gnuplot

An Interactive Plotting Program

Thomas Williams & Colin Kelley

Version 3.6 organized by: David Denholm

Major contributors (alphabetic order):

John Campbell

Robert Cunningham

David Denholm

Gershon Elber

Roger Fearick

Carsten Grammes

Thomas Koenig

David Kotz

Ed Kubaitis

Russell Lang

Alexander Lehmann

Carsten Steger

Tom Tkacik

Jos Van der Woude

Alex Woo

Copyright (C) 1986 - 1993, 1997 Thomas Williams, Colin Kelley

Mailing list for comments: info-gnuplot@dartmouth.edu

Mailing list for bug reports: bug-gnuplot@dartmouth.edu

This manual was prepared by Dick Crawford

10 October 1997

gnuplot
Commands
Graphical User Interfaces
Bugs

Copyright
Introduction
Seeking-assistance
What's New in version 3.6
Command-line-editing
Comments
Coordinates
Environment
Expressions
Glossary
Plotting
Start-up
Substitution
Syntax
Time/Date data

      Copyright (C) 1986 - 1993, 1997   Thomas Williams, Colin Kelley

      AUTHORS

      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 is a command-driven interactive function and data plotting program. It 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 for more details).

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.

Many gnuplot commands have multiple options. These options must appear in the proper order, although unwanted ones may be omitted in most cases. Thus if the entire command is "command a b c", then "command a c" will probably work, but "command c a" will fail.

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 comment). 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.

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 the plot command (if on-line, type help plot). Simple Plots Demo

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

      comp.graphics.apps.gnuplot

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. Administrative requests should be sent to

      majordomo@dartmouth.edu

Send a message with the body (not the subject) consisting of the single word "help" (without the quotes) for more details.

The address for mailing to list members is:

      info-gnuplot@dartmouth.edu

Bug reports and code contributions should be mailed to:

      bug-gnuplot@dartmouth.edu

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

      info-gnuplot-beta@dartmouth.edu

There is also a World Wide Web page with up-to-date information, including known bugs:

      http://www.cs.dartmouth.edu/gnuplot
 

Before seeking help, please check the FAQ (Frequently Asked Questions) list. If you do not have a copy of the FAQ, you may request a copy by email from the Majordomo address above, or see the WWW gnuplot page.

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 info-gnuplot, 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.

2. Greatly expanded using command. See plot using.

3. set timefmt allows for the use of dates as input and output for time series plots. See Time/Date data and timedat.dem.

4. Multiline labels and font selection in some drivers.

5. Minor (unlabeled) tics. See set mxtics.

6. key options for moving the key box in the page (and even outside of the plot), putting a title on it and a box around it, and more. See set key.

7. Multiplots on a single logical page with set multiplot.

8. Enhanced postscript driver with super/subscripts and font changes. (This was a separate driver (enhpost) that was available as a patch for 3.5.)

9. Second axes: use the top and right axes independently of the bottom and left, both for plotting and labels. See plot.

10. Special datafile names '-' and "". See plot special-filenames.

11. Additional coordinate systems for labels and arrows. See coordinates.

12. set size can try to plot with a specified aspect ratio.

13. set missing now treats missing data correctly.

14. The call command: load with arguments.

15. More flexible range commands with reverse and writeback keywords.

16. set encoding for multi-lingual encoding.

17. New x11 driver with persistent and multiple windows.

18. New plotting styles: xerrorbars, histeps, financebars and more. See set style.

19. New tic label formats, including "%l %L" which uses the mantissa and exponents to a given base for labels. See set format.

20. New drivers, including cgm for inclusion into MS-Office applications and gif for serving plots to the WEB.

21. Smoothing and spline-fitting options for plot. See plot smooth.

22. set margin and set origin give much better control over where a graph appears on the page.

23. set border now controls each border individually.

24. The new commands if and reread allow command loops.

25. Point styles and sizes, line types and widths can be specified on the plot command. Line types and widths can also be specified for grids, borders, tics and arrows. See plot with. Furthermore these types may be combined and stored for further use. See set linestyle.

26. Text (labels, tic labels, and the time stamp) can be written vertically by those terminals capable of doing so.

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:

      Line-editing:

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

      History:

      ^P    moves back through history.
      ^N    moves 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:

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

The Atari version of readline defines some additional key aliases:

      Undo            - same as ^L.
      Home            - same as ^A.
      Ctrl Home       - same 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.

The commands set arrow, set key, and set label 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 or screen.

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); and 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.

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.

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 and set timefmt. gnuplot will also accept an integer expression, which will be interpreted as seconds from 1 January 2000.

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) and, of course, by later explicit changes.

On Unix, AmigaDOS, 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 AmigaDOS, AtariTOS, MS-DOS and OS/2, gnuplot is used. On VMS, SYS$LOGIN: is used. See help start-up.

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

On Unix, AtariTOS and AmigaDOS, 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 interface is used, BGI is used to point to the full path of the BGI drivers directory. Furthermore, SVGA is used to name the Super VGA BGI driver in 800x600 resolution and its mode of operation is Name.Mode. E.g., if the Super VGA driver is

      C:\TC\BGI\SVGADRV.BGI

and mode 3 is used for 800x600 resolution, then use the following:

      set BGI=C:\TC\BGI
      set SVGA=SVGADRV.3

FIT_SCRIPT may be used to specify a gnuplot command to be executed when a fit is interrupted---see fit. FIT_LOG specifies the filename of the logfile maintained by fit.

Functions
Operators
User-defined

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.

abs
acos
acosh
arg
asin
asinh
atan
atan2
atanh
besj0
besj1
besy0
besy1
ceil
cos
cosh
erf
erfc
exp
floor
gamma
ibeta
inverf
igamma
imag
invnorm
int
lgamma
log
log10
norm
rand
real
sgn
sin
sinh
sqrt
tan
tanh
column
tm_hour
tm_mday
tm_min
tm_mon
tm_sec
tm_wday
tm_yday
tm_year
valid

The arg function returns the phase of a complex number in radians or degrees, as selected by set angles.

The atan function returns the arc tangent (inverse tangent) of its argument. atan returns its argument in radians or degrees, as selected by set angles.

The atan2 function returns the arc tangent (inverse tangent) of the ratio of the real parts of its arguments. atan2 returns its argument in radians or degrees, as selected by set angles, in the correct quadrant.

The cos function returns the cosine of its argument. cos accepts its argument in radians or degrees, as selected by set angles.

The sin function returns the sine of its argument. sin expects its argument to be in radians or degrees, as selected by set angles.

The tan function returns the tangent of its argument. tan expects its argument to be in radians or degrees, as selected by set angles.

column(x) may be used only in expressions as part of using manipulations to fits or datafile plots. See plot datafile using.

valid(x) may be used only in expressions as part of using manipulations to fits or datafile plots. See plot datafile using. Use of functions and complex variables for airfoils

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.

Unary
Binary
Ternary

    Symbol      Example    Explanation
      -           -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.

    Symbol       Example      Explanation
      **          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

    Symbol       Example      Explanation
      ?:          a?b:c     ternary operation

      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 data-file using for an explanation of the using syntax.

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>

Examples:

      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)

Note that the variable pi is already defined. But it is in no way magic; you may redefine it to be whatever you like.

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

See show functions and show variables.

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 always remove the 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 graph may have up to four labelled axes. Various 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. 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. "z" also occurs in commands used with 3-d plotting.

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

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 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.

splot can plot surfaces and contours in addition to lines. See splot datafile for information about the requisite file structure for both of these; see set isosamples for information about defining the grid for a 3-d function. See set contour and set cntrparam for information about contours.

When gnuplot is run, it looks for an initialization file to load. This file is called .gnuplot on Unix and AmigaDOS 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 AmigaDOS, Atari(single)TOS, MS-DOS and OS/2, the environment variable gnuplot should contain the name of this directory). 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.

Command-line substitution can be used anywhere on the gnuplot command line.

Example:

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

The general rules of syntax and punctuation in gnuplot are that keywords and options are order-dependent. 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 braces [], text and file names are enclosed in quotes, and a few miscellaneous things are enclosed in parentheses. Brackets {} 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.)

Braces are used to delimit ranges, whether they are given on 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.

Brackets 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.

Text may be enclosed in single- or double-quotes. Backslash processing of sequences like \n (newline) and \345 (octal character code) is performed for double-quoted strings, but not for single-quoted strings.

The 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.

but

      '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.

At present you should not embed \n inside {} when using the enhanced option of the postscript terminal.

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.

Back-quotes are used to enclose system commands for substitution.

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.

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 command set format defines 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"
      set xrange ["03/21":"03/22"]
      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.

cd
call
clear
exit
fit
help
if
load
pause
plot
print
pwd
quit
replot
reread
reset
save
set-show
shell
splot
test
update

      cd '<directory-name>'

      cd 'subdir'
      cd ".."

      cd "c:\newdata"

      cd 'c:\newdata'

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. $ followed by any character other than a digit 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.

Syntax:

      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).

Example:

If the file 'calltest.gp' contains the line:

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

entering the command:

      call 'calltest.gp' "abcd" 1.2 + "'quoted'" -- "$2"

will display:

      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.

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.

Example:

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

Please see set multiplot, set size, and set origin for details of these commands.

