calculation or the calculation of the positions of other text to be written in the margin. This can lead to tic labels overwriting other text if the axis is very close to the border.

43.37 Mouse

The command set mouse enables mouse actions. Currently the pm, x11, ggi, windows and wxt terminals are mouse enhanced. There are two mouse modes. The 2d-graph mode works for 2d graphs and for maps (i.e. splots with set view having z-rotation 0, 90, 180, 270 or 360 degrees, including set view map) and it allows tracing the position over graph, zooming, annotating graph etc. For 3d graphs splot, the view and scaling of the graph can be changed with mouse buttons 1 and 2. If additionally to these buttons the modifier <ctrl> is hold down, the coordinate system only is rotated which is useful for large data sets. A vertical motion of Button 2 with the shift key hold down changes the ticslevel.

Mousing is not available in multiplot mode. When multiplot is finished using unset multiplot, then the mouse will be turned on again and acts on the last plot (like replot does).


     set mouse {doubleclick <ms>} {nodoubleclick} \  
               {{no}zoomcoordinates} \  
               {noruler | ruler {at x,y}} \  
               {polardistance{deg|tan} | nopolardistance} \  
               {format <string>} \  
               {clipboardformat <int>/<string>} \  
               {mouseformat <int>/<string>} \  
               {{no}labels} {labeloptions <string>} \  
               {{no}zoomjump} {{no}verbose}  
     unset mouse

The doubleclick resolution is given in milliseconds and used for Button 1 which copies the current mouse position to the clipboard. If you want that to be done by single clicking a value of 0 ms can be used. The default value is 300 ms.

The option zoomcoordinates determines if the coordinates of the zoom box are drawn at the edges while zooming. This is on by default.

The options noruler and ruler switch the ruler off and on, the latter optionally at given coordinates. This corresponds to the default key binding ’r’.

The option polardistance determines if the distance between the mouse cursor and the ruler is also shown in polar coordinates (distance and angle in degrees or tangent (slope)). This corresponds to the default key binding ’5’.

The format option takes a fprintf like format string which determines how floating point numbers are printed to the drivers window and the clipboard. The default is "% #g".

clipboardformat and mouseformat are used for formatting the text on Button1 and Button2 actions – copying the coordinates to the clipboard and temporarily annotating the mouse position. This corresponds to the key bindings ’1’, ’2’, ’3’, ’4’ (see the drivers’s help window). If the argument is a string this string is used as c format specifier and should contain two float specifiers, e.g. set mouse mouseformat "mouse = %5.2g, %10.2f". Use set mouse mouseformat "" to turn this string off again.

The following formats are available (format 6 may only be selected if the format string was specified already):

0   real coordinates in  brackets e.g. [1.23, 2.45]  
1   real coordinates w/o brackets e.g.  1.23, 2.45  
2   x == timefmt                       [(as set by ‘set timefmt‘), 2.45]  
3   x == date                          [31. 12. 1999, 2.45]  
4   x == time                          [23:59, 2.45]  
5   x == date / time                   [31. 12. 1999 23:59, 2.45]  
6   alt. format, specified as string   ""

Choose the option labels to get real gnuplot labels on Button 2. (The default is nolabels which makes Button 2 drawing only temporary annotations at the mouse positions). The labels are drawn with the current setting of mouseformat. labeloptions controls which options are passed to the set label command. The default is "pointstyle 1" which will plot a small plus at the label position. Note that the pointsize is taken from the set pointsize command. Labels can be removed by holding the Ctrl-Key down while clicking with Button 2 on the label’s point. The threshold for how close you must be to the label is also determined by the pointsize.

If the option zoomjump is on, the mouse pointer will be automatically offset a small distance after starting a zoom region with button 3. This can be useful to avoid a tiny (or even empty) zoom region. zoomjump is off by default.

If the option verbose is turned on the communication commands are shown during execution. This option can also be toggled by hitting 6 in the driver’s window. verbose is off by default.

Press ’h’ in the driver’s window for a short summary of the mouse and key bindings. This will also display user defined bindings or hotkeys which can be defined using the bind command, see help for bind (p. 73). Note, that user defined hotkeys may override the default bindings.

Press ’q’ in the driver’s window to close the window. This key cannot be overridden with the bind command.

See also help for bind (p. 73) and label (p. 351).

43.37.1 X11 mouse

If multiple X11 plot windows have been opened using the set term x11 <n> terminal option, then only the current plot window supports the entire range of mouse commands and hotkeys. The other windows will, however, continue to display mouse coordinates at the lower left.

For consistency with other screen terminals, X11 mouse support is turned on by default, wherever the standard input comes from. However, on some UNIX flavors, special input devices as /dev/null might not be select-able; using such devices with the mouse turned on will hang gnuplot. Please turn off mousing with unset mouse if you are in this situation.

43.38 Multiplot

The command set multiplot places gnuplot in the multiplot mode, in which several plots are placed on the same page, window, or screen.


     set multiplot { layout <rows>,<cols>  
                     {rowsfirst|columnsfirst} {downwards|upwards}  
                     {title <page title>}  
                     {scale <xscale>{,<yscale>}} {offset <xoff>{,<yoff>}}  
     unset multiplot

For some terminals, no plot is displayed until the command unset multiplot is given, which causes the entire page to be drawn and then returns gnuplot to its normal single-plot mode. For other terminals, each separate plot command produces an updated display, either by redrawing all previous ones and the newly added plot, or by just adding the new plot to the existing display.

The area to be used by the next plot is not erased before doing the new plot. The clear command can be used to do this if wanted, as is typically the case for "inset" plots.

Any labels or arrows that have been defined will be drawn for each plot according to the current size and origin (unless their coordinates are defined in the screen system). Just about everything else that can be set is applied to each plot, too. If you want something to appear only once on the page, for instance a single time stamp, you’ll need to put a set time/unset time pair around one of the plot, splot or replot commands within the set multiplot/unset multiplot block.

The multiplot title is separate from the individual plot titles, if any. Space is reserved for it at the top of the page, spanning the full width of the canvas.

The commands set origin and set size must be used to correctly position each plot if no layout is specified or if fine tuning is desired. See set origin (p. 386) and set size (p. 434) for details of their usage.


     set multiplot  
     set size 0.4,0.4  
     set origin 0.1,0.1  
     plot sin(x)  
     set size 0.2,0.2  
     set origin 0.5,0.5  
     plot cos(x)  
     unset multiplot

This displays a plot of cos(x) stacked above a plot of sin(x).

set size and set origin refer to the entire plotting area used for each plot. Please also see set term size (p. 30). If you want to have the axes themselves line up, you can guarantee that the margins are the same size with the set margin commands. See set margin (p. 375) for their use. Note that the margin settings are absolute, in character units, so the appearance of the graph in the remaining space will depend on the screen size of the display device, e.g., perhaps quite different on a video display and a printer.

With the layout option you can generate simple multiplots without having to give the set size and set origin commands before each plot: Those are generated automatically, but can be overridden at any time. With layout the display will be divided by a grid with <rows> rows and <cols> columns. This grid is filled rows first or columns first depending on whether the corresponding option is given in the multiplot command. The stack of plots can grow downwards or upwards. Default is rowsfirst and downwards.

Each plot can be scaled by scale and shifted with offset; if the y-values for scale or offset are omitted, the x-value will be used. unset multiplot will turn off the automatic layout and restore the values of set size and set origin as they were before set multiplot layout.


     set size 1,1  
     set origin 0,0  
     set multiplot layout 3,2 columnsfirst scale 1.1,0.9  
     [ up to 6 plot commands here ]  
     unset multiplot

The above example will produce 6 plots in 2 columns filled top to bottom, left to right. Each plot will have a horizontal size of 1.1/2 and a vertical size of 0.9/3.

See also

multiplot demo (multiplt.dem)

43.39 Mx2tics

Minor tic marks along the x2 (top) axis are controlled by set mx2tics. Please see set mxtics (p. 381).

43.40 Mxtics

Minor tic marks along the x axis are controlled by set mxtics. They can be turned off with unset mxtics. Similar commands control minor tics along the other axes.


     set mxtics {<freq> | default}  
     unset mxtics  
     show mxtics

The same syntax applies to mytics, mztics, mx2tics, my2tics and mcbtics.

<freq> is the number of sub-intervals (NOT the number of minor tics) between major tics (the default for a linear axis is either two or five depending on the major tics, so there are one or four minor tics between major tics). Selecting default will return the number of minor ticks to its default value.

If the axis is logarithmic, the number of sub-intervals will be set to a reasonable number by default (based upon the length of a decade). This will be overridden if <freq> is given. However the usual minor tics (2, 3, ..., 8, 9 between 1 and 10, for example) are obtained by setting <freq> to 10, even though there are but nine sub-intervals.

To set minor tics at arbitrary positions, use the ("<label>" <pos> <level>, ...) form of set {x|x2|y|y2|z}tics with <label> empty and <level> set to 1.

The set m{x|x2|y|y2|z}tics commands work only when there are uniformly spaced major tics. If all major tics were placed explicitly by set {x|x2|y|y2|z}tics, then minor tic commands are ignored. Implicit major tics and explicit minor tics can be combined using set {x|x2|y|y2|z}tics and set {x|x2|y|y2|z}tics add.


     set xtics 0, 5, 10  
     set xtics add (7.5)  
     set mxtics 5

Major tics at 0,5,7.5,10, minor tics at 1,2,3,4,6,7,8,9

     set logscale y  
     set ytics format ""  
     set ytics 1e-6, 10, 1  
     set ytics add ("1" 1, ".1" 0.1, ".01" 0.01, "10^-3" 0.001, \  
                    "10^-4" 0.0001)  
     set mytics 10

Major tics with special formatting, minor tics at log positions

By default, minor tics are off for linear axes and on for logarithmic axes. They inherit the settings for axis|border and {no}mirror specified for the major tics. Please see set xtics (p. 514) for information about these.

43.41 My2tics

Minor tic marks along the y2 (right-hand) axis are controlled by set my2tics. Please see set mxtics (p. 381).

43.42 Mytics

Minor tic marks along the y axis are controlled by set mytics. Please see set mxtics (p. 381).

43.43 Mztics

Minor tic marks along the z axis are controlled by set mztics. Please see set mxtics (p. 381).

43.44 Offsets

Offsets provide a mechanism to put a boundary around the data inside of an autoscaled graph.


     set offsets <left>, <right>, <top>, <bottom>  
     unset offsets  
     show offsets

Each offset may be a constant or an expression. Each defaults to 0. Left and right offsets are given in units of the x axis, top and bottom offsets in units of the y axis. A positive offset expands the graph in the specified direction, e.g., a positive bottom offset makes ymin more negative. Negative offsets, while permitted, can have unexpected interactions with autoscaling and clipping.