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

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

[xrange] and [yrange] are of the form [{variable=}{<min>}{:<max>}], allowing the range of the fit to be limited temporarily in a manner analogous to plot. <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 modifiers for datafiles (using, every,...) in plot are available here (except smooth)---see plot datafile for full details. The default columns for x and y are 1 and 2. These may be changed by the using x:y mechanism. If using has a third entry (a column or an expression), it will be interpreted as the standard deviation of each y value and will be used to compute the weight; otherwise all data will be weighted equally. If four columns are specified, they are x:y:z:error---note that an error _must_ be specified in order to perform a 3-d fit. If errors are not available, a constant value can be specified, e.g., using ...:(1).

Initial values for the parameters to be fit may be specified in a (load-)file wherein each line is of 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 that is initialized but will not be adjusted. It is not necessary (but sometimes useful for clarity) to specify them at all. The keyword # FIXED has to appear in exactly this form.

The other means of specifying the adjustable parameters is to provide a comma-separated list of variable names after the via keyword. If any of these variables do not yet exist within the current gnuplot session, they are created with an initial value of 1.0, but the fit is more likely to converge if a more appropriate starting value is given. If this form is used, it may prove beneficial to iterate the fit, allowing only one or two variables to be adjusted at a time until a reasonably close fit is obtained, before allowing fit to vary all parameters.

After each iteration step, detailed information is given about the fit's state, both on the screen and on a logfile "fit.log". This file will never be erased but always appended to so that the fit's history isn't lost. After each iteration step, the fit may be interrupted by pressing Ctrl-C (any key _but_ Ctrl-C under MSDOS and Atari Multitasking Systems). Then you have the options of stopping (and accepting the current parameter values), continuing the iteration of the fit, or executing a gnuplot command specified by an environment variable FIT_SCRIPT. A plot or load command may be useful in this context.

Special gnuplot variables:

      FIT_LIMIT

may be specified to change the default epsilon limit (1e-5). When the sum of squared residuals changes between two iteration steps by less than a factor of this number, the fit is considered to have 'converged'.

Once the fit is converged, the final values may be stored in (load-)file suitable for use as an initial-value file, as discussed above. Please see update for details.

      FIT_MAXITER

may be specified to limit the number of iterations performed without convergence by FIT_LIMIT. A value of 0 (or not defining it at all) means that there is no limit.

[FIT_SKIP was available in previous releases of gnufit. Its functionality is now obtained using the every modifier for datafiles. FIT_INDEX was previously available in order to allow multi-branch fitting. Multi-branch fitting in 2-d can now be done as a pseudo-3-d fit in which the y values are the dataline number (using 1:-1:...) or index (using 1:-2:...).]

Environment variables:

      FIT_LOG

changes the logfile's path from './fit.log' (write permission is necessary).

      FIT_SCRIPT

specifies a command to be executed after an user interrupt.

Examples:

      f(x) = a*x**2 + b*x + c
      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 f(x,y) 'surface.dat' using 1:2:3:(1) via a, b, c
  See the `fit` demos. 

Introduction To Fitting
Errors In Fitting
Tips and Tricks

fit is used to find a set of parameters to be used in a parametric function to make it fit to your data optimally. The quantity to be minimized is the sum of squared differences between your input data points and the function values at the same places, usually called 'chisquared' (i.e. the Greek letter chi, to the power of 2). (To be precise, the differences will be divided by the input data errors before being squared; see fit errors for details.)

Now that you know why it's called 'least squares fitting', let's see why it's 'nonlinear'. That's because the function's dependence on the parameters (not the data!) may be non-linear. Of course, this might not tell you much if you didn't know already, so let me try to describe it. If the fitting problem were to be linear, the target function would have to be a sum of simple, non-parametric functions, each multiplied by one parameter. (For example, consider the function f(x) = c*sin(x), where we want to find the best value for the constant c. This is nonlinear in x, of course, but it is linear in c. Since the fitting procedure solves for c, it has a linear equation to solve.) For such a linear case, the task of fitting can be performed by comparatively simple linear algebra in one direct step. But fit can do more for you: the parameters may be used in your function in any way you can imagine. To handle this more general case, however, it has to perform an iteration, i.e. it will repeat a sequence of steps until it finds the fit to have 'converged', or until you stop it.

Generally, the function to be fitted will come from some kind of theory (some prefer the term 'model' here) that makes a prediction about how the data should behave, and fit is then used to find the free parameters of the theory. This is a typical task in scientific work, where you have lots of data that depend in more or less complicated ways on the values you're interested in. The results will then usually be of the form 'the measured data can be described by the {foo} theory, for the following set of parameters', and then a set of values is given, together with the errors of your determination of these values.

This reasoning implies that fit is probably _not_ your tool of choice if all you really want is a smooth line through your data points. If you want this, the smooth option to plot is what you've been looking for, not fit. See plot datafile smooth for details.

One of the most important things to keep in mind when using fit is the handling of errors. Here, this term refers to the measurement errors accompanying both your input data and resulting parameters.

The reason for the importance of input data errors to fitting is that normally the single measurements aren't all of the same quality, so they shouldn't have the same importance in determining the results. That's one major reason for dividing the differences between data and function by the input errors, also known as 'weighting', in the computation of chisquared.

By weighting, deviations from your function at places where the data have large errors will have a smaller part in chisquared, as the division will make them smaller compared to the better measurements. Another reason for the division is that, for mathematical reasons, chisquared has to be a dimensionless variable, i.e. chisquared should be something like '15.3', not '15.3 square seconds'.

Without input data errors being given, all data will be weighted equally, and the resulting errors of the parameters won't have much of a real meaning. Therefore, you should always try to find a sensible set of y-errors for your data. An important example is that of data representing a histogram. In such a case, the square root of the y value is often the correct input error to use.

Once the fit iteration has stopped, it will display a load of valuable information which you will have to learn to interpret before you can use it. The 'sum of squares residuals' is the distance between the data and your fit function, shortly called 'chisquared'. This is what fit tries to minimize. To quickly test if your fit went well, check that this is about the same as the number of data points minus the number of parameters (all this is only valid if you supplied y-errors, and the number of data points is large enough). For details on this, look up the 'Chi-squared distribution' in your favourite statistics textbook.

If chisquared is much larger than that, then your function didn't fit the data very well. Try another, more general one, or allow more of the parameters to be adjusted by fit. Another possible reason could be that the y-errors you supplied were a bit optimistic, i.e. too small.

If, on the other hand, chisquared is too small, then the function fit the data _too_ well. Either the given y-errors were too large, or the function is too general. You should try to restrict it by either fixing some parameters, or just make it simpler one way or the other.

If all else went well, you'll see a list of the resulting parameter values, together with estimates of the errors of these values. And you should always be aware of this: they're _estimates_, not more. You'll have to get used to both fit and kind of problems you usually apply it to before you can use these errors for anything serious. To start with, the errors reported by fit are insensitive to the global scale of the y-errors, i.e. if you multiply all y-errors by a constant, the resulting parameter errors don't change.

And, to repeat this once more: if you didn't supply y-errors, the parameter errors will normally be meaningless.

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 batch operation (possibly unattended), where you just supply the startup values in a file and perhaps later use update to copy the results back into another file (or the same one).

The via var1, var2, ... form is best used interactively. Using the command history mechanism built into gnuplot, you can easily edit the list of parameters to be fitted or supply new startup values for the next try. This is particularly useful for hard problems, where a direct fit to all the parameters at once won't work, at least not without really _good_ values to start with. To find such a set of good starting parameters, you can iterate several times, fitting only some of the parameters each time, until the values are close enough to the goal that the final fit (to all the parameters at once) will work.

A general word about starting values: fit may, and often will, get really badly lost in searching for the optimal parameter set if you start it way off any possible solution. The main reason for this is that nonlinear fitting is not guaranteed to converge to a global optimum. It can get stuck in a local optimum, and there's no way for the routine to find out about that. You'll have to use your own judgement in checking whether this has happened to you or not.

To partly avoid that problem, you should put all starting values at least roughly into the vicinity of the solution. At least the order of magnitude should be correct, if possible. The better your starting values are, the less error-prone the fit. A good way to find starting values is to draw data and fit-function into one plot, and iterate, changing the values and replot-ting until reasonable similarity is reached. The same plot is also useful to check if the fit got stuck in a non-global minimum.

Make sure that there is no mutual dependency among parameters of the function you are fitting. E.g., 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 quotient of the largest and the smallest absolute parameter values, the slower the fit will converge. If the quotient 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'll 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 then not nonlinear any more. It should take only a really small number of iterations to converge on a linear case, maybe even only one.

In prescriptions for analysing data from practical experimentation courses, you'll often find descriptions how to first fit your data to some functions, maybe in a multi-step process accounting for several aspects of the underlying theory one by one, and then extract the data you really wanted from the fitting parameters of that function. With fit, this last step can often be eliminated by rewriting the model function to directly use the desired final parameters. Transforming data can also be avoided quite often, although sometimes at the cost of a harder fit problem. If you think this contradicts the previous paragraph about keeping the fit function as simple as possible, you're correct.

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

      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.

      if (<condition>) <command-line>

      pi=3
      if (pi!=acos(-1)) print "?Fixing pi!"; pi=acos(-1); print pi

      ?Fixing pi!
      3.14159265358979

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

See reread for an example of how if and reread can be used together to perform a loop.

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 comment for information about comments in commands.

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

Syntax:

      load "<input-file>"

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

Examples:

      load 'work.gnu'
      load "func.dat"

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. See also call.

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.

Syntax:

      pause <time> {"<string>"}

<time> may be any integer constant or expression. Choosing -1 will wait until a carriage return is hit, zero (0) won't pause at all, and a positive integer will wait the specified number of seconds. pause 0 is synonymous with print.