Offsets are ignored in splots.


     set offsets 0, 0, 2, 2  
     plot sin(x)

This graph of sin(x) will have a y range [-3:3] because the function will be autoscaled to [-1:1] and the vertical offsets are each two.

43.45 Origin

The set origin command is used to specify the origin of a plotting surface (i.e., the graph and its margins) on the screen. The coordinates are given in the screen coordinate system (see coordinates (p. 41) for information about this system).


     set origin <x-origin>,<y-origin>

43.46 Output

By default, screens are displayed to the standard output. The set output command redirects the display to the specified file or device.


     set output {"<filename>"}  
     show output

The filename must be enclosed in quotes. If the filename is omitted, any output file opened by a previous invocation of set output will be closed and new output will be sent to STDOUT. (If you give the command set output "STDOUT", your output may be sent to a file named "STDOUT"! ["May be", not "will be", because some terminals, like x11 or wxt, ignore set output.])

MSDOS users should note that the \ character has special significance in double-quoted strings, so single-quotes should be used for filenames in different directories.

When both set terminal and set output are used together, it is safest to give set terminal first, because some terminals set a flag which is needed in some operating systems. This would be the case, for example, if the operating system needs to know whether or not a file is to be formatted in order to open it properly.

On machines with popen functions (Unix), output can be piped through a shell command if the first non-whitespace character of the filename is ’|’. For instance,

     set output "|lpr -Plaser filename"  
     set output "|lp -dlaser filename"

On MSDOS machines, set output "PRN" will direct the output to the default printer. On VMS, output can be sent directly to any spooled device. It is also possible to send the output to DECnet transparent tasks, which allows some flexibility.

43.47 Parametric

The set parametric command changes the meaning of plot (splot) from normal functions to parametric functions. The command unset parametric restores the plotting style to normal, single-valued expression plotting.


     set parametric  
     unset parametric  
     show parametric

For 2-d plotting, a parametric function is determined by a pair of parametric functions operating on a parameter. An example of a 2-d parametric function would be plot sin(t),cos(t), which draws a circle (if the aspect ratio is set correctly — see set size (p. 434)). gnuplot will display an error message if both functions are not provided for a parametric plot.

For 3-d plotting, the surface is described as x=f(u,v), y=g(u,v), z=h(u,v). Therefore a triplet of functions is required. An example of a 3-d parametric function would be cos(u)*cos(v),cos(u)*sin(v),sin(u), which draws a sphere. gnuplot will display an error message if all three functions are not provided for a parametric splot.

The total set of possible plots is a superset of the simple f(x) style plots, since the two functions can describe the x and y values to be computed separately. In fact, plots of the type t,f(t) are equivalent to those produced with f(x) because the x values are computed using the identity function. Similarly, 3-d plots of the type u,v,f(u,v) are equivalent to f(x,y).

Note that the order the parametric functions are specified is xfunction, yfunction (and zfunction) and that each operates over the common parametric domain.

Also, the set parametric function implies a new range of values. Whereas the normal f(x) and f(x,y) style plotting assume an xrange and yrange (and zrange), the parametric mode additionally specifies a trange, urange, and vrange. These ranges may be set directly with set trange, set urange, and set vrange, or by specifying the range on the plot or splot commands. Currently the default range for these parametric variables is [-5:5]. Setting the ranges to something more meaningful is expected.

43.48 Plot

The show plot command shows the current plotting command as it results from the last plot and/or splot and possible subsequent replot commands.

In addition, the show plot add2history command adds this current plot command into the history. It is useful if you have used replot to add more curves to the current plot and you want to edit the whole command now.

43.49 Pm3d

pm3d is an splot style for drawing palette-mapped 3d and 4d data as color/gray maps and surfaces. It uses a pm3d algorithm which allows plotting gridded as well as non-gridded data without preprocessing, even when the data scans do not have the same number of points.

Drawing of color surfaces is available on terminals supporting filled colored polygons with color mapping specified by palette. Currently supported terminals include

 Screen terminals:  
   OS/2 Presentation Manager  
   Linux VGA (vgagl)  
   AquaTerm (Mac OS X)  
   wxWidgets (wxt)  
   pslatex, pstex, epslatex  
   gif, png, jpeg  

Let us first describe how a map/surface is drawn. The input data come from an evaluated function or from an splot data file. Each surface consists of a sequence of separate scans (isolines). The pm3d algorithm fills the region between two neighbouring points in one scan with another two points in the next scan by a gray (or color) according to z-values (or according to an additional ’color’ column, see help for using (p. 185)) of these 4 corners; by default the 4 corner values are averaged, but this can be changed by the option corners2color. In order to get a reasonable surface, the neighbouring scans should not cross and the number of points in the neighbouring scans should not differ too much; of course, the best plot is with scans having same number of points. There are no other requirements (e.g. the data need not be gridded). Another advantage is that the pm3d algorithm does not draw anything outside of the input (measured or calculated) region.

Surface coloring works with the following input data:

1. splot of function or of data file with one or three data columns: The gray/color scale is obtained by mapping the averaged (or corners2color) z-coordinate of the four corners of the above-specified quadrangle into the range [min_color_z,max_color_z] of zrange or cbrange providing a gray value in the range [0:1]. This value can be used directly as the gray for gray maps. The normalized gray value can be further mapped into a color — see set palette (p. 400) for the complete description.

2. splot of data file with two or four data columns: The gray/color value is obtained by using the last-column coordinate instead of the z-value, thus allowing the color and the z-coordinate be mutually independent. This can be used for 4d data drawing.

Other notes:

1. The term ’scan’ referenced above is used more among physicists than the term ’iso_curve’ referenced in gnuplot documentation and sources. You measure maps recorded one scan after another scan, that’s why.

2. The ’gray’ or ’color’ scale is a linear mapping of a continuous variable onto a smoothly varying palette of colors. The mapping is shown in a rectangle next to the main plot. This documentation refers to this as a "colorbox", and refers to the indexing variable as lying on the colorbox axis. See set colorbox (p. 288), set cbrange (p. 536).

3. To use pm3d coloring to generate a two-dimensional plot rather than a 3D surface, use set view map or set pm3d map.

Syntax (the options can be given in any order):

     set pm3d {  
                { at <bst combination> }  
                { interpolate <steps in scan>,<steps between scans> }  
                { scansautomatic | scansforward | scansbackward | depthorder }  
                { flush { begin | center | end } }  
                { ftriangles | noftriangles }  
                { clip1in | clip4in }  
                { corners2color { mean|geomean|median|min|max|c1|c2|c3|c4 } }  
                { hidden3d <linestyle> | nohidden3d }  
                { implicit | explicit }  
                { map }  
     show pm3d  
     unset pm3d

Color surface is drawn if data or function style is set to pm3d globally or via ’with’ option, or if the option implicit is on — then the pm3d surface is combined with the line surface mesh. See bottom of this section for mode details.

Color surface can be drawn at the base or top (then it is a gray/color planar map) or at z-coordinates of surface points (gray/color surface). This is defined by the at option with a string of up to 6 combinations of b, t and s. For instance, at b plots at bottom only, at st plots firstly surface and then top map, while at bstbst will never by seriously used.

Colored quadrangles are plotted one after another. When plotting surfaces (at s), the later quadrangles overlap (overdraw) the previous ones. (Gnuplot is not virtual reality tool to calculate intersections of filled polygon meshes.) You may try to switch between scansforward and scansbackward to force the first scan of the data to be plotted first or last. The default is scansautomatic where gnuplot makes a guess about scans order. On the other hand, the depthorder option completely reorders the qudrangles. The rendering is performed after a depth sorting, which allows to visualize even complicated surfaces; see pm3d depthorder (p. 399) for more details.

If two subsequent scans do not have same number of points, then it has to be decided whether to start taking points for quadrangles from the beginning of both scans (flush begin), from their ends (flush end) or to center them (flush center). Note, that flush (center|end) are incompatible with scansautomatic: if you specify flush center or flush end and scansautomatic is set, it is silently switched to scansforward.

If two subsequent scans do not have the same number of points, the option ftriangles specifies whether color triangles are drawn at the scan tail(s) where there are not enough points in either of the scan. This can be used to draw a smooth map boundary.

Clipping with respect to x, y coordinates of quadrangles can be done in two ways. clip1in: all 4 points of each quadrangle must be defined and at least 1 point of the quadrangle must lie in the x and y ranges. clip4in: all 4 points of each quadrangle must lie in the x and y ranges.

There is a single gray/color value associated to each drawn pm3d quadrangle (no smooth color change among vertices). The value is calculated from z-coordinates from the surrounding corners according to corners2color <option>. The options ’mean’ (default), ’geomean’ and ’median’ produce various kinds of surface color smoothing, while options ’min’ and ’max’ choose minimal or maximal value, respectively. This may not be desired for pixel images or for maps with sharp and intense peaks, in which case the options ’c1’, ’c2’, ’c3’ or ’c4’ can be used instead to assign the quadrangle color based on the z-coordinate of only one corner. Some experimentation may be needed to determine which corner corresponds to ’c1’, as the orientation depends on the drawing direction. Because the pm3d algorithm does not extend the colored surface outside the range of the input data points, the ’c<j>’ coloring options will result in pixels along two edges of the grid not contributing to the color of any quadrangle. For example, applying the pm3d algorithm to the 4x4 grid of data points in script demo/pm3d.dem (please have a look) produces only (4-1)x(4-1)=9 colored rectangles.

Another drawing algorithm, which would draw quadrangles around a given node by taking corners from averaged (x,y)-coordinates of its surrounding 4 nodes while using node’s color, could be implemented in the future. This is already done for drawing images (2D grids) via image and rgbimage styles.

Notice that ranges of z-values and color-values for surfaces are adjustable independently by set zrange, set cbrange, as well as set log for z or cb. Maps can be adjusted by the cb-axis only; see also set view map (p. 500) and set colorbox (p. 288).

The option hidden3d takes as the argument a linestyle which must be created by set style line .... (The style need not to be present when setting pm3d, but it must be present when plotting). If set, lines are drawn using the specified line style, taking into account hidden line removal. This is by far more efficient than using the command set hidden3d as it doesn’t really calculate hidden line removal, but just draws the filled polygons in the correct order. So the recommended choice when using pm3d is

     set pm3d at s hidden3d 100  
     set style line 100 lt 5 lw 0.5  
     unset hidden3d  
     unset surf  
     splot x*x+y*y

There used to be an option {transparent|solid} to this command. Now you get the same effect from set grid {front|layerdefault}, respectively.

The set pm3d map is an abbreviation for set pm3d at b; set view map; set style data pm3d; set style func pm3d;. It is used for backwards compatibility, when set view map was not available. Take care that you properly use zrange and cbrange for input data point filtering and color range scaling, respectively; and also set (no)surface seems to have a (side?) effect.

The option interpolate will interpolate grid points into a finer mesh, and color each quadrangle appropriately. For data files, this will smoothen the color surface, and enhance spikes in a color surface. For functions, interpolation makes little sense, except to trade off precision for memory. It would usually make more sense to use samples and isosamples when working with functions.

The coloring setup as well as the color box drawing are determined by set palette. There can be only one palette for the current plot. Drawing of several surfaces with different palettes can be achieved by multiplot with fixed origin and size; don’t forget to use set palette maxcolors when your terminal is running out of available colors.

On gnuplot start-up, mode is explicit. For historical and thus compatibility reasons, commands set pm3d; (i.e. no options) and set pm3d at X ... (i.e. at is the first option) sets mode implicit. Further, set pm3d; sets up the other options to their default.

If the option implicit is on, all surface plots will be plotted additionally to the default type, e.g.

     splot ’fred.dat’ with lines, ’lola.dat’ with lines

would give both plots (meshes) additionally to a pm3d surface. That’s what you are used to after set pm3d;.

If the option explicit is on (or implicit is off) only plots specified by the with pm3d attribute are plotted with a pm3d surface, e.g.:

     splot ’fred.dat’ with lines, ’lola.dat’ with pm3d

would plot ’fred.dat’ with lines (and only lines) and ’lola.dat’ with a pm3d surface.

If you set the default data or function style to pm3d, e.g.:

     set style data pm3d

then the options implicit and explicit have no effect.

Note that when plotting several plots, they are plotted in the order given on the command line. This can be of interest especially for filled surfaces which can overwrite and therefore hide part of earlier plots.

If with pm3d is specified in the splot command line, then it accepts the ’at’ option. The following plots draw three color surfaces at different altitudes:

     set border 4095  
     set pm3d at s  
     splot 10*x with pm3d at b, x*x-y*y, x*x+y*y with pm3d at t

See also help for set palette (p. 400), set cbrange (p. 536), set colorbox (p. 288), x11 pm3d (p. 752) and definitely the demo file demo/pm3d.dem.

43.49.1 Depthorder

By default the quadrangles making up a pm3d solid surface are rendered in the order they are encountered along the surface grid points. This order may be controlled by the options scansautomatic|scansforward|scansbackward. These scan options are not in general compatible with hidden-surface removal.

Gnuplot does not do true hidden surface removal for solid surfaces, but often it is sufficient to render the component quadrangles in order from furthest to closest. This mode may be selected using the options

     set pm3d depthorder hidden3d

The depthorder option orders the solid quadrangles; the hidden3d option similarly orders the bounding lines (if drawn). Note that the global option set hidden3d does not affect pm3d surfaces.

43.50 Palette

Palette is a color storage for use by pm3d, filled color contours or polygons, color histograms, color gradient background, and whatever it is or it will be implemented... Here it stands for a palette of smooth "continuous" colors or grays, but let’s call it just a palette.

Color palettes require terminal entries for filled color polygons and palettes of smooth colors, are currently available for terminals listed in help for set pm3d. The range of color values are adjustable independently by set cbrange and set log cb. The whole color palette is visualized in the colorbox.


     set palette  
     set palette {  
                { gray | color }  
                { gamma <gamma> }  
                {   rgbformulae <r>,<g>,<b>  
                  | defined { ( <gray1> <color1> {, <grayN> <colorN>}... ) }  
                  | file ’<filename>’ {datafile-modifiers}  
                  | functions <R>,<G>,<B>  
                { model { RGB | HSV | CMY | YIQ | XYZ } }  
                { positive | negative }  
                { nops_allcF | ps_allcF }  
                { maxcolors <maxcolors> }  
     show palette  
     show palette palette <n> {{float | int}}  
     show palette gradient  
     show palette fit2rgbformulae  
     show palette rgbformulae  
     show palette colornames

set palette (i.e. without options) sets up the default values. Otherwise, the options can be given in any order. show palette shows the current palette properties.

show palette gradient displays the gradient defining the palette (if appropriate). show palette rgbformulae prints the available fixed gray –> color transformation formulae. show palette colornames prints the implemented color names.

show palette palette <n> prints to screen or to the file given by set output table of RGB triplets calculated for the current palette settings and a palette having <n> discrete colors. The default wide table can be limited to 3 columns of r,g,b float values [0..1] or integer values [0..255] by options float or int, respectively. This way, the current gnuplot color palette can be loaded into other imaging applications, for example Octave. Additionally to this textual list of RGB table, you can enjoy command test palette to draw graphically the R,G,B profiles for the current palette.

The following options determine the coloring properties.

Figure using this palette can be gray or color. For instance, in pm3d color surfaces the gray of each small spot is obtained by mapping the averaged z-coordinate of the 4 corners of surface quadrangles into the range [min_z,max_z] providing range of grays [0:1]. This value can be used directly as the gray for gray maps. The color map requires a transformation gray –> (R,G,B), i.e. a mapping [0:1] –> ([0:1],[0:1],[0:1]).

Basically two different types of mappings can be used: Analytic formulae to convert gray to color, or discrete mapping tables which are interpolated. palette rgbformulae and palette functions use analytic formulae whereas palette defined and palette file use interpolated tables. palette rgbformulae reduces the size of postscript output to a minimum.

The command show palette fit2rgbformulae finds the best matching set palette rgbformulae for the current set palette. Naturally, it makes sense to use it for non-rgbformulae palettes. This command can be found useful mainly for external programs using the same rgbformulae definition of palettes as gnuplot, like zimg (


set palette gray switches to a gray only palette. set palette rgbformulae, set palette defined, set palette file and set palette functions switch to a color mapping. set palette color is an easy way to switch back from the gray palette to the last color mapping.

Automatic gamma correction via set palette gamma <gamma> can be done for gray maps only (set palette gray). Linear mapping to gray is for gamma equals 1, see test palette (p. 560). Gamma is ignored for color mappings.

Most terminals support only discrete number of colors (e.g. 256 colors in gif). All entries of the palette remaining after the default gnuplot linetype colors declaration are allocated for pm3d by default. Then multiplot could fail if there are no more color positions in the terminal available. Then you should use set palette maxcolors <maxcolors> with a reasonably small value. This option can also be used to separate levels of z=constant in discrete steps, thus to emulate filled contours. Default value of 0 stays for allocating all remaining entries in the terminal palette or for to use exact mapping to RGB.

RGB color space might not be the most useful color space to work in. For that reason you may change the color space with model to one of RGB, HSV, CMY, YIQ and XYZ. Using color names for set palette defined tables and a color space other than RGB will result in funny colors. All explanation have been written for RGB color space, so please note, that R can be H, C, Y, or X, depending on the actual color space (G and B accordingly).

All values for all color spaces are limited to [0,1].

RGB stands for Red, Green and Blue; CMY stands for Cyan, Magenta and Yellow; HSV stands for Hue, Saturation, and Value; YIQ is the color model used by the U.S. Commercial Color Television Broadcasting, it is basically an RGB recoding with downward compatibility for black and white television; XYZ are the three primary colors of the color model defined by the ’Commission Internationale de l’Eclairage’ (CIE). For more information on color models see:  ncs/color/glossary.htm


43.50.1 Rgbformulae

For rgbformulae three suitable mapping functions have to be chosen. This is done via rgbformulae <r>,<g>,<b>. The available mapping functions are listed by show palette rgbformulae. Default is 7,5,15, some other examples are 3,11,6, 21,23,3 or 3,23,21. Negative numbers, like 3,-11,-6, mean inverted color (i.e. 1-gray passed into the formula, see also positive (p. 404) and negative (p. 404) options below).

Some nice schemes in RGB color space

  7,5,15   ... traditional pm3d (black-blue-red-yellow)  
  3,11,6   ... green-red-violet  
  23,28,3  ... ocean (green-blue-white); try also all other permutations  
  21,22,23 ... hot (black-red-yellow-white)  
  30,31,32 ... color printable on gray (black-blue-violet-yellow-white)  
  33,13,10 ... rainbow (blue-green-yellow-red)  
  34,35,36 ... AFM hot (black-red-yellow-white)

A full color palette in HSV color space

  3,2,2    ... red-yellow-green-cyan-blue-magenta-red

Please note that even if called rgbformulae the formulas might actually determine the <H>,<S>,<V> or <X>,<Y>,<Z> or ... color components as usual.

Use positive and negative to invert the figure colors.

Note that it is possible to find a set of the best matching rgbformulae for any other color scheme by the command

  show palette fit2rgbformulae

43.50.2 Defined

Gray-to-rgb mapping can be manually set by use of palette defined: A color gradient is defined and used to give the rgb values. Such a gradient is a piecewise linear mapping from gray values in [0,1] to the RGB space [0,1]x[0,1]x[0,1]. You have to specify the gray values and the corresponding RGB values in between a linear interpolation shall take place:


     set palette  defined { ( <gray1> <color1> {, <grayN> <colorN>}... ) }

<grayX> are gray values which are mapped to [0,1] and <colorX> are the corresponding rgb colors. The color can be specified in three different ways:

    <color> :=  { <r> <g> <b> | ’<color-name>’ | ’#rrggbb’ }

Either by three numbers (each in [0,1]) for red, green and blue, separated by whitespace, or the name of the color in quotes or X style color specifiers also in quotes. You may freely mix the three types in a gradient definition, but the named color "red" will be something strange if RGB is not selected as color space. Use show palette colornames for a list of known color names.

Please note, that even if written as <r>, this might actually be the <H> component in HSV color space or <X> in CIE-XYZ space, or ... depending on the selected color model.

The <gray> values have to form an ascending sequence of real numbers; the sequence will be automatically rescaled to [0,1].

set palette defined (without a gradient definition in braces) switches to RGB color space and uses a preset full-spectrum color gradient. Use show palette gradient to display the gradient.


To produce a gray palette (useless but instructive) use:

     set palette model RGB  
     set palette defined ( 0 "black", 1 "white" )

To produce a blue yellow red palette use (all equivalent):

     set palette defined ( 0 "blue", 1 "yellow", 2 "red" )  
     set palette defined ( 0 0 0 1, 1 1 1 0, 2 1 0 0 )  
     set palette defined ( 0 "#0000ff", 1 "#ffff00", 2 "#ff0000" )

To produce some rainbow-like palette use:

     set palette defined ( 0 "blue", 3 "green", 6 "yellow", 10 "red" )

Full color spectrum within HSV color space:

     set palette model HSV  
     set palette defined ( 0 0 1 1, 1 1 1 1 )  
     set palette defined ( 0 0 1 0, 1 0 1 1, 6 0.8333 1 1, 7 0.8333 0 1)

To produce a palette with few colors only use:

     set palette model RGB maxcolors 4  
     set palette defined ( 0 "blue", 1 "green", 2 "yellow", 3 "red" )

’Traffic light’ palette (non-smooth color jumps at gray = 1/3 and 2/3).

     set palette model RGB  
     set palette defined (0 "dark-green", 1 "green", 1 "yellow", \  
                          2 "dark-yellow", 2 "red", 3 "dark-red" )

43.50.3 Functions

Use set palette functions <Rexpr>, <Gexpr>, <Bexpr> to define three formulae for the R(gray), G(gray) and B(gray) mapping. The three formulae may depend on the variable gray which will take values in [0,1] and should also produce values in [0,1]. Please note that <Rexpr> might be a formula for the H-value if HSV color space has been chosen (same for all other formulae and color spaces).