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).

Examples:

      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."

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 for differences.

Syntax:

      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).

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).

Examples:

      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

data-file
errorbars
parametric
ranges
title
with

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

Syntax:

      {s}plot '<file_name>' {index <index list>}
                            {every <every list>}
                            {thru <thru expression>}
                            {using <using list>}
                            {smooth <option>}

The modifiers index, every, thru, using, and smooth are discussed separately. In brief, 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.

Data files should contain one data point per record. Records beginning with # (or ! on VMS) will be treated as comments and ignored. Each data point represents an (x,y) pair. For plots with error bars (see set style errorbars), 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 on 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.

Data may be written in exponential format with the exponent preceded by the letter e, E, d, D, q, or Q.

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---pairs of blank records separate indexes (see plot datafile index). Data separated by double blank records are treated as if they were in separate data files.

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).

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 '-', '-'
      1 1
      19 19
      e
      1 1
      19 19
      e

every
example datafile
index
smooth
special-filenames
thru
using

      plot 'file' every {<point_incr>}
                          {:{<line_incr>}
                            {:{<start_point>}
                              {:{<start_line>}
                                {:{<end_point>}
                                  {:<end_line>}}}}}

      every :::3::3    # selects just the fourth dataline ('0' is first)
      every :::::9     # selects the first 10 datalines
      every 2:2        # selects every other point in every other dataline
      every ::5::15    # selects points 5 through 15 in each dataline
 Simple Plot Demos ,
 Non-parametric splot demos , and
 Parametric splot demos.

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

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

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

      plot 'file' index 4:5
  splot with indices demo. 

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.

Syntax:

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

unique plots 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) 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 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, four for acsplines, and three for the others.

The smooth options have no effect on function plots.

acsplines
bezier
csplines
sbezier
unique

acsplines approximates the data with a "natural smoothing spline". After the data are made monotonic in x (see smooth unique), 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.

Example:

      sw(x,S)=1/(x*x*S)
      plot 'data_file' using 1:2:(sw($3,100)) smooth acsplines

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

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

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. See demos.

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
      2
      4
      6

      10
      12
      14
      e
      2
      4
      6

      10
      12
      14
      e

does indeed work,

      plot '-', '-'
      2
      4
      6
      e
      10
      12
      14
      e

is a lot easier to type.

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

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.

      plot 'file' thru f(x)

      plot 'file' using 1:(f($2))

      plot 'file' thru f(y)

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

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

N.B.---the call command also uses $'s as a special character. See call for details about how to include a column number in a call argument list.

If the using list has but a single entry, that <entry> will be used for y and the data point number 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. Additional entries are usually errors in x and/or y. See set style for details about plotting styles that make use of error information, and fit for use of error information in curve fitting.

'scanf' accepts several numerical specifications but gnuplot requires all inputs to be double-precision floating-point variables, so lf is the only permissible specifier. '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" or requires use of double-quotes rather than single-quotes.

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.)

      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.

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

If you want to leave text in your data files, it is always safe to put the comment character (#) in the first column of the text lines. Feeble using demos.

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).

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

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

See plot using, plot with, and set style for more information.

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

Examples:

      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
  Parametric Mode Demos. 

      [{<dummy-var>=}{{<min>}:{<max>}}]
      [{{<min>}:{<max>}}]

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.

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.

Examples:

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'

See Demo.

      title "<title>" | notitle

By default the line title is the function or 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 for details.

Examples:

This plots y=x with the title 'x':

      plot x

This plots x squared with title "x^2" and "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

      with <style> { {linestyle | ls <line_style>}
                     | {{linetype | lt <line_type>}
                        {linewidth | lw <line_width>}
                        {pointtype | pt <point_type>}
                        {pointsize | ps <point_size>}} }

where <style> is either lines, points, linespoints, impulses, dots, steps, fsteps, histeps, errorbars, xerrorbars, yerrorbars, xyerrorbars, boxes, boxerrorbars, boxxyerrorbars, financebars, candlesticks or vector. Some of these styles require additional information. See set style <style> for details of each style.

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

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

The keywords may be abbreviated as indicated.

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

Examples:

This plots sin(x) with impulses:

      plot sin(x) with impulses

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

      plot x*y w points, x**2 + y**2

This plots tan(x) with the default function style, "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 ('exper.dat' should have three or four data columns):

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

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 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

See set style to change the default styles. Styles demos.

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.

Syntax:

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

See expressions.

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.

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, and similarly you can use replot to add a plot from a binary file only if the previous command 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 for ways to edit the last plot (splot) command.

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 for details.) The reread command has no effect if input from standard input.

Examples:

Suppose the file "looper" contains the commands

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

and from within gnuplot you submit the commands

      a=0
      load 'looper'

The result will be four 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

      n_p=6
      c_p=1
      set nokey
      set yrange [0:10]
      set multiplot
      call 'plotter' 'data'
      set nomultiplot

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. Reread Animation Demo

The reset command causes all options that can be set with the set command to take on their default values. The only exceptions are that the terminal set with set term and the output file set with set output are left unchanged. This command is useful, e.g., to restore the default 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 save command saves user-defined functions, variables, set options, or all three, plus the last plot (splot) command to the specified file.

Syntax:

      save  {<option>} "<filename>"

where <option> is functions, variables 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.

The filename must be enclosed in quotes.

Examples:

      save "work.gnu"
      save functions 'func.dat'
      save var 'var.dat'
      save set "options.dat"

The set command sets _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.

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.

angles
arrow
autoscale
bar
bmargin
border
boxwidth
clabel
clip
cntrparam
contour
data style
dgrid3d
dummy
encoding
format
function style
functions
grid
hidden3d
isosamples
key
label
linestyle
lmargin
locale
logscale
mapping
margin
missing
multiplot
mx2tics
mxtics
my2tics
mytics
mztics
offsets
origin
output
parametric
pointsize
polar
rmargin
rrange
samples
size
style
surface
terminal
tics
ticslevel
ticscale
timestamp
timefmt
title
tmargin
trange
urange
variables
view
vrange
x2data
x2dtics
x2label
x2mtics
x2range
x2tics
x2zeroaxis
xdata
xdtics
xlabel
xmtics
xrange
xtics
xzeroaxis
y2data
y2dtics
y2label
y2mtics
y2range
y2tics
y2zeroaxis
ydata
ydtics
ylabel
ymtics
yrange
ytics
yzeroaxis
zdata
zdtics
zero
zeroaxis
zlabel
zmtics
zrange
ztics

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.

Syntax:

      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. Note that the output of inverse hyperbolic functions of complex arguments are effected, however; if these functions are used, set angles radians must be in effect:

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

but

      set angles degrees
      y=sinh(x)
      print y         #prints {1.16933, 0.154051}
      print asinh(y)  #prints {57.29578, 5.729578}
  Polar plot using `set angles`. 

      set arrow {<tag>} {from <position>} {to <position>} {{no}head}
                { {linestyle | ls <line_style>}
                  | {linetype | lt <line_type>}
                    {linewidth | lw <line_width} }
      set noarrow {<tag>}
      show arrow

The <position>s are specified by either x,y or x,y,z, and may be preceded by first, second, graph, or screen to select the coordinate system. Unspecified coordinates default to 0. The endpoints can be specified in one of four coordinate systems---first or second axes, graph or screen. See coordinates 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.

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, arrows have heads.

The line style may be selected from a user-defined list of line styles (see set linestyle) 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.

Examples:

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 delete arrow number 2, use:

      set noarrow 2

To delete all arrows, use:

      set noarrow

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

      show arrow
  Arrows Demos. 

      set autoscale {<axes>{min|max}}
      set noautoscale {<axes>{min|max}}
      show autoscale

where <axes> is either x, y, z, 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.

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 for additional information about ranges.

The behavior of autoscaling remains consistent in parametric mode, (see set parametric). 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).

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.

Examples:

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 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:

      set noautoscale

This disables autoscaling of the z axis only:

      set noautoscale z

parametric mode
polar 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.

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 polar demos

      set bar {small | large | <size>}
      show bar

The command set bmargin sets the size of the bottom margin. Please see set margin for details.

The set border and set noborder commands control the display of the graph borders for the plot and splot commands.

Syntax:

      set border {<integer>}
      set noborder
      show border

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 and the top four bits control the edges on top of the splot. In detail, the <integer> should be the sum of the appropriate entries from the following table:

                        plot border
                            and
                        splot base    splot verticals    splot top
      bottom (south)         1              16              256
      left   (west)          2              32              512
      top    (north)         4              64             1024
      right  (east)          8             128             2048

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

To have tics on edges other than bottom and left, disable the usual tics and enable the second axes.

Examples:

Draw all borders:

      set border

Draw only the SOUTHWEST borders:

      set border 3

Draw a complete box around a splot:

      set border 4095

Draw a partial box, omitting the front vertical:

      set border 127+256+512

Draw only the NORTHEAST borders:

      set noxtics; set noytics; set x2tics; set y2tics; set border 12

Borders Demo.

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

Syntax:

      set boxwidth {<width>}
      show boxwidth

If a data file is plotted without the width being specified in the third, fourth, or fifth column (or using entry), or if a function is plotted, the width of each box is set by the set boxwidth command. (If a width is given both in the file and by the set boxwidth command, the one in the file is used.) If the width is not specified in one of these ways, the width of each box will be calculated automatically so that it touches the adjacent boxes. 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 set style boxerrorbars 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)

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.

Syntax:

      set clabel {'<format>'}
      set noclabel
      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.

See also set contour.

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

Syntax:

      set clip <clip-type>
      set noclip <clip-type>
      show clip

Three clip types are supported by gnuplot: points, one, and two. One, two, or all three clip types may be active for a single graph.

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
      set noclip

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

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

points---Eventually all drawings are done with piecewise linear strokes. This number controls the number of points used to approximate a curve. It is relevant only for cubicspline and bspline modes.

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---Approximate number of contour levels. Selection of the levels is controlled by auto (default), discrete, and incremental. For auto, 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 discrete, contours will be generated at z = <z1>, <z2> ... as specified. The number of discrete levels is limited to MAX_DISCRETE_LEVELS, defined in plot.h to be 30. If incremental, contours are generated at values of z beginning at <start> and increasing by <increment> until <end> is reached. If <end> is not specified, MAX_DISCRETE_LEVELS will be generated.

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

Examples:

      set cntrparam bspline
      set cntrparam points 7
      set cntrparam order 10

To select 5 levels automatically:

      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 (retaining the current settings of auto, discr. and increment's start and increment value, while changing its end):

      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 for control of where the contours are drawn, and set clabel for control of the format of the contour labels. Contours Demo and contours with User Defined Levels.

set contour enables contour drawing for surfaces. This option is available for splot only.

Syntax:

      set contour {base | surface | both}
      set nocontour
      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 for the parameters that affect the drawing of contours, and set clabel for control of labelling of the contours.

The surface can be switched off (see set surface), giving a contour-only graph. Though it is possible to use set view to enlarge the plot to fill the screen, better results can be obtained by writing the contour information out to a file, and rereading it as a 2-d datafile plot:

      set nosurface
      set contour
      set cntrparam ...
      set term table
      set out 'filename'
      splot ...
      set out
      # contour info now in filename
      set term <whatever>
      plot 'filename'

In order to draw contours, the data must be organized as "grid data". In such a file all of the points for a single y value are listed, then all the points for the next y, 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 value group from the next. See also plot datafile.

If contours are desired from non-grid data, set dgrid3d can be used to create an appropriate grid. See set dgrid3d for more information. Contours Demo and contours with User Defined Levels.

      set data style <style-choice>
      show data style

See set style for the choices. If no choice is given, the choices are listed. show data style shows the current default data plotting style.

      set dgrid3d {<row_size>} {,{<col_size>} {,<norm>}}
      set nodgrid3d
      show dgrid3d

The third parameter, norm, controls the weighting: each point is weighted inversely by its distance (from the grid point) raised to the norm power. (Actually it's not quite the distance: 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.) Thus the closer the data point is to a grid point, the more effect it has on that grid point. In gnuplot, this distance computation is optimized for norms that are powers of 2, specifically 1, 2, 4, 8, and 16, but any non-negative integer can be used.

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.

Examples:

      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 the L2 norm (a norm of 2 is to be used in the distance computation). The second only modifies the norm to be used to L4. Dgrid3d Demo.

      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.

Examples:

      set dummy u,v
      set dummy ,s

The second example sets the second variable to s.

      set encoding <value>
      show encoding

      set format {<axes>} {"<format-string>"}
      set format {<axes>} {'<format-string>'}
      show format

where <axes> is either x, y, z, xy, x2, y2 or nothing (which is the same as xy). The length of the string representing a ticmark (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 set noxtics or set noytics.

Newline (\n) is accepted in the format string. Use double-quotes rather than single-quotes to enable such interpretation. See also syntax.

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 %%".

The acceptable formats (if not in date/time mode) are:

      Format       Explanation
      %f           floating point notation
      %e or %E     exponential notation; an "e" or "E" before the power
      %g or %G     the shorter of %e (or %E) and %f
      %x or %X     hex
      %o or %O     octal
      %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; "#", 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 releases of 'printf' may not support all of these modifiers but may also support others; in case of doubt, check the appropriate documentation and then experiment.

Examples:

      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 "%.0P%%"; 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 date/time, the format string must contain valid codes for the 'strftime' function (outside of gnuplot, type "man strftime"). See set timefmt for a list of the allowed input format codes.

In date/time mode, the acceptable formats are:

      Format       Explanation
      %a           abbreviated name of day of the week
      %A           full name of day of the week
      %b or %h     abbreviated name of the month
      %B           full name of the month
      %d           day of the month, 1--31
      %D           shorthand for "%m/%d/%y"
      %H or %k     hour, 0--24
      %I or %l     hour, 0--12
      %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"
      %R           shorthand for %H:%M"
      %S           second, 0--60
      %T           shorthand for "%H:%M:%S"
      %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.

Examples:

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"

See also set xtics for more information about tic labels. See demo.

      set function style <style-choice>
      show function style

See set style for the choices. If no choice is given, the choices are listed. show function style shows the current default function plotting style.

      show functions

For information about the definition and usage of functions in gnuplot, please see expressions and user-defined. Splines as User Defined Functions. Use of functions and complex variables for airfoils

      set grid {{no}{m}xtics} {{no}{m}ytics} {{no}{m}ztics}
               {{no}{m}x2tics} {{no}{m}y2tics}
               {polar {<angle>}} {<major_linetype> {<minor_linetype>}}
      set nogrid
      show grid

The grid can be enabled and disabled for the major and/or minor tic marks on any axis, and the linetype can be specified for major and minor grid lines. But note that <major_linetype> and <minor_linetype> are indices in the default linetype list provided by the terminal; user-defined linetypes (via the set linestyle command) are not accessible for grid lines.

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.

Z grid lines are drawn on the back of the plot. This looks better if a partial box is drawn around the plot---see set border.

The set hidden3d command enables hidden line removal for explicit surface plotting (see splot).

Syntax:

      set hidden3d
      set nohidden3d
      show hidden3d

Hidden line removal may be used for both explicit functions and for explicit data. It now works for parametric surfaces as well. This mode is meaningful only when surfaces are splotted with lines.

When this flag is set, both the hidden portion of the surface and possibly its hidden contours (see set contour) as well as the hidden grid will be removed. Each surface has its hidden parts removed with respect to itself and to other surfaces, if more than one surface is plotted. But contours drawn on the surface (set contour surface) don't seem to work. Labels and arrows are always visible and are unaffected. Hidden Line Removal Demo and Complex Hidden Line Demo.

      set isosamples <iso_1> {,<iso_2>}
      show isosamples

When a surface plot is being done without the removal of hidden lines, set samples also has an effect on the number of points being evaluated---it controls the number of points sampled along each isoline. See set samples.

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 and plot with for more information.

Syntax:

      set key {  left | right | top | bottom | outside | below
               | <position>}
              {Left | Right} {{no}reverse}
              {samplen <sample_length>} {spacing <vertical_spacing>}
              {width <width_increment>}
              {title "<text>"} {{no}box {<linetype>}}
      set nokey
      show key

By default the key is placed in the upper right corner of the graph. The keywords left, right, top, bottom, outside and below may be used to place the key in the other corners inside the graph or to the right (outside) or below the graph. They may be given alone or combined.

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 {<linetype>}) in a specified linetype. But note that <linetype> is an index in the default linetype list provided by the terminal; user-defined linetypes (via the set linestyle command) are not accessible for the key box.

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 it is not drawn. <sample_length> must be an integer.

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.

A title can be put on the key (title "<text>")---see also syntax for the distinction between text in single- or double-quotes. The key title uses the same justification as do the plot titles.

The defaults for set key are right, top, Right, noreverse, samplen 4, spacing 1.25, title "", and nobox. The default <linetype> is the same as that used for the plot borders. Entering set key with no options returns the key to its default configuration.

The <position> can be a simple x,y,z as in previous versions, but these can be preceded by one of four keywords (first, second, graph, screen) which selects the coordinate system in which the position is specified. See coordinates for more details.

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.

Some or all of the key may be outside of the graph boundary, although this may interfere with other labels and may cause an error on some devices. If you use the keywords outside or below, gnuplot makes space for the keys and the graph becomes smaller. Putting keys outside to the right, they occupy as few columns as possible, and putting them below, as many columns as possible (depending of the length of the labels), thus stealing as little space from the graph as possible.

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 for details.

Examples:

This places the key at the default location:

      set key

This disables the key:

      set nokey

This places a key at coordinates 2,3.5,2 in the default (first) coordinate system:

      set key 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

      set label {<tag>} {"<label_text>"} {at <position>}
                {<justification>} {{no}rotate} {font "<name><,size>"}
      set nolabel {<tag>}
      show label

The <position> is specified by either x,y or x,y,z, and may be preceded by first, second, graph, or screen to select the coordinate system. See coordinates 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.

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 parameter <justification>, 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 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 and set timefmt.

The EEPIC, Imagen, LaTeX, and TPIC drivers allow \\ in a string to specify a newline.

Examples:

To set a label at (1,2) to "y=x", use:

      set label "y=x" at 1,2

To set a label of the sign Sigma of size 24 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:

      set nolabel 2

To delete all labels, use:

      set nolabel

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

Each terminal has a default set of line and point types, which can be seen by using the command test. set linestyle 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.

Syntax:

      set linestyle <index> {linetype | lt <line_type>}
                            {linewidth | lw <line_width>}
                            {pointtype | pt <point_type>}
                            {pointsize | ps <point_size>}
      set nolinestyle
      show linestyle

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 styles; both may be used.