To produce a full color palette use:

     set palette model HSV functions gray, 1, 1

A nice black to gold palette:

     set palette model XYZ functions gray**0.35, gray**0.5, gray**0.8

A gamma-corrected black and white palette

     gamma = 2.2  
     color(gray) = gray**(1./gamma)  
     set palette model RGB functions color(gray), color(gray), color(gray)

43.50.4 File

set palette file is basically a set palette defined (<gradient>) where <gradient> is read from a datafile. Either 4 columns (gray,R,G,B) or just three columns (R,G,B) have to be selected via the using data file modifier. In the three column case, the line number will be used as gray. The gray range is automatically rescaled to [0,1]. The file is read as a normal data file, so all datafile modifiers can be used. Please note, that R might actually be e.g. H if HSV color space is selected.

As usual <filename> may be ’-’ which means that the data follow the command inline and are terminated by a single e on a line of its own.

Use show palette gradient to display the gradient.


Read in a palette of RGB triples each in range [0,255]:

     set palette file ’some-palette’ using ($1/255):($2/255):($3/255)

Equidistant rainbow (blue-green-yellow-red) palette:

     set palette model RGB file "-"  
     0 0 1  
     0 1 0  
     1 1 0  
     1 0 0  

Binary palette files are supported as well, see binary general (p. 155). Example: put 64 triplets of R,G,B doubles into file palette.bin and load it by

     set palette file "palette.bin" binary record=64 using 1:2:3

43.50.5 Gamma correction

For gray mappings gamma correction can be turned on by set palette gamma <gamma>. <gamma> defaults to 1.5 which is quite suitable for most terminals.

For color mappings no automatic gamma correction is done by gnuplot. However, you may easily implement gamma correction. Here is an example for a gray scale image by use of explicit functions for the red, green and blue component with slightly different values of gamma


     set palette model RGB  
     set palette functions gray**0.64, gray**0.67, gray**0.70

To use gamma correction with interpolated gradients specify intermediate gray values with appropriate colors. Instead of

     set palette defined ( 0 0 0 0, 1 1 1 1 )

use e.g.

     set palette defined ( 0 0 0 0, 0.5 .73 .73 .73, 1 1 1 1 )

or even more intermediate points until the linear interpolation fits the "gamma corrected" interpolation well enough.

43.50.6 Postscript

In order to reduce the size of postscript files, the gray value and not all three calculated r,g,b values are written to the file. Therefore the analytical formulae are coded directly in the postscript language as a header just before the pm3d drawing, see /g and /cF definitions. Usually, it makes sense to write therein definitions of only the 3 formulae used. But for multiplot or any other reason you may want to manually edit the transformations directly in the postscript file. This is the default option nops_allcF. Using the option ps_allcF writes postscript definitions of all formulae. This you may find interesting if you want to edit the postscript file in order to have different palettes for different surfaces in one graph. Well, you can achieve this functionality by multiplot with fixed origin and size.

If pm3d map has been plotted from gridded or almost regular data with an output to a postscript file, then it is possible to reduce the size of this postscript file up to at about 50% by the enclosed awk script pm3dCompress.awk. This you may find interesting if you intend to keep the file for including it into your publication or before downloading a very large file into a slow printer. Usage:

   awk -f pm3dCompress.awk >

If pm3d map has been plotted from rectangular gridded data with an output to a postscript file, then it is possible to reduce the file size even more by the enclosed awk script pm3dConvertToImage.awk. Usage:

   awk -f pm3dConvertToImage.awk < >

You may manually change the postscript output from gray to color and vice versa and change the definition of <maxcolors>.

43.50.7 Colornames

Gnuplot knows a limited number of color names. You can use these to define the color range spanned by a pm3d palette, or to assign a terminal-independent color to a particular linetype or linestyle. To see the list of known color names, use the command show palette colornames (p. 424). See set palette (p. 400), linestyle (p. 450).

43.51 Pointsize

The set pointsize command scales the size of the points used in plots.


     set pointsize <multiplier>  
     show pointsize

The default is a multiplier of 1.0. Larger pointsizes may be useful to make points more visible in bitmapped graphics.

The pointsize of a single plot may be changed on the plot command. See plot with (p. 213) for details.

Please note that the pointsize setting is not supported by all terminal types.

43.52 Polar

The set polar command changes the meaning of the plot from rectangular coordinates to polar coordinates.


     set polar  
     unset polar  
     show polar

There have been changes made to polar mode in version 3.7, so that scripts for gnuplot versions 3.5 and earlier will require modification. The main change is that the dummy variable t is used for the angle so that the x and y ranges can be controlled independently. Other changes are: 1) tics are no longer put along the zero axes automatically — use set xtics axis nomirror; set ytics axis nomirror; 2) the grid, if selected, is not automatically polar — use set grid polar; 3) the grid is not labelled with angles — use set label as necessary.

In polar coordinates, the dummy variable (t) is an angle. The default range of t is [0:2*pi], or, if degree units have been selected, to [0:360] (see set angles (p. 240)).

The command unset polar changes the meaning of the plot back to the default rectangular coordinate system.

The set polar command is not supported for splots. See the set mapping (p. 372) command for similar functionality for splot (p. 539)s.

While in polar coordinates the meaning of an expression in t is really r = f(t), where t is an angle of rotation. The trange controls the domain (the angle) of the function, and the x and y ranges control the range of the graph in the x and y directions. Each of these ranges, as well as the rrange, may be autoscaled or set explicitly. See set xrange (p. 508) for details of all the ranges (p. 200) commands.


     set polar  
     plot t*sin(t)  
     plot [-2*pi:2*pi] [-3:3] [-3:3] t*sin(t)

The first plot uses the default polar angular domain of 0 to 2*pi. The radius and the size of the graph are scaled automatically. The second plot expands the domain, and restricts the size of the graph to [-3:3] in both directions.

You may want to set size square to have gnuplot try to make the aspect ratio equal to unity, so that circles look circular. See also

polar demos (polar.dem)


polar data plot (poldat.dem).

43.53 Print

The set print command redirects the output of the print command to a file.


     set print  
     set print "-"  
     set print "<filename>"  
     set print "<filename>" append  
     set print "|<shell_command>"

Without "<filename>", the output file is restored to <STDERR>. The <filename> "-" means <STDOUT>. The append flag causes the file to be opened in append mode. A <filename> starting with "|" is opened as a pipe to the <shell_command> on platforms that support piping.

43.54 Object

This command defines a single object, which will appear in all subsequent 2D plots. You may define as many objects as you like. Currently the only object type supported is rectangle. Each rectangle is specified by a pair of points that define diagonal vertices. A default set of style properties (fill, color, border) are inherited from those set by the command set style rectangle, but each rectangle can also be given individual style properties.


   set object <index> rectangle  
       {from <position> {to|rto} <position> |  
        center <position> size <w>,<h> |  
        at <position> size <w>,<h>}  
       {front|back|behind} {fc|fillcolor <colorspec>} {fs <fillstyle>}  
       {default} {lw|linewidth <width>}

The position of the rectangle may be specified by giving the position of two diagonal corners (bottom left and top right) or by giving the position of the center followed by the width and the height. In either case the positions may be given in axis, graph, or screen coordinates. See coordinates (p. 41). The options at and center are synonyms.

Setting front will draw the rectangle in front of all plot elements, but behind any labels that are also marked front. Setting back will place the rectangle behind all plot curves and labels. Setting behind will place the rectangle behind everything including the axes and back rectangles, and can be used to provide a colored background for the entire graph or page.

The fill color of the rectangle is taken from the <colorspec>. fillcolor may be abbreviated fc. The fill style is taken from <fillstyle>. See colorspec (p. 68) and fillstyle (p. 445). If the keyword default is given, these properties are inherited from the default settings of at the time a plot is drawn. See set style rectangle (p. 456).


   # Force the entire area enclosed by the axes to have background color cyan  
   set object 1 rect from graph 0, graph 0 to graph 1, graph 1 back  
   set object 1 rect fc rgb "cyan" fillstyle solid 1.0

   # Position a red square with lower left at 0,0 and upper right at 2,3  
   set object 2 rect from 0,0 to 2,3 fc lt 1

   # Position an empty rectangle (no fill) with a blue border  
   set object 3 rect from 0,0 to 2,3 fs empty border 3

   # Return fill and color to the default style but leave vertices unchanged  
   set object 2 rect default

43.55 Rmargin

The command set rmargin sets the size of the right margin. Please see set margin (p. 375) for details.

43.56 Rrange

The set rrange command sets the range of the radial coordinate for a graph in polar mode. Please see set xrange (p. 508) for details.

43.57 Samples

The sampling rate of functions, or for interpolating data, may be changed by the set samples command.


     set samples <samples_1> {,<samples_2>}  
     show samples

By default, sampling is set to 100 points. A higher sampling rate will produce more accurate plots, but will take longer. This parameter has no effect on data file plotting unless one of the interpolation/approximation options is used. See plot smooth (p. 171) re 2-d data and set cntrparam (p. 281) and set dgrid3d (p. 312) re 3-d data.

When a 2-d graph is being done, only the value of <samples_1> is relevant.

When a surface plot is being done without the removal of hidden lines, the value of samples specifies the number of samples that are to be evaluated for the isolines. Each iso-v line will have <sample_1> samples and each iso-u line will have <sample_2> samples. If you only specify <samples_1>, <samples_2> will be set to the same value as <samples_1>. See also set isosamples (p. 335).

43.58 Size


     set size {{no}square | ratio <r> | noratio} {<xscale>,<yscale>}  
     show size

The <xscale> and <yscale> values are scale factors for the size of the plot, which includes the graph, labels, and margins.

Important note:

     In earlier versions of gnuplot, some terminal types used the values from  
     ‘set size‘ to control also the size of the output canvas; others did not.  
     In version 4.2 almost all terminals now follow the following convention:

set term <terminal_type> size <XX>, <YY> controls the size of the output file, or canvas. Please see individual terminal documentation for allowed values of the size parameters. By default, the plot will fill this canvas.

set size <XX>, <YY> scales the plot itself relative to the size of the canvas. Scale values less than 1 will cause the plot to not fill the entire canvas. Scale values larger than 1 will cause only a portion of the plot to fit on the canvas. Please be aware that setting scale values larger than 1 may cause problems on some terminal types.

ratio causes gnuplot to try to create a graph with an aspect ratio of <r> (the ratio of the y-axis length to the x-axis length) within the portion of the plot specified by <xscale> and <yscale>.