Not all terminals support the line width and point size features, so these terminals obviously cannot fully support set linestyle.

Note that this feature is not completely installed; linestyles defined by this mechanism may be used with 'plot', 'splot', 'replot', and 'set arrow', but not by other commands that allow the default index to be used, such as 'set grid'.

Example: 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 linestyle 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 function style 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 function style linespoints
      plot p(x) lt 1 pt 3, q(x) ls 1

will create a plot of f(x) using the default triangles connected by a red line and q(x) using small triangles connected by a green line.

The command set lmargin sets the size of the left margin. Please see set margin for details.

      set locale {"<locale>"}

      set logscale <axes> <base>
      set nologscale <axes>
      show logscale

      set logscale xz

      set logscale y 2

      set nologscale z

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.

Syntax:

      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 polar and azimuthal angles theta and phi (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.

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 plots. Mapping Demos.

      set bmargin {<margin>}
      set lmargin {<margin>}
      set rmargin {<margin>}
      set tmargin {<margin>}
      show margin

The 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.

The set missing command allows you to tell gnuplot what character is used in a data file to denote missing data.

Syntax:

      set missing {"<character>"}
      show missing

Example:

      set missing "?"

would mean that, when plotting a file containing

         1 1
         2 ?
         3 2

the middle line would be ignored.

There is no default character for missing.

The command set multiplot places gnuplot in the multiplot mode, in which several plots are placed on the same page, window, or screen.

Syntax:

      set multiplot
      set nomultiplot

For some terminals, no plot is displayed until the command set nomultiplot 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 a plot, but the screen may not be cleared between 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/set notime pair around one of the plot, splot or replot commands within the set multiplot/set nomultiplot block.

The commands set origin and set size must be used to correctly position each plot; see set origin and set size for details of their usage.

Example:

      set size 0.7,0.7
      set origin 0.1,0.1
      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)
      set nomultiplot

displays a plot of cos(x) stacked above a plot of sin(x). Note the initial set size and set origin. While these are not always required, their inclusion is recommended. Some terminal drivers require that bounding box information be available before any plots can be made, and the form given above guarantees that the bounding box will include the entire plot array rather than just the bounding box of the first plot.

set size and set origin refer to the entire plotting area used for each plot. If you want to have the axes themselves line up, perhaps to avoid having to label all of them, you need to guarantee that the margins are the same size. This can be done with the set margin commands. Please see set margin for their use. See demo.

Minor tic marks along the x2 (top) axis are controlled by set mx2tics. Please see set mxtics.

      set mxtics {<freq> | default}
      set nomxtics
      show mxtics

The same syntax applies to mytics, mztics, mx2tics and my2tics.

<freq> is the number of sub-intervals (NOT the number of minor tics) between major tics (ten is the default for a linear axis, so there are nine 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.

Minor tics can be used only with uniformly spaced major tics. Since major tics can be placed arbitrarily by set {x|x2|y|y2|z}tics, minor tics cannot be used if major tics are explicitly set.

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 for information about these.

Minor tic marks along the y2 (right-hand) axis are controlled by set my2tics. Please see set mxtics.

Minor tic marks along the y axis are controlled by set mytics. Please see set mxtics.

Minor tic marks along the z axis are controlled by set mztics. Please see set mxtics.

      set offsets <left>, <right>, <top>, <bottom>
      set nooffsets
      show offsets

Offsets are ignored in splots.

Example:

      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.

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 for information about this system).

Syntax:

      set origin <x-origin>,<y-origin>

      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, 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 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.

The set parametric command changes the meaning of plot (splot) from normal functions to parametric functions. The command set noparametric restores the plotting style to normal, single-valued expression plotting.

Syntax:

      set parametric
      set noparametric
      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). 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.

      set pointsize <pointsize>
      show pointsize

The pointsize of a single plot may be changed on the plot command. See plot with for details.

Please note that the pointsize setting is not supported with all terminal types.

      set polar
      set nopolar
      show polar

There have been changes made to polar mode in version 3.6, 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).

The command set nopolar 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 command for similar functionality for splots.

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 for details of all the set range commands.

Example:

      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. Polar demos Polar Data Plot.

The command set rmargin sets the size of the right margin. Please see set margin for details.

The set rrange command sets the range of the radial coordinate for a graph in polar mode. Please see set xrange for details.

      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 smooth options is used.

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 isoline. 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.

      set size {{no}square | ratio <r> | noratio} {<xscale>,<yscale>}
      show size

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 its default aspect ratio (1.0), but do not return <xscale> or <yscale> to their default values (also 1.0).

ratio and square have no effect on 3-d plots.

set size is relative to the default size, which differs from terminal to terminal. Since gnuplot fills as much of the available plotting area as possible by default, it is safer to use set size to decrease the size of a plot than to increase it. See set terminal for the default sizes.

On some terminals, changing the size of the plot will result in text being misplaced.

Examples:

To set the size to normal size use:

      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 demo.

Default styles are chosen with the set function style and set data style commands. See plot with for information about how to override the default plotting style for individual functions and data sets.

Syntax:

      set function style <style>
      set data style <style>
      show function style
      show data style

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) can be used with splots or function plots. Neither boxes nor any of the steps styles can be used with splots. If an inappropriate style is specified, it will be changed to points.

For 2-d data with more than two columns, gnuplot is picky about the allowed errorbar 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), 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, boxxyerrorbars, and boxerrorbars are allowed. An illegal style will be changed to yerrorbars.

Five-column data allow only the boxerrorbars, financebars, and candlesticks styles. (The last two of these are primarily used for plots of financial prices.) An illegal style will be changed to boxerrorbars before plotting.

Six- and seven-column data only allow the xyerrorbars and boxxyerrorbars styles. Illegal styles will be changed to xyerrorbars before plotting.

For more information about error bars, please see plot errorbars.

boxerrorbars
boxes
boxxyerrorbars
candlesticks
dots
financebars
fsteps
histeps
impulses
lines
linespoints
points
steps
vector
xerrorbars
xyerrorbars
yerrorbars

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 Demo.

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.

The candlesticks 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 an open rectangle, 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 width of the rectangle may be changed by set bar. The symbol will be unchanged if the low and high prices are interchanged or if the opening and closing prices are interchanged. See set bar and financebars. See demos.

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 bar. The symbol will be unchanged if the high and low prices are interchanged. See set bar and candlesticks. See demos.

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 demo.

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 demo.

histeps is only a plotting style; gnuplot does not have the ability to create bins and determine their population from some data set.

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.

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 for its usage.

linespoints may be abbreviated lp.

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 for its usage.

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 demo.

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 bar is used---see set bar for details).

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 bar is used---see set bar 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

The yerrorbars (or errorbars) style is only relevant to 2-d data plots. yerrorbars is like dots, 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 bar is used---see set bar for details). See demo.

      set surface
      set nosurface
      show surface

gnuplot supports many different graphics devices. Use set terminal to tell gnuplot what kind of output to generate.

Syntax:

      set terminal {<terminal-type>}
      show terminal

If <terminal-type> is omitted, gnuplot will list the available terminal types. <terminal-type> may be abbreviated.

Use set output to redirect this output to a file or device.

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 additional options. For example, see dumb, iris4d, hpljii or postscript.

This document may describe drivers that are not available to you because they were not installed, or it may not describe all the drivers that are available to you, depending on its output format.

aed767
gpic
regis
tek410x
tek40
x11
xlib
aifm
cgm
corel
dumb
dxf
dxy800a
excl
fig
hp2623a
hp2648
hp500c
hpgl
hpljii
hppj
imagen
kyo
mif
pbm
png
postscript
qms
table
tgif
tkcanvas
vx384
epson-180dpi
latex
pslatex and pstex
eepic
tpic
pstricks
texdraw
mf

      set terminal gpic {<x> <y>}

      groff -p -mpic -Tps file.pic > file.ps.

The output from pic can be pipe-lined into eqn, so it is possible to put complex functions in a graph with the set label and set {x/y}label commands. For instance,

      set ylab '@space 0 int from 0 to x alpha ( t ) roman d t@'

will label the y axis with a nice integral if formatted with the command:

      gpic filename.pic | geqn -d@@ -Tps | groff -m[macro-package] -Tps
          > filename.ps

Figures made this way can be scaled to fit into a document. The pic language is easy to understand, so the graphs can be edited by hand if need be. All co-ordinates in the pic-file produced by gnuplot are given as x+gnuplotx and y+gnuploty. By default x and y are given the value 0. If this line is removed with an editor in a number of files, one can put several graphs in one figure like this (default size is 5.0x3.0 inches):

      .PS 8.0
      x=0;y=3
      copy "figa.pic"
      x=5;y=3
      copy "figb.pic"
      x=0;y=0
      copy "figc.pic"
      x=5;y=0
      copy "figd.pic"
      .PE

This will produce an 8-inch-wide figure with four graphs in two rows on top of each other.

One can also achieve the same thing by the command

      set terminal gpic x y

for example, using

      .PS 6.0
      copy "trig.pic"
      .PE

      set terminal regis {4 | 16}

gnuplot provides the x11 terminal type for use with X servers. This terminal type is set automatically at startup if the DISPLAY environment variable is set, if the TERM environment variable is set to xterm, or if the -display command line option is used.

Syntax:

      set terminal x11 {reset} {<n>}

Multiple plot windows are supported: set terminal x11 <n> directs the output to plot window number n. If n>0, the terminal number will be appended to the window title and the icon will be labeled gplt <n>. The active window may distinguished by a change in cursor (from default to crosshair.)