The meaning of a negative value for <r> is different. If <r>=-1, gnuplot tries to set the scales so that the unit has the same length on both the x and y axes (suitable for geographical data, for instance). If <r>=-2, the unit on y has twice the length of the unit on x, and so on.

The success of gnuplot in producing the requested aspect ratio depends on the terminal selected. The graph area will be the largest rectangle of aspect ratio <r> that will fit into the specified portion of the output (leaving adequate margins, of course).

square is a synonym for ratio 1.

Both noratio and nosquare return the graph to the default aspect ratio of the terminal, but do not return <xscale> or <yscale> to their default values (1.0).

ratio and square have no effect on 3-d plots, but do affect 3D projections created using set view map. Similarly set view equal forces the x and y axes of a 3D onto the same scale.


To set the size so that the plot fills the available canvas:

     set size 1,1

To make the graph half size and square use:

     set size square 0.5,0.5

To make the graph twice as high as wide use:

     set size ratio 2

See also

airfoil demo.

43.59 Style

Default plotting styles are chosen with the set style data and set style function commands. See plot with (p. 213) for information about how to override the default plotting style for individual functions and data sets. See plotting styles (p. 456) for a complete list of styles.


     set style function <style>  
     set style data <style>  
     show style function  
     show style data

Default styles for specific plotting elements may also be set.


     set style arrow <n> <arrowstyle>  
     set style fill <fillstyle>  
     set style histogram <histogram style options>  
     set style line <n> <linestyle>

43.59.1 Set style arrow

Each terminal has a default set of arrow and point types, which can be seen by using the command test. set style arrow defines a set of arrow types and widths and point types and sizes so that you can refer to them later by an index instead of repeating all the information at each invocation.


     set style arrow <index> default  
     set style arrow <index> {nohead | head | heads}  
                             {size <length>,<angle>{,<backangle>}}  
                             {filled | empty | nofilled}  
                             {front | back}  
                             { {linestyle | ls <line_style>}  
                               | {linetype | lt <line_type>}  
                                 {linewidth | lw <line_width} }  
     unset style arrow  
     show style arrow

<index> is an integer that identifies the arrowstyle.

If default is given all arrow style parameters are set to their default values.

If the linestyle <index> already exists, only the given parameters are changed while all others are preserved. If not, all undefined values are set to the default values.

Specifying nohead produces arrows drawn without a head — a line segment. This gives you yet another way to draw a line segment on the plot. By default, arrows have one head. Specifying heads draws arrow heads on both ends of the line.

Head size can be controlled by size <length>,<angle> or size <length>,<angle>,<backangle>, where <length> defines length of each branch of the arrow head and <angle> the angle (in degrees) they make with the arrow. <Length> is in x-axis units; this can be changed by first, second, graph, screen, or character before the <length>; see coordinates (p. 41) for details. <Backangle> only takes effect when filled or empty is also used. Then, <backangle> is the angle (in degrees) the back branches make with the arrow (in the same direction as <angle>). The fig terminal has a restricted backangle function. It supports three different angles. There are two thresholds: Below 70 degrees, the arrow head gets an indented back angle. Above 110 degrees, the arrow head has an acute back angle. Between these thresholds, the back line is straight.

Specifying filled produces filled arrow heads (if heads are used). Filling is supported on filled-polygon capable terminals, see help of pm3d (p. 390) for their list, otherwise the arrow heads are closed but not filled. The same result (closed but not filled arrow head) is reached by specifying empty. Further, filling and outline is obviously not supported on terminals drawing arrows by their own specific routines, like metafont, metapost, latex or tgif.

The line style may be selected from a user-defined list of line styles (see set style line (p. 450)) or may be defined here by providing values for <line_type> (an index from the default list of styles) and/or <line_width> (which is a multiplier for the default width).

Note, however, that if a user-defined line style has been selected, its properties (type and width) cannot be altered merely by issuing another set style arrow command with the appropriate index and lt or lw.

If front is given, the arrows are written on top of the graphed data. If back is given (the default), the arrow is written underneath the graphed data. Using front will prevent a arrow from being obscured by dense data.


To draw an arrow without an arrow head and double width, use:

     set style arrow 1 nohead lw 2  
     set arrow arrowstyle 1

See also ‘set arrow‘ for further examples.

43.59.2 Set style data

The set style data command changes the default plotting style for data plots.


     set style data <plotting-style>  
     show style data

See plotting styles (p. 456) for the choices. If no choice is given, the choices are listed. show style data shows the current default data plotting style.

43.59.3 Set style fill

The set style fill command is used to set the style of boxes, histograms, candlesticks and filledcurves.


     set style fill {empty | solid {<density>} | pattern {<n>}}  
                    {border {<linetype>} | noborder}

The default fillstyle is empty.

The solid option causes filling with a solid color, if the terminal supports that. The <density> parameter specifies the intensity of the fill color. At a <density> of 0.0, the box is empty, at <density> of 1.0, the inner area is of the same color as the current linetype. Some terminal types can vary the density continuously; others implement only a few levels of partial fill. If no <density> parameter is given, it defaults to 1.

The pattern option causes filling to be done with a fill pattern supplied by the terminal driver. The kind and number of available fill patterns depend on the terminal driver. If multiple datasets using filled boxes are plotted, the pattern cycles through all available pattern types, starting from pattern <n>, much as the line type cycles for multiple line plots.

The empty option causes filled boxes not to be filled. This is the default.

By default, border, the box is bounded by a solid line of the current linetype. border <lt> specifies that a border is to be drawn using linetype <lt>. noborder specifies that no bounding lines are drawn.

43.59.4 Set style function

The set style function command changes the default plotting style for function plots.


     set style function <plotting-style>  
     show style function

See plotting styles (p. 456) for the choices. If no choice is given, the choices are listed. show style function shows the current default function plotting style.

43.59.5 Set style increment


     set style increment {default|userstyles}  
     show style increment

By default, successive plots within the same graph will use successive linetypes from the default set for the current terminal type. However, choosing set style increment user allows you to step through the user-defined line styles rather than through the default linetypes.


     set style line 1 lw 2 lc rgb "gold"  
     set style line 2 lw 2 lc rgb "purple"  
     set style line 4 lw 1 lc rgb "sea-green"  
     set style increment user

     plot f1(x), f2(x), f3(x), f4(x)

should plot functions f1, f2, f4 in your 3 newly defined line styles. If a user-defined line style is not found then the corresponding default linetype is used instead. E.g. in the example above, f3(x) will be plotted using the default linetype 3.

43.59.6 Set style line

Each terminal has a default set of line and point types, which can be seen by using the command test. set style line defines a set of line types and widths and point types and sizes so that you can refer to them later by an index instead of repeating all the information at each invocation.


     set style line <index> default  
     set style line <index> {{linetype  | lt} <line_type> | <colorspec>}  
                            {{linecolor | lc} <colorspec>}  
                            {{linewidth | lw} <line_width>}  
                            {{pointtype | pt} <point_type>}  
                            {{pointsize | ps} <point_size>}  
     unset style line  
     show style line

If default is given all line style parameters are set to their default values.

If the linestyle <index> already exists, only the given parameters are changed while all others are preserved. If not, all undefined values are set to the default values.

The line and point types are taken from the default types for the terminal currently in use. The line width and point size are multipliers for the default width and size (but note that <point_size> here is unaffected by the multiplier given on set pointsize).

The defaults for the line and point types is the index. The defaults for the width and size are both unity.

Linestyles created by this mechanism do not replace the default linetype styles; both may be used. If you want plots to use the defined styles in preference to the default linetypes, please see set style increment (p. 447).

Not all terminals support the linewidth and pointsize features; if not supported, the option will be ignored.

Terminal-independent colors may be assigned using either linecolor <colorspec> or linetype <colorspec>, abbreviated lc or lt. This requires giving a RGB color triple, a known palette color name, a fractional index into the current palette, or a constant value from the current mapping of the palette onto cbrange. See colors (p. 64), colorspec (p. 68), set palette (p. 400), colornames (p. 424), cbrange (p. 536).

set style line <n> linetype <lt> will set both a terminal-dependent dot/dash pattern and color. The commandsset style line <n> linecolor <colorspec> or set style line <n> linetype <colorspec> will set a new line color while leaving the existing dot-dash pattern unchanged.

In 3d mode (splot command), the special keyword palette is allowed as a shorthand for "linetype palette z". The color value corresponds to the z-value (elevation) of the splot, and varies smoothly along a line or surface.

Examples: Suppose that the default lines for indices 1, 2, and 3 are red, green, and blue, respectively, and the default point shapes for the same indices are a square, a cross, and a triangle, respectively. Then

     set style line 1 lt 2 lw 2 pt 3 ps 0.5

defines a new linestyle that is green and twice the default width and a new pointstyle that is a half-sized triangle. The commands

     set style function lines  
     plot f(x) lt 3, g(x) ls 1

will create a plot of f(x) using the default blue line and a plot of g(x) using the user-defined wide green line. Similarly the commands

     set style function linespoints  
     plot p(x) lt 1 pt 3, q(x) ls 1

will create a plot of p(x) using the default triangles connected by a red line and q(x) using small triangles connected by a green line.

     splot sin(sqrt(x*x+y*y))/sqrt(x*x+y*y) w l pal

creates a surface plot using smooth colors according to palette. Note, that this works only on some terminals. See also set palette (p. 400), set pm3d (p. 390).

     set style line 10 linetype 1 linecolor rgb "cyan"

will assign linestyle 10 to be a solid cyan line on any terminal that supports rgb colors.

43.59.7 Plotting styles

The commands set style data and set style function change the default plotting style for subsequent plot and splot commands.

The types used for all line and point styles (i.e., solid, dash-dot, color, etc. for lines; circles, squares, crosses, etc. for points) will be either those specified on the plot or splot command or will be chosen sequentially from the types available to the terminal in use. Use the command test to see what is available.

None of the styles requiring more than two columns of information (e.g., errorbars or errorlines) can be used with splots or function plots. Neither boxes, filledcurves nor any of the steps styles can be used with splots. If an inappropriate style is specified, it will be changed to points.

The above caveat does not apply to plot with labels, for which the third column specifies a data source rather than coordinate information. See set style labels (p. 482).

For 2-d data with more than two columns, gnuplot is picky about the allowed errorbars and errorlines styles. The using option on the plot command can be used to set up the correct columns for the style you want. (In this discussion, "column" will be used to refer both to a column in the data file and an entry in the using list.)