Plot windows remain open even when the gnuplot driver is changed to a different device. A plot window can be closed by pressing the letter q while that window has input focus, or by choosing close from a window manager menu.

All plot windows can be closed by specifying reset, which actually terminates the subprocess which maintains the windows (unless -persist was specified).

For terminal type x11, gnuplot accepts (when initialized) the standard X Toolkit options and resources such as geometry, font, and background from from the command line arguments or a configuration file. See the X(1) man page (or its equivalent) for a description of the options.

Many other options are available for the x11 terminal. These may be specified either as command-line options when gnuplot is invoked or as resources in the configuration file "/.Xdefaults". Thus they are chosen at start-up time and cannot be altered during a single gnuplot session.

command-line-options

In addition to the X Toolkit options, the following options may be specified on the command line when starting gnuplot or as resources in your ".Xdefaults" file:

 -mono    forces monochrome rendering on color displays.
 -gray    requests grayscale rendering on grayscale or color displays.
            (Grayscale displays receive monochrome rendering by default.)
 -clear   requests that the window be cleared momentarily before a
            new plot is displayed.
 -tvtwm   requests that geometry specifications for position of the
            window be made relative to the currently displayed portion
            of the virtual root.
 -raise   raise plot window after each plot
 -noraise do not raise plot window after each plot
 -persist plot windows survive after main gnuplot program exits

The options are shown above in their command-line syntax. When entered as resources in ".Xdefaults", they require a different syntax.

Example:

      gnuplot*gray: on

gnuplot provides a command line option (-pointsize v) and a resource (gnuplot*pointsize: v) to control the size of points plotted with the points plotting style. The value v is a real number (greater than 0 and less than or equal to ten) used as a scaling factor for point sizes. For example, -pointsize 2 uses points twice the default size, and -pointsize 0.5 uses points half the normal size.

Plot windows will automatically be closed at the end of the session unless the -persist option was given.

For monochrome displays, gnuplot does not honor foreground or background colors. The default is black-on-white. -rv or gnuplot*reverseVideo: on requests white-on-black.

For color displays, gnuplot honors the following resources (shown here with their default values). The values may be color names as listed in the X11 rgb.txt file on your system, hexadecimal RGB color specifications (see X11 documentation), or a color name followed by a comma and an intensity value from 0 to 1. For example, blue,.5 means a half intensity blue.

 gnuplot*background: white
 gnuplot*textColor: black
 gnuplot*borderColor: black
 gnuplot*axisColor: black
 gnuplot*line1Color: red
 gnuplot*line2Color: green
 gnuplot*line3Color: blue
 gnuplot*line4Color: magenta
 gnuplot*line5Color: cyan
 gnuplot*line6Color: sienna
 gnuplot*line7Color: orange
 gnuplot*line8Color: coral

The command-line syntax for these is, for example,

Example:

      gnuplot -background coral

When -gray is selected, gnuplot honors the following resources for grayscale or color displays (shown here with their default values). Note that the default background is black.

 gnuplot*background: black
 gnuplot*textGray: white
 gnuplot*borderGray: gray50
 gnuplot*axisGray: gray50
 gnuplot*line1Gray: gray100
 gnuplot*line2Gray: gray60
 gnuplot*line3Gray: gray80
 gnuplot*line4Gray: gray40
 gnuplot*line5Gray: gray90
 gnuplot*line6Gray: gray50
 gnuplot*line7Gray: gray70
 gnuplot*line8Gray: gray30

gnuplot honors the following resources for setting the width (in pixels) of plot lines (shown here with their default values.) 0 or 1 means a minimal width line of 1 pixel width. A value of 2 or 3 may improve the appearance of some plots.

 gnuplot*borderWidth: 2
 gnuplot*axisWidth: 0
 gnuplot*line1Width: 0
 gnuplot*line2Width: 0
 gnuplot*line3Width: 0
 gnuplot*line4Width: 0
 gnuplot*line5Width: 0
 gnuplot*line6Width: 0
 gnuplot*line7Width: 0
 gnuplot*line8Width: 0

gnuplot honors the following resources for setting the dash style used for plotting lines. 0 means a solid line. A two-digit number jk (j and k are >= 1 and <= 9) means a dashed line with a repeated pattern of j pixels on followed by k pixels off. For example, '16' is a "dotted" line with one pixel on followed by six pixels off. More elaborate on/off patterns can be specified with a four-digit value. For example, '4441' is four on, four off, four on, one off. The default values shown below are for monochrome displays or monochrome rendering on color or grayscale displays. For color displays, the default for each is 0 (solid line) except for axisDashes which defaults to a '16' dotted line.

 gnuplot*borderDashes: 0
 gnuplot*axisDashes: 16
 gnuplot*line1Dashes: 0
 gnuplot*line2Dashes: 42
 gnuplot*line3Dashes: 13
 gnuplot*line4Dashes: 44
 gnuplot*line5Dashes: 15
 gnuplot*line6Dashes: 4441
 gnuplot*line7Dashes: 42
 gnuplot*line8Dashes: 13

The size or aspect ratio of a plot may be changed by resizing the gnuplot window.

The xlib terminal driver supports the X11 Windows System. It generates gnulib_x11 commands. set term x11 behaves similarly to set terminal xlib; set output "|gnuplot_x11". xlib has no options, but see x11.

      set terminal aifm {<color>} {"<fontname>"} {<fontsize>}

<color> is either color or monochrome; "<fontname>" is the name of a valid PostScript font; <fontsize> is the size of the font in PostScript points, before scaling by the set size command. Selecting default sets all options to their default values: monochrome, "Helvetica", and 14pt.

Since AI does not really support multiple pages, multiple graphs will be drawn directly on top of one another. However, each graph will be grouped individually, making it easy to separate them inside AI (just pick them up and move them).

Examples:

      set term aifm
      set term aifm 22
      set size 0.7,1.4; set term aifm color "Times-Roman" 14

      set terminal cgm {<mode>} {<color>} {<rotation>}
                       {width <plot_width>} {linewidth <line_width>}
                       {"<font>"} {<fontsize>}

where <mode> is landscape, portrait, or default; <color> is either color or monochrome; <rotation> is either rotate or norotate; <plot_width> is the width of the page in points; <line_width> is the line width in points; <font> is the name of a font; and <fontsize> is the size of the font in points.

By default, cgm uses rotated text for the Y axis label.

The first five options can be in any order. Selecting default sets all options to their default values.

Examples:

      set terminal cgm landscape color rotate width 432 linewidth 1 \
                     'Arial Bold' 12                  # defaults
      set terminal cgm 14 linewidth 2  14  # wider lines & larger font
      set terminal cgm portrait 'Times Roman Italic' 12

font
fontsize
linewidth
rotate
size
width
winword6

      Arial
      Arial Italic
      Arial Bold
      Arial Bold Italic
      Times Roman
      Times Roman Italic
      Times Roman Bold
      Times Roman Bold Italic
      Helvetica
      Roman

Fonts are scaled assuming the page is 6 inches wide. If the size command is used to change the aspect ratio of the page or the CGM file is converted to a different width (e.g. it is imported into a document in which the margins are not 6 inches apart), the resulting font sizes will be different. To change the assumed width, use the width option.

The linewidth option sets the width of lines in pt. The default width is 1 pt. Scaling is affected by the actual width of the page, as discussed under the fontsize and width options

All distances in the CGM file are in abstract units. The application that reads the file determines the size of the final page. By default, the width of the final page is assumed to be 6 inches (15.24 cm). This distance is used to calculate the correct font size, and may be changed with the width option. The keyword should be followed by the width in points. (Here, a point is 1/72 inch, as in PostScript. This unit is known as a "big point" in TeX.) gnuplot arithmetic can be used to convert from other units, as follows:

      set terminal cgm width 432            # default
      set terminal cgm width 6*72           # same as above
      set terminal cgm width 10/2.54*72     # 10 cm wide

                          | {monochrome | color
                               {<fontname> {"<fontsize>" 
                                  {<xsize> <ysize> {<linewidth> }}}}}

      set terminal dumb {[no]feed} {<xsize> <ysize>}

      set term dumb nofeed
      set term dumb 79 49 # VGA screen---why would anyone do that?

The dxf terminal driver creates pictures that can be imported into AutoCad (Release 10.x). It has no options of its own, but some features of its plots may be modified by other means. The default size is 120x80 AutoCad units, which can be changed by set size. dxf uses seven colors (white, red, yellow, green, cyan, blue and magenta), which can be changed only by modifying the source file. If a black-and-white plotting device is used, the colors are mapped to differing line thicknesses. See the description of the AutoCad print/plot command.

      set terminal fig {monochrome | color} {small | big}
                       {pointsmax <max_points>}
                       {landscape | portrait}
                       {metric | inches}
                       {fontsize <fsize>}
                       {size <xsize> <ysize>}
                       {thickness <units>}
                       {depth <layer>}

monochrome and color determine whether the picture is black-and-white or color. small and big produce a 5x3 or 8x5 inch graph in the default landscape mode and 3x5 or 5x8 inches in portrait mode. <max_points> sets the maximum number of points per polyline. Default units for editing with "xfig" may be metric or inches. fontsize sets the size of the text font to <fsize> points. size sets (overrides) the size of the drawing area to <xsize>*<ysize> in units of inches or centimeters depending on the inches or metric setting in effect. depth sets the default depth layer for all lines and text. The default depth is 10 to leave room for adding material with "xfig" on top of the plot.

thickness sets the default line thickness, which is 1 if not specified. Overriding the thickness can be achieved by adding a multiple of 100 to the to the linetype value for a plot command. In a similar way the depth of plot elements (with respect to the default depth) can be controlled by adding a multiple of 1000 to <linetype>. The depth is then <layer> + <linetype>/1000 and the thickness is (<linetype>%1000)/100 or, if that is zero, the default line thickness. Additional point-plot symbols are also available with the fig driver. The symbols can be used through pointtype values % 100 above 50, with different fill intensities controlled by <pointtype> % 5 and outlines in black (for <pointtype> % 10 < 5) or in the current color. Available symbols are

        50 - 59:  circles
        60 - 69:  squares
        70 - 79:  diamonds
        80 - 89:  upwards triangles
        90 - 99:  downwards triangles

The size of these symbols is linked to the font size. The depth of symbols is by default one less than the depth for lines to achieve nice error bars. If <pointtype> is above 1000, the depth is <layer> + <pointtype>/1000-1. If <pointtype>%1000 is above 100, the fill color is (<pointtype>%1000)/100-1.

Available fill colors are (from 1 to 9): black, blue, green, cyan, red, magenta, yellow, white and dark blue (in monochrome mode: black for 1 to 6 and white for 7 to 9).

See plot with for details of <linetype> and <pointtype>.

The big option is a substitute for the bfig terminal in earlier versions, which is no longer supported.

Examples:

      set terminal fig monochrome small pointsmax 1000  # defaults

      plot 'file.dat' with points linetype 102 pointtype 759

would produce circles with a blue outline of width 1 and yellow fill color.

      plot 'file.dat' using 1:2:3 with err linetype 1 pointtype 554

would produce errorbars with black lines and circles filled red. These circles are one layer above the lines (at depth 9 by default).

To plot the error bars on top of the circles use

      plot 'file.dat' using 1:2:3 with err linetype 1 pointtype 2554

      set terminal hp500c {<res>} {<comp>}

The international character sets ISO-8859-1 and CP850 are recognized via set encoding iso_8859_1 or set encoding cp850 (see set encoding for details).

Syntax:

      set terminal hpgl {<number_of_pens>} {eject}

The selection

      set terminal hpgl 8 eject

is equivalent to the previous hp7550 terminal, and the selection

      set terminal hpgl 4

is equivalent to the previous hp7580b terminal.

The pcl5 driver supports the Hewlett-Packard Laserjet III. It actually uses HPGL-2, but there is a name conflict among the terminal devices. It has several options

Syntax:

      set terminal pcl5 {<mode>} {<font>} {<fontsize>}

where <mode> is landscape, or portrait, <font> is stick, univers, or cg_times, and <fontsize> is the size in points.

With pcl5 international characters are handled by the printer; you just put the appropriate 8-bit character codes into the text strings. You don't need to bother with set encoding.

HPGL graphics can be imported by many software packages.

      set terminal hpljii | hpdj {<res>}

The hp500c terminal is similar to hpdj; hp500c additionally supports color and compression.

      set terminal hppj {FNT5X9 | FNT9X17 | FNT13X25}

      set terminal imagen {<fontsize>} {portrait | landscape}
                          {[<horiz>,<vert>]}

where fontsize defaults to 12 points and the layout defaults to landscape. <horiz> and <vert> are the number of graphs in the horizontal and vertical directions; these default to unity.

Example:

      set terminal imagen portrait [2,3]

puts six graphs on the page in three rows of two in portrait orientation.

The mif terminal driver produces Frame Maker MIF format version 3.00. It plots in MIF Frames with the size 15*10 cm, and plot primitives with the same pen will be grouped in the same MIF group. Plot primitives in a gnuplot page will be plotted in a MIF Frame, and several MIF Frames are collected in one large MIF Frame. The MIF font used for text is "Times".

Several options may be set in the MIF 3.00 driver.

Syntax:

      set terminal mif {colour | monochrome} {polyline | vectors}
                       {help | ?}

colour plots lines with line types >= 0 in colour (MIF sep. 2--7) and monochrome plots all line types in black (MIF sep. 0). polyline plots curves as continuous curves and vectors plots curves as collections of vectors. help and ? print online help on standard error output---both print a short description of the usage; help also lists the options;

Examples:

      set term mif colour polylines    # defaults
      set term mif                     # defaults
      set term mif vectors
      set term mif help

      set terminal pbm {<fontsize>} {<mode>}

where <fontsize> is small, medium, or large and <mode> is monochrome, gray or color. The default plot size is 640 pixels wide and 480 pixels high; this may be changed by set size.

The output of the pbm driver depends upon <mode>: monochrome produces a portable bitmap (one bit per pixel), gray a portable graymap (three bits per pixel) and color a portable pixmap (color, four bits per pixel).

The output of this driver can be used with Jef Poskanzer's excellent PBMPLUS package, which provides programs to convert the above PBMPLUS formats to GIF, TIFF, MacPaint, Macintosh PICT, PCX, X11 bitmap and many others.

Examples:

      set terminal pbm small monochrome             # defaults
      set size 2,2; set terminal pbm color medium

      set terminal png {small | medium | large}
                       {monochrome | gray | color}

      set terminal postscript {<mode>} {color | monochrome}
                              {solid | dashed} {<duplexing>}
                              {enhanced | noenhanced}
                              {"<fontname>"} {<fontsize>}

 Default size of a PostScript plot is 10 inches wide and 7 inches high.

      set terminal postscript default       # old postscript
      set terminal postscript enhanced      # old enhpost
      set terminal postscript landscape 22  # old psbig
      set terminal postscript eps 14        # old epsf1
      set terminal postscript eps 22        # old epsf2
      set size 0.7,1.4; set term post portrait color "Times-Roman" 14

enhanced postscript

 Control      Examples        Explanation
  ^           a^x             superscript
  _           a_x             subscript
  @           @x or a@^b_c    phantom box (occupies no width)
  &           &{space}        inserts space of specified length

If the encoding vector has been changed by set encoding, the default encoding vector can be used instead by following the slash with a dash. This is unnecessary if you use the Symbol font, however---since /Symbol uses its own encoding vector, gnuplot will not apply any other encoding vector to it.

The phantom box is useful for a@^b_c to align superscripts and subscripts, and for overwriting an accent on a letter. (The latter is tricky; it is much easier to use set encoding iso_8859_1 to change to the ISO Latin-1 encoding vector, which contains a large variety of letters with accents or other diacritical marks.) It is common sense to put the shorter of the two in the phantom box.

Space equal in length to a string can be inserted using the '&' character. Thus

        'abc&{def}ghi'

would produce

        'abc   ghi'.

You can access special symbols numerically by specifying \character-code (in octal), e.g., {/Symbol \245} is the symbol for infinity.

You can escape control characters using \, e.g., \\, \{, and so on.

But be aware that strings in double-quotes are parsed differently than those enclosed in single-quotes. The major difference is that backslashes may need to be doubled when in double-quoted strings.

Examples (these are hard to describe in words---try them!):

      set xlabel 'Time (10^6 {/Symbol m}s)'
      set title '{/Symbol=18 \362@_{/=9.6 0}^{/=12 x}} \
                 {/Helvetica e^{-{/Symbol m}^2/2} d}{/Symbol m}'

The file "ps_guide.ps" in the /docs subdirectory of the gnuplot source distribution contains more examples of the enhanced syntax.

Instead of producing a picture, the table terminal prints out evaluation results in a multicolumn ASCII table of X Y Z values. For those times when you really want to see the numbers, now you can see them on the screen or save to a file. This can be useful if you want to use the contouring engine of gnuplot to work out the contours of your data and then save them for further use, perhaps for plotting with plot. See set contour for an example. The same trick can be used to save gridded data (set dgrid3d).

The tgif driver supports different pointsizes (with set pointsize), different label fonts and font sizes (e.g. set label "Hallo" at x,y font "Helvetica,34") and multiple graphs on the page. The proportions of the axes are not changed.

Syntax:

      set terminal tgif {portrait | landscape} {<[x,y]>}
                        {solid | dashed}
                        {"<fontname>"} {<fontsize>}

where <[x,y]> specifies the number of graphs in the x and y directions on the page, "<fontname>" is the name of a valid PostScript font, and <fontsize> specifies the size of the PostScript font. Defaults are portrait, [1,1], dashed, "Helvetica", and 18.

The solid option is usually prefered if lines are colored, as they often are in the editor. Hardcopy will be black-and-white, so dashed should be chosen for that.

Multiplot is implemented in two different ways.

The first multiplot implementation is the standard gnuplot multiplot feature:

      set terminal tgif
      set output "file.obj"
      set multiplot
      set origin x01,y01
      set size  xs,ys
      plot ...
           ...
      set origin x02,y02
      plot ...
      set nomultiplot

See set multiplot for further information.

The second version is the [x,y] option for the driver itself. The advantage of this implementation is that everything is scaled and placed automatically without the need for setting origins and sizes; the graphs keep their natural x/y proportions of 3/2 (or whatever is fixed by set size).

If both multiplot methods are selected, the standard method is chosen and a warning message is given.