For three columns, only xerrorbars, yerrorbars (or errorbars), xerrorlines, yerrorlines (or errorlines), boxes, and boxerrorbars are allowed. If another plot style is used, the style will be changed to yerrorbars. The boxerrorbars style will calculate the boxwidth automatically.

For four columns, only xerrorbars, yerrorbars (or errorbars), xyerrorbars, xerrorlines, yerrorlines (or errorlines), xyerrorlines, boxxyerrorbars, and boxerrorbars are allowed. An illegal style will be changed to yerrorbars.

Five-column data allow only the boxerrorbars, financebars, and candlesticks styles. An illegal style will be changed to boxerrorbars before plotting.

Six- and seven-column data only allow the xyerrorbars, xyerrorlines, and boxxyerrorbars styles. Illegal styles will be changed to xyerrorbars before plotting.

For more information about error bars with and without lines, please see plot errorlines (p. 196) and plot errorbars (p. 194).

43.59.8 Set style rectangle

Rectangles defined with the set object rectangle command can have individual styles. However, if a rectangle is not assigned a private style then it inherits a default that is taken from the set style rectangle command.


   set style rectangle {front|back} {fillcolor <colorspec>} {fs <fillstyle>}  
                       {lw|linewidth <lw>}

See colorspec (p. 68) and fillstyle (p. 445). fillcolor may be abbreviated as fc.


   set style rectangle back fc rgb "white" fs solid 1.0 border -1  
   set style rectangle fc linsestyle 3 fs pattern 2 noborder

The default values correspond to solid fill with the background color and a black border.

Boxerrorbars The boxerrorbars style is only relevant to 2-d data plotting. It is a combination of the boxes and yerrorbars styles. The boxwidth will come from the fourth column if the y errors are in the form of "ydelta" and the boxwidth was not previously set equal to -2.0 (set boxwidth -2.0) or from the fifth column if the y errors are in the form of "ylow yhigh". The special case boxwidth = -2.0 is for four-column data with y errors in the form "ylow yhigh". In this case the boxwidth will be calculated so that each box touches the adjacent boxes. The width will also be calculated in cases where three-column data are used.

The box height is determined from the y error in the same way as it is for the yerrorbars style — either from y-ydelta to y+ydelta or from ylow to yhigh, depending on how many data columns are provided. See also

errorbar demo.

Boxes The boxes style is only relevant to 2-d plotting. It draws a box centered about the given x coordinate from the x axis (not the graph border) to the given y coordinate. The width of the box is obtained in one of three ways. If it is a data plot and the data file has a third column, this will be used to set the width of the box. If not, if a width has been set using the set boxwidth command, this will be used. If neither of these is available, the width of each box will be calculated automatically so that it touches the adjacent boxes.

The interior of the boxes is drawn according to the current fillstyle. See set style fill (p. 445) for details. Alternatively a new fillstyle may be specified in the plot command.

For fillstyle empty the box is filled with the background color.

For fillstyle solid the box is filled with a solid rectangle of the current drawing color. There is an optional parameter <density> that controls the fill density; it runs from 0 (background color) to 1 (current drawing color).

For fillstyle pattern the box is filled in the current drawing color with a pattern, if supported by the terminal driver.


To plot a data file with solid filled boxes with a small vertical space separating them (bargraph):

     set boxwidth 0.9 relative  
     set style fill solid 1.0  
     plot ’file.dat’ with boxes

To plot a sine and a cosine curve in pattern-filled boxes style:

     set style fill pattern  
     plot sin(x) with boxes, cos(x) with boxes

The sin plot will use pattern 0; the cos plot will use pattern 1. Any additional plots would cycle through the patterns supported by the terminal driver.

To specify explicit fillstyles for each dataset:

    plot ’file1’ with boxes fs solid 0.25, \  
         ’file2’ with boxes fs solid 0.50, \  
         ’file3’ with boxes fs solid 0.75, \  
         ’file4’ with boxes fill pattern 1, \  
         ’file5’ with boxes fill empty

Currently only the following terminal drivers support fillstyles other than empty: x11, windows, pm, wxt, postscript, fig, pbm, png, gif, hpdj, hppj, hpljii, hp500c, jpeg, nec_cp6, epson_180dpi, epson_60dpi, epson_lx800, okidata, starc and tandy_60dpi. The BeOS driver (be) is untested.

Boxxyerrorbars The boxxyerrorbars style is only relevant to 2-d data plotting. It is a combination of the boxes and xyerrorbars styles.

The box width and height are determined from the x and y errors in the same way as they are for the xyerrorbars style — either from xlow to xhigh and from ylow to yhigh, or from x-xdelta to x+xdelta and from y-ydelta to y+ydelta , depending on how many data columns are provided.

If filled-box support is present, then the interior of the boxes is drawn according to the current fillstyle. See set style fill (p. 445) and boxes (p. 458) for details. Alternatively a new fillstyle may be specified in the plot command.

Candlesticks The candlesticks style can be used for 2-d data plotting of financial data or for generating box-and-whisker plots of statistical data. Five columns of data are required; in order, these should be the x coordinate (most likely a date) and the opening, low, high, and closing prices. The symbol is a rectangular box, centered horizontally at the x coordinate and limited vertically by the opening and closing prices. A vertical line segment at the x coordinate extends up from the top of the rectangle to the high price and another down to the low. The vertical line will be unchanged if the low and high prices are interchanged.

The width of the rectangle can be controlled by the set boxwidth command. For backwards compatibility with earlier gnuplot versions, when the boxwidth parameter has not been set then the width of the candlestick rectangle is controlled by set bars <width>.

By default the vertical line segments have no crossbars at the top and bottom. If you want crossbars, which are typically used for box-and-whisker plots, then add the keyword whiskerbars to the plot command. By default these whiskerbars extend the full horizontal width of the candlestick, but you can modify this by specifying a fraction of the full width.

By default the rectangle is empty if (open < close), and filled with three vertical bars if (close < open). If filled-boxes support is present, then the rectangle is colored according to set style fill <fillstyle>. See set bars (p. 263) and financebars (p. 470). See also

finance demos


Note: To place additional symbols, such as the median value, on a box-and-whisker plot requires additional plot commands as in this example:

 # Data columns: X Min 1stQuartile Median 3rdQuartile Max  
 set bars 4.0  
 set style fill empty  
 plot ’stat.dat’ using 1:3:2:6:5 with candlesticks title ’Quartiles’, \  
      ’’         using 1:4:4:4:4 with candlesticks lt -1 notitle

 # Plot with crossbars on the whiskers, crossbars are 50% of full width  
 plot ’stat.dat’ using 1:3:2:6:5 with candlesticks whiskerbars 0.5

See ‘set boxwidth‘, ‘set bars‘ and ‘set style fill‘.

Dots The dots style plots a tiny dot at each point; this is useful for scatter plots with many points. For some terminals (post, pdf) the size of the dot can be controlled by changing the linewidth.

Filledcurves The filledcurves style is only relevant to 2-d plotting. Three variants are possible. The first two variants require either a function or two columns of input data, and may be further modified by the options listed below. The first variant, closed, treats the curve itself as a closed polygon. This is the default if there are two columns of input data.

The second variant is to fill the area between the curve and a given axis, a horizontal or vertical line, or a point.

The third variant requires three columns of input data: the x coordinate and two y coordinates corresponding to two curves sampled at the same set of x coordinates; the area between the two curves is filled. This is the default if there are three or more columns of input data.


   set style [data | function] filledcurves [option]  
   plot ... with filledcurves [option]

where the option can be

   [closed | {above | below} {x1 | x2 | y1 | y2}[=<a>] | xy=<x>,<y>]

The first two plot variants can be further modified by the options

   filledcurves closed   ... just filled closed curve,  
   filledcurves x1       ... x1 axis,  
   filledcurves x2       ... x2 axis, etc for y1 and y2 axes,  
   filledcurves y1=0     ... line y=0 (at y1 axis) ie parallel to x1 axis,  
   filledcurves y2=42    ... line y=42 (at y2 axis) ie parallel to x2, etc,  
   filledcurves xy=10,20 ... point 10,20 of x1,y1 axes (arc-like shape).

Example of filling the area between two input curves.

fill between curves demo.

   plot ’data’ using 1:2:3 with filledcurves

The above and below options apply both to commands of the form

   ... filledcurves above {x1|x2|y1|y2}=<val>

and to commands of the form

   ... using 1:2:3 with filledcurves below

In either case the option limits the filled area to one side of the bounding line or curve.

Note: Not all terminal types support this plotting mode.

Zoom of a filled curve drawn from a datafile may produce empty or incorrect area because gnuplot is clipping points and lines, and not areas.

If the values of <a>, <x>, <y> are out of the drawing boundary, then they are moved to the graph boundary. Then the actually filled area in the case of option xy=<x>,<y> will depend on xrange and yrange.

Financebars The financebars style is only relevant for 2-d data plotting of financial data. Five columns of data are required; in order, these should be the x coordinate (most likely a date) and the opening, low, high, and closing prices. The symbol is a vertical line segment, located horizontally at the x coordinate and limited vertically by the high and low prices. A horizontal tic on the left marks the opening price and one on the right marks the closing price. The length of these tics may be changed by set bars. The symbol will be unchanged if the high and low prices are interchanged. See set bars (p. 263) and candlesticks (p. 461), and also the

finance demo.

Fsteps The fsteps style is only relevant to 2-d plotting. It connects consecutive points with two line segments: the first from (x1,y1) to (x1,y2) and the second from (x1,y2) to (x2,y2). See also

steps demo.

Histeps The histeps style is only relevant to 2-d plotting. It is intended for plotting histograms. Y-values are assumed to be centered at the x-values; the point at x1 is represented as a horizontal line from ((x0+x1)/2,y1) to ((x1+x2)/2,y1). The lines representing the end points are extended so that the step is centered on at x. Adjacent points are connected by a vertical line at their average x, that is, from ((x1+x2)/2,y1) to ((x1+x2)/2,y2).

If autoscale is in effect, it selects the xrange from the data rather than the steps, so the end points will appear only half as wide as the others. See also

steps demo.

histeps is only a plotting style; gnuplot does not have the ability to create bins and determine their population from some data set.

Histograms The histograms style is only relevant to 2-d plotting. It produces a bar chart from a sequence of data columns in parallel. Each element of the plot command must specify a single input data source (e.g. one column of the input file), possibly with associated tic values or key titles. Four styles of histogram layout are currently supported.

     set style histogram clustered {gap <gapsize>}  
     set style histogram errorbars {gap <gapsize>} {<linewidth>}  
     set style histogram rowstacked  
     set style histogram columnstacked

The default style corresponds to set style histogram clustered gap 2. In this style, each set of parallel data values is collected into a group of boxes clustered at the x-axis coordinate corresponding to their sequential position (row #) in the selected datafile columns. Thus if <n> datacolumns are selected, the first cluster is centered about x=1, and contains <n> boxes whose heights are taken from the first entry in the corresponding <n> data columns. This is followed by a gap and then a second cluster of boxes centered about x=2 corresponding to the second entry in the respective data columns, and so on. The default gap width of 2 indicates that the empty space between clusters is equivalent to the width of 2 boxes. All boxes derived from any one column are given the same fill color and/or pattern (see set style fill (p. 445)).

Each cluster of boxes is derived from a single row of the input data file. It is common in such input files that the first element of each row is a label. Labels from this column may be placed along the x-axis underneath the appropriate cluster of boxes with the xticlabels option to using.

The errorbars style is very similar to the clustered style, except that it requires two columns of input for each entry. The first column is treated as the height (y-value) of that box, exactly as for the clustered style. The second column is treated as an error magnitude, and used to generate a vertical error bar at the top of the box. The appearance of the error bar is controlled by the current value of set bars and by the optional <linewidth> specification.

Two styles of stacked histogram are supported, chosen by the command set style histogram {rowstacked|columnstacked}. In these styles the data values from the selected columns are collected into stacks of boxes. Positive values stack upwards from y=0; negative values stack downwards. Mixed positive and negative values will produce both an upward stack and a downward stack. The default stacking mode is rowstacked.

The rowstacked style places a box resting on the x-axis for each data value in the first selected column; the first data value results in a box a x=1, the second at x=2, and so on. Boxes corresponding to the second and subsequent data columns are layered on top of these, resulting in a stack of boxes at x=1 representing the first data value from each column, a stack of boxes at x=2 representing the second data value from each column, and so on. All boxes derived from any one column are given the same fill color and/or pattern (see set style fill (p. 445)).

The columnstacked style is similar, except that each stack of boxes is built up from a single data column. Each data value from the first specified column yields a box in the stack at x=1, each data value from the second specified column yields a box in the stack at x=2, and so on. In this style the color of each box is taken from the row number, rather than the column number, of the corresponding data field.

Box widths may be modified using the set boxwidth command. Box fill styles may be set using the set style fill command.

Histograms always use the x1 axis, but may use either y1 or y2. If a plot contains both histograms and other plot styles, the non-histogram plot elements may use either the x1 or the x2 axis.


To plot a data file containing multiple columns of data as a histogram of clustered boxes (the default style):

     set boxwidth 0.9 relative  
     set style data histograms  
     set style fill solid 1.0 border -1  
     plot ’file.dat’ using 2, ’’ using 4, ’’ using 6

This will produce a plot with clusters of three boxes (vertical bars) centered at each integral value on the x axis. If the first column of the input file contains labels, they may be placed along the x-axis using the variant command

     plot ’file.dat’ using 2, ’’ using 4, ’’ using 6:xticlabels(1)

If the file contains both a magnitude and an error estimate for each value, then error bars can be added to the plot. The following commands will add error bars extending from (y-<error>) to (y+<error>), capped by horizontal bar ends drawn the same width as the box itself. The error bars and bar ends are drawn with linewidth 2 using the border linetype from the current fill style.

     set bars fullwidth  
     set style fill solid border -1  
     set style histogram errorbars gap 2 lw 2  
     plot ’file.dat’ using 2:3, ’’ using 4:5, ’’ using 6:7:xticlabels(1)

To plot the same data as a rowstacked histogram:

     set style histogram rows  
     plot ’file.dat’ using 2, ’’ using 4, ’’ using 6:xtic(1)

This will produce a plot in which each vertical bar contains a stack of three segments, corresponding in height to the values found in columns 2, 4 and 6 of the datafile.

Finally, the commands

     set style histogram columnstacked  
     plot ’file.dat’ using 2, ’’ using 4, ’’ using 6

will produce three vertical stacks. The stack at x=1 will contain a box for each entry in column 2 of the datafile. The stack at x=2 will contain a box for each parallel entry in column 4 of the datafile, and the stack at x=3 a box for each entry of column 6. Because this interchanges gnuplot’s usual interpretation of input rows and columns, the specification of key titles and x-axis tic labels must also be modified.

     set style histogram columnstacked  
     plot ’’ u 5:key(1)            # uses first column to generate key titles  
     plot ’’ u 5 title columnhead  # uses first row to generate xtic labels

Newhistogram More than one set of histograms can appear in a single plot. In this case you can force a gap between them, and a separate label for each set, by using the plot command newhistogram {"<title>"} {<linetype>}{at <x-coord>}. For example

     set style histogram  cluster  
     plot newhistogram "Set A", ’a’ using 1, ’’ using 2, ’’ using 3, \  
          newhistogram "Set B", ’b’ using 1, ’’ using 2, ’’ using 3

The labels "Set A" and "Set B" will appear beneath the respective sets of histograms, under the overall x axis label.

The newhistogram command can also be used to force histogram coloring to begin with a specific color (linetype). By default colors will continue to increment successively even across histogram boundaries. Here is an example using the same coloring for multiple histograms

     plot newhistogram "Set A" lt 4, ’a’ using 1, ’’ using 2, ’’ using 3, \  
          newhistogram "Set B" lt 4, ’b’ using 1, ’’ using 2, ’’ using 3

The at <x-coord> option only applies to column-stacked histograms.

Image The image style is intendend for plotting 2D images. It may be used for both plot and splot in the form of 3D data (x,y,value) or projected 4D data (x,y,z,value), respectively. It is assumed that in the viewing plane the image data forms an equidistant sampling grid in the viewing plane along two, not necessarily orthogonal, directions. In other words, groups of four adjacent points are assumed to form the same size parallelogram. The variable value in the tuples represent a palette color (gray value) for indexing in the current palette.

The image style will attempt to create a properly positioned and scaled data matrix to match the plot borders for those terminals supporting palettes and images. Such output is efficient and draws quickly. However, when a terminal driver does not support palettes and images, or when image support is not implemented, the image style reverts to drawing filled rectangular boxes for pixels, which is not as efficient. General parallelogram-shaped images currently always have filled parallelograms for pixels.

The coordinate of each data point of an image will lie at the center of a pixel. That is, an M x N set of data will form an image with M x N pixels. This is slightly different than pm3d elements where an M x N set of data will form a surface of (M-1) x (N-1) elements. The scan directions for the image data grid can be any of eight possible combinations.

Here are some specific comments about particular terminal drivers:

x11 and wxt - Pixels are either repeated or decimated to fit the display

     resolution; no other processing (filtering) is done.  Thus, aliasing may  
     occur when decimating images having high spatial frequency content.

postscript (pslatex, epslatex, pstex) - Image is copied in its original

     resolution, and sample interpolation is turned off.

See also rgbimage (p. 484).

Impulses The impulses style displays a vertical line from the x axis (not the graph border), or from the grid base for splot, to each point.

Labels The labels style is available only if gnuplot is built with configuration option –enable-datastrings. For a 2-D plot with labels you must specify 3 input data columns; the text string found in the third column is printed at the X and Y coordinates generated by the first two column specifiers. The font, color, rotation angle and other properties of the printed text may be specified as additional command options (see set label (p. 351)). The example below will generate a 2-D plot with text labels taken from column 4 of the input file (tc lt 2 is shorthand for textcolor linetype 2, which is green).

 plot ’datafile’ using 1:(0.5 * $2):4 with labels font "arial,11" tc lt 2

The labels style can also be used in 3-D plots. In this case four input column specifiers are required, corresponding to X Y Z and text.

 splot ’datafile’ using 1:2:3:4 with labels

See also datastrings (p. 43), set style data (p. 444).

Lines The lines style connects adjacent points with straight line segments. See also linetype (p. 63), linewidth (p. 450), and linestyle (p. 450).

Linespoints The linespoints style does both lines and points, that is, it draws a small symbol at each point and then connects adjacent points with straight line segments. The command set pointsize may be used to change the size of the points. See set pointsize (p. 424) for its usage.

linespoints may be abbreviated lp.

Points The points style displays a small symbol at each point. The command set pointsize may be used to change the size of the points. See set pointsize (p. 424) for its usage.

Steps The steps style is only relevant to 2-d plotting. It connects consecutive points with two line segments: the first from (x1,y1) to (x2,y1) and the second from (x2,y1) to (x2,y2). See also

steps demo.

Rgbimage The rgbimage style is intended for plotting 2D images and is similar in concept to image. See image (p. 480) for details. The difference is that 5D data (x,y,r,g,b) for plot and 6D data (x,y,z,r,g,b) for splot describe the coordinates and color components of an image.

See also image (p. 480).

Vectors The 2D vectors style draws a vector from (x,y) to (x+xdelta,y+ydelta). Thus it requires four columns of data. It also draws a small arrowhead at the end of the vector. The 3D vectors style is similar, but requires six columns of data. splot with vectors is supported only for set mapping cartesian. The keywords "with vectors" may be followed by arrow style specifications. See arrowstyle (p. 441) for more details.


     plot ’file.dat’ using 1:2:3:4 with vectors head filled lt 2  
     splot ’file.dat’ using 1:2:3:(1):(1):(1) with vectors filled head lw 2

set clip one and set clip two affect vectors drawn in 2D. Please see set clip (p. 278) and arrowstyle (p. 441).

Xerrorbars The xerrorbars style is only relevant to 2-d data plots. xerrorbars is like dots, except that a horizontal error bar is also drawn. At each point (x,y), a line is drawn from (xlow,y) to (xhigh,y) or from (x-xdelta,y) to (x+xdelta,y), depending on how many data columns are provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details).

Xyerrorbars The xyerrorbars style is only relevant to 2-d data plots. xyerrorbars is like dots, except that horizontal and vertical error bars are also drawn. At each point (x,y), lines are drawn from (x,y-ydelta) to (x,y+ydelta) and from (x-xdelta,y) to (x+xdelta,y) or from (x,ylow) to (x,yhigh) and from (xlow,y) to (xhigh,y), depending upon the number of data columns provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details).

If data are provided in an unsupported mixed form, the using filter on the plot command should be used to set up the appropriate form. For example, if the data are of the form (x,y,xdelta,ylow,yhigh), then you can use

     plot ’data’ using 1:2:($1-$3):($1+$3):4:5 with xyerrorbars

Yerrorbars The yerrorbars (or errorbars) style is only relevant to 2-d data plots. yerrorbars is like points, except that a vertical error bar is also drawn. At each point (x,y), a line is drawn from (x,y-ydelta) to (x,y+ydelta) or from (x,ylow) to (x,yhigh), depending on how many data columns are provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details). See also

errorbar demo.

Xerrorlines The xerrorlines style is only relevant to 2-d data plots. xerrorlines is like linespoints, except that a horizontal error line is also drawn. At each point (x,y), a line is drawn from (xlow,y) to (xhigh,y) or from (x-xdelta,y) to (x+xdelta,y), depending on how many data columns are provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details).

Xyerrorlines The xyerrorlines style is only relevant to 2-d data plots. xyerrorlines is like linespoints, except that horizontal and vertical error bars are also drawn. At each point (x,y), lines are drawn from (x,y-ydelta) to (x,y+ydelta) and from (x-xdelta,y) to (x+xdelta,y) or from (x,ylow) to (x,yhigh) and from (xlow,y) to (xhigh,y), depending upon the number of data columns provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details).

If data are provided in an unsupported mixed form, the using filter on the plot command should be used to set up the appropriate form. For example, if the data are of the form (x,y,xdelta,ylow,yhigh), then you can use

     plot ’data’ using 1:2:($1-$3):($1+$3):4:5 with xyerrorlines

Yerrorlines The yerrorlines (or errorlines) style is only relevant to 2-d data plots. yerrorlines is like linespoints, except that a vertical error line is also drawn. At each point (x,y), a line is drawn from (x,y-ydelta) to (x,y+ydelta) or from (x,ylow) to (x,yhigh), depending on how many data columns are provided. A tic mark is placed at the ends of the error bar (unless set bars is used — see set bars (p. 263) for details). See also

errorbar demo.

43.60 Surface

The command set surface controls the display of surfaces by splot.


     set surface  
     unset surface  
     show surface

The surface is drawn with the style specified by with, or else the appropriate style, data or function.

Whenever unset surface is issued, splot will not draw points or lines corresponding to the function or data file points. Contours may still be drawn on the surface, depending on the set contour option. unset surface; set contour base is useful for displaying contours on the grid base. See also set contour (p. 291).

43.61 Table

When table mode is enabled, plot and splot commands print out a multicolumn ASCII table of X Y {Z} R values rather than creating an actual plot on the current terminal. The character R takes on one of three values: "i" if the point is in the active range, "o" if it is out-of-range, or "u" if it is undefined. The data format is determined by the format of the axis labels (see set format (p. 322)), and the columns are separated by single spaces. This can be useful if you want to generate contours and then save them for further use, perhaps for plotting with plot; see set contour (p. 291) for example. The same method can be used to save interpolated data (see set samples (p. 433) and set dgrid3d (p. 312)).


     set table {"outfile"}  
     plot <whatever>  
     unset table

Tabular output is written to the named file, if any, otherwise it is written to the current value of set output. You must explicitly unset table in order to go back to normal plotting on the current terminal.

43.62 Terminal

gnuplot supports many different graphics devices. Use set terminal to tell gnuplot what kind of output to generate. Use set output to redirect that output to a file or device.


     set terminal {<terminal-type> | push | pop}  
     show terminal

If <terminal-type> is omitted, gnuplot will list the available terminal types. <terminal-type> may be abbreviated.

If both set terminal and set output are used together, it is safest to give set terminal first, because some terminals set a flag which is needed in some operating systems.

Several terminals have many additional options. For example, see png (p. 685), or postscript (p. 690). The options used by a previous invocation set term <term> <options> of a given <term> are remembered, thus subsequent set term <term> does not reset them. This helps in printing, for instance, when switching among different terminals — previous options don’t have to be repeated.

The command set term push remembers the current terminal including its settings while set term pop restores it. This is equivalent to save term and load term, but without accessing the filesystem. Therefore they can be used to achieve platform independent restoring of the terminal after printing, for instance. After gnuplot’s startup, the default terminal or that from startup file is pushed automatically. Therefore portable scripts can rely that set term pop restores the default terminal on a given platform unless another terminal has been pushed explicitly.

For a complete list of available terminal types, see terminal (p. 564).

43.63 Termoption

The set termoption command allows you to change the behaviour of the current terminal without requiring a new set terminal command. Only one option can be changed per command, and only a small number of options can be changed this way. Currently the only options accepted are

    set termoption {no}enhanced  
    set termoption font "<fontname>{,<fontsize>}"

43.64 Tics

Control of the major (labelled) tics on all axes at once is possible with the set tics command.

Fine control of the major (labelled) tics on all axes at once is possible with the set tics command. The tics may be turned off with the unset tics command, and may be turned on (the default state) with set tics. Similar commands (by preceding ’tics’ by the axis name) control the major tics on a single axis.


     set tics {axis | border} {{no}mirror}  
              {in | out} {scale {default | <major> {,<minor>}}}  
              {{no}rotate {by <ang>}} {offset <offset> | nooffset}  
              { format "formatstring" } { font "name{,<size>}" }  
              { textcolor <colorspec> }  
     set tics {front | back}  
     unset tics  
     show tics

The options in the first set above can be applied individually to any or all axes, i.e., x, y, z, x2, y2, and cb.

Set tics front or back applies to all axes at once, but only for 2D plots (not splot). It controls whether the tics are placed behind or in front of the plot elements, in the case that there is overlap.

axis or border tells gnuplot to put the tics (both the tics themselves and the accompanying labels) along the axis or the border, respectively. If the axis is very close to the border, the axis option will move the tic labels to outside the border in case the border is printed (see set border (p. 264)). The relevant margin settings will usually be sized badly by the automatic layout algorithm in this case.

mirror tells gnuplot to put unlabelled tics at the same positions on the opposite border. nomirror does what you think it does.

in and out change the tic marks to be drawn inwards or outwards.

With scale, the size of the tic marks can be adjusted. If <minor> is not specified, it is 0.5*<major>. The default size 1.0 for major tics and 0.5 for minor tics is requested by scale default.

rotate asks gnuplot to rotate the text through 90 degrees, which will be done if the terminal driver in use supports text rotation. norotate cancels this. rotate by <ang> asks for rotation by <ang> degrees, supported by some terminal types.

The defaults are border mirror norotate for tics on the x and y axes, and border nomirror norotate for tics on the x2 and y2 axes. For the z axis, the default is nomirror.

The <offset> is specified by either x,y or x,y,z, and may be preceded by first, second, graph, screen, or character to select the coordinate system. <offset> is the offset of the tics texts from their default positions, while the default coordinate system is character. See coordinates (p. 41) for details. nooffset switches off the offset.

set tics with no options restores to place tics inwards. Every other options are retained.

See also set xtics (p. 514) for more control of major (labelled) tic marks and set mxtics for control of minor tic marks. These commands provide control at a axis by axis basis.

43.65 Ticslevel

See set xyplane (p. 527).

43.66 Ticscale

The set ticscale command is deprecated, use set tics scale instead.

43.67 Timestamp

The command set timestamp places the time and date of the plot in the left margin.


     set timestamp {"<format>"} {top|bottom} {{no}rotate}  
                   {offset <xoff>{,<yoff>}} {font "<fontspec>"}  
     unset timestamp  
     show timestamp

The format string allows you to choose the format used to write the date and time. Its default value is what asctime() uses: "%a %b %d %H:%M:%S %Y" (weekday, month name, day of the month, hours, minutes, seconds, four-digit year). With top or bottom you can place the timestamp at the top or bottom of the left margin (default: bottom). rotate lets you write the timestamp vertically, if your terminal supports vertical text. The constants <xoff> and <yoff> are offsets that let you adjust the position more finely. <font> is used to specify the font with which the time is to be written.

The abbreviation time may be used in place of timestamp.


     set timestamp "%d/%m/%y %H:%M" offset 80,-2 font "Helvetica"

See set timefmt (p. 495) for more information about time format strings.

43.68 Timefmt

This command applies to timeseries where data are composed of dates/times. It has no meaning unless the command set xdata time is given also.


     set timefmt "<format string>"  
     show timefmt

The string argument tells gnuplot how to read timedata from the datafile. The valid formats are:

Time Series timedata Format Specifiers


%d day of the month, 1–31
%m month of the year, 1–12
%y year, 0–99
%Y year, 4-digit
%j day of the year, 1–365
%H hour, 0–24
%M minute, 0–60
%s seconds since the Unix epoch (1970-01-01 00:00 UTC)
%S second, 0–60
%b three-character abbreviation of the name of the month
%B name of the month

Any character is allowed in the string, but must match exactly. \t (tab) is recognized. Backslash-octals (\nnn) are converted to char. If there is no separating character between the time/date elements, then %d, %m, %y, %H, %M and %S read two digits each, %Y reads four digits and %j reads three digits. %b requires three characters, and %B requires as many as it needs.

Spaces are treated slightly differently. A space in the string stands for zero or more whitespace characters in the file. That is, "%H %M" can be used to read "1220" and "12 20" as well as "12 20".

Each set of non-blank characters in the timedata counts as one column in the using n:n specification. Thus 11:11 25/12/76 21.0 consists of three columns. To avoid confusion, gnuplot requires that you provide a complete using specification if your file contains timedata.

Since gnuplot cannot read non-numerical text, if the date format includes the day or month in words, the format string must exclude this text. But it can still be printed with the "%a", "%A", "%b", or "%B" specifier: see set format (p. 322) for more details about these and other options for printing timedata. (gnuplot will determine the proper month and weekday from the numerical values.)

See also set xdata (p. 503) and Time/date (p. 106) for more information.


     set timefmt "%d/%m/%Y\t%H:%M"

tells gnuplot to read date and time separated by tab. (But look closely at your data — what began as a tab may have been converted to spaces somewhere along the line; the format string must match what is actually in the file.) See also

time data demo.

43.69 Title

The set title command produces a plot title that is centered at the top of the plot. set title is a special case of set label.


     set title {"<title-text>"} {offset <offset>} {font "<font>{,<size>}"}  
               {{textcolor | tc} {<colorspec> | default}} {{no}enhanced}  
     show title

If <offset> is specified by either x,y or x,y,z the title is moved by the given offset. It may be preceded by first, second, graph, screen, or character to select the coordinate system. See coordinates (p. 41) for details. By default, the character coordinate system is used. For example, "set title offset 0,-1" will change only the y offset of the title, moving the title down by roughly the height of one character. The size of a character depends on both the font and the terminal.

<font> is used to specify the font with which the title is to be written; the units of the font <size> depend upon which terminal is used.

textcolor <colorspec> changes the color of the text. <colorspec> can be a linetype, an rgb color, or a palette mapping. See help for colorspec (p. 68) and palette (p. 400).

noenhanced requests that the title not be processed by the enhanced text mode parser, even if enhanced text mode is currently active.

set title with no parameters clears the title.

See syntax (p. 101) for details about the processing of backslash sequences and the distinction between single- and double-quotes.

43.70 Tmargin

The command set tmargin sets the size of the top margin. Please see set margin (p. 375) for details.

43.71 Trange

The set trange command sets the parametric range used to compute x and y values when in parametric or polar modes. Please see set xrange (p. 508) for details.

43.72 Urange

The set urange and set vrange commands set the parametric ranges used to compute x, y, and z values when in splot parametric mode. Please see set xrange (p. 508)