Examples of single plots (or standard multiplot):

      set terminal tgif                  # defaults
      set terminal tgif "Times-Roman" 24
      set terminal tgif landscape
      set terminal tgif landscape solid

Examples using the built-in multiplot mechanism:

      set terminal tgif portrait [2,4]  # portrait; 2 plots in the x-
                                        # and 4 in the y-direction
      set terminal tgif [1,2]           # portrait; 1 plot in the x-
                                        # and 2 in the y-direction
      set terminal tgif landscape [3,3] # landscape; 3 plots in both
                                        # directions

This terminal driver generates tk canvas widget commands. To use it, rebuild gnuplot (after uncommenting or inserting the appropriate line in "term.h"), then

 gnuplot> set term tkcanvas
 gnuplot> set output 'plot.file'

After invoking "wish", execute the following sequence of tcl commands:

 % source plot.file
 % canvas .c
 % pack .c
 % gnuplot .c

The code generated by gnuplot creates a tcl procedure called "gnuplot" that takes the name of a canvas as its argument. When the procedure is, called, it clears the canvas, finds the size of the canvas and draws the plot in it, scaled to fit.

The current version of tkcanvas supports neither multiplot nor replot.

      set terminal nec-cp6 {monochrome | colour | draft}

which defaults to monochrome. With each of these drivers, a binary copy is required on a PC to print. Do not use print---use instead copy file /b lpt1:.

      set terminal latex | emtex {courier | roman} {<fontsize>}

fontsize may be any size you specify. The default is 10-point Roman.

Unless your driver is capable of building fonts at any size (e.g. dvips), stick to the standard 10, 11 and 12 point sizes.

METAFONT users beware: METAFONT does not like odd sizes.

      set terminal pslatex | |pstex {<color>} {<dashed>} {<rotate>}
                                    {auxfile} {<font_size>}

<color> is either color or monochrome. <rotate> is either rotate or norotate and determines if the y-axis label is rotated. <font_size> is used to scale the font from its usual size.

If auxfile is specified, it directs the driver to put the PostScript commands into an auxiliary file instead of directly into the LaTeX file. This is useful if your pictures are large enough that dvips cannot handle them. The name of the auxiliary PostScript file is derived from the name of the TeX file given on the set output command; it is determined by replacing the trailing .tex (actually just the final extent in the file name---and the option will be turned off if there is no extent) with .ps in the output file name. Remember to close the file before leaving gnuplot.

Examples:

      set term pslatex monochrome dashed rotate       # set to defaults

To write the PostScript commands into the file "foo.ps":

      set term pslatex auxfile
      set output "foo.tex"; plot ...: set output

The eepic terminal driver supports the extended LaTeX picture environment. It is an alternative to the latex driver.

The output of this terminal is intended for use with the "eepic.sty" macro package for LaTeX. To use it, you need "eepic.sty", "epic.sty" and a printer driver that supports the "tpic" \specials. If your printer driver doesn't support those \specials, "eepicemu.sty" will enable you to use some of them.

Although dotted and dashed lines are possible with eepic and are tempting, they do not work well for high-sample-rate curves, fusing the dashes all together into a solid line. For now, the eepic driver creates only solid lines. There is another gnuplot driver (tpic) that supports dashed lines, but it cannot be used if your DVI driver doesn't support "tpic" \specials.

The eepic terminal has no options.

The tpic terminal driver supports the LaTeX picture environment with tpic \specials. It is an alternative to the latex and eepic terminal drivers. Options are the point size, line width, and dot-dash interval.

Syntax:

      set terminal tpic <pointsize> <linewidth> <interval>

where pointsize and linewidth are integers in milli-inches and interval is a float in inches. If a non-positive value is specified, the default is chosen: pointsize = 40, linewidth = 6, interval = 0.1.

The pstricks driver is intended for use with the "pstricks.sty" macro package for LaTeX. It is an alternative to the eepic and latex drivers. You need "pstricks.sty", and, of course, a printer that understands PostScript. Ghostscript understands PostScript, too.

PSTricks is available via anonymous ftp from the /pub directory at Princeton.EDU. This driver definitely does not come close to using the full capability of the PSTricks package.

Syntax:

      set terminal pstricks {hacktext | nohacktext} {unit | nounit}

The first option invokes an ugly hack that gives nicer numbers; the second has to do with plot scaling. The defaults are hacktext and nounit.

 It has no options.

To use a picture in a document, the METAFONT program must be run with the output file from gnuplot as input. Thus, the user needs a basic knowledge of the font creating process and the procedure for including a new font in a document. However, if the METAFONT program is set up properly at the local site, an unexperienced user could perform the operation without much trouble.

The text support is based on a METAFONT character set. Currently the Computer Modern Roman font set is input, but the user is in principal free to chose whatever fonts he or she needs. The METAFONT source files for the chosen font must be available. Each character is stored in a separate picture variable in METAFONT. These variables may be manipulated (rotated, scaled etc.) when characters are needed. The drawback is the interpretation time in the METAFONT program. On some machines (i.e. PC) the limited amount of memory available may also cause problems if too many pictures are stored.

The mf terminal has no options.

METAFONT Instructions

  set terminal mf

  set output "myfigures.mf"

- Quit gnuplot.

- Generate a TFM and GF file by running METAFONT on the output of gnuplot. Since the picture is quite large (5*3 in), you will have to use a version of METAFONT that has a value of at least 150000 for memmax. On Unix systems these are conventionally installed under the name bigmf. For the following assume that the command virmf stands for a big version of METAFONT. For example:

- Invoke METAFONT:

    virmf '&plain'

- Select the output device: At the METAFONT prompt ('*') type:

    \mode:=CanonCX;     % or whatever printer you use

- Optionally select a magnification:

    mag:=1;             % or whatever you wish

- Input the gnuplot-file:

    input myfigures.mf

On a typical Unix machine there will usually be a script called "mf" that executes virmf '&plain', so you probably can substitute mf for virmf &plain. This will generate two files: mfput.tfm and mfput.$$$gf (where $$$ indicates the resolution of your device). The above can be conveniently achieved by typing everything on the command line, e.g.: virmf '&plain' '\mode:=CanonCX; mag:=1; input myfigures.mf' In this case the output files will be named myfigures.tfm and myfigures.300gf.

- Generate a PK file from the GF file using gftopk:

  gftopk myfigures.300gf myfigures.300pk

The name of the output file for gftopk depends on the DVI driver you use. Ask your local TeX administrator about the naming conventions. Next, either install the TFM and PK files in the appropriate directories, or set your environment variables properly. Usually this involves setting TEXFONTS to include the current directory and doing the same thing for the environment variable that your DVI driver uses (no standard name here...). This step is necessary so that TeX will find the font metric file and your DVI driver will find the PK file.

- To include your pictures in your document you have to tell TeX the font:

  \font\gnufigs=myfigures

Each picture you made is stored in a single character. The first picture is character 0, the second is character 1, and so on... After doing the above step, you can use the pictures just like any other characters. Therefore, to place pictures 1 and 2 centered in your document, all you have to do is:

  \centerline{\gnufigs\char0}
  \centerline{\gnufigs\char1}

in plain TeX. For LaTeX you can, of course, use the picture environment and place the picture wherever you wish by using the \makebox and \put macros.

This conversion saves you a lot of time once you have generated the font; TeX handles the pictures as characters and uses minimal time to place them, and the documents you make change more often than the pictures do. It also saves a lot of TeX memory. One last advantage of using the METAFONT driver is that the DVI file really remains device independent, because no \special commands are used as in the eepic and tpic drivers.

      set tics {<direction>}
      show tics

See also set xtics for more control of major (labelled) tic marks and set mxtics for control of minor tic marks.

Using splot, one can adjust the relative height of the vertical (Z) axis using set ticslevel. The numeric argument provided specifies the location of the bottom of the scale (as a fraction of the z-range) above the xy-plane. The default value is 0.5. Negative values are permitted, but tic labels on the three axes may overlap.

To place the xy-plane at a position 'pos' on the z-axis, ticslevel should be set equal to (pos - zmin) / (zmin - zmax).

Syntax:

      set ticslevel {<level>}
      show tics

See also set view.

      set ticscale {<major> {<minor>}}
      show tics

      set timestamp {"<format>"} {top|bottom} {{no}rotate}
                    {<xoff>}{,<yoff>} {"<font>"}
      set notimestamp
      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 <off> are offsets from the default position given in character screen coordinates. <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.

Example:

      set timestamp "%d/%m/%y %H:%M" 80,-2 "Helvetica"

See set timefmt for more information about time format strings.

      set timefmt "<format string>"
      show timefmt

The string argument tells gnuplot how to read timedata from the datafile. The valid formats are:

      Format       Explanation
      %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           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 date/time 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 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 and Time/date for more information.

Example:

      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.) Time Data Demo

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.

Syntax:

      set title {"<title-text>"} {<xoff>}{,<yoff>} {"<font>,{<size>}"}
      show title

Specifying constants <xoff> or <yoff> as optional offsets for the title will move the title <xoff> or <yoff> character screen coordinates (not graph coordinates). For example, "set title ,-1" will change only the y offset of the title, moving the title down by roughly the height of one character.

<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.

set title with no parameters clears the title.

See syntax for details about the processing of backslash sequences and the distinction between single- and double-quotes.

The command set tmargin sets the size of the top margin. Please see set margin for details.

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 for details.

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 for details.

      show variables

The set view command sets the viewing angle for splots. It controls how the 3-d coordinates of the plot are mapped into the 2-d screen space. It provides controls for both rotation and scaling of the plotted data, but supports orthographic projections only.

Syntax:

      set view <rot_x> {,{<rot_z>}{,{<scale>}{,<scale_z>}}}
      show view

where <rot_x> and <rot_z> control the rotation angles (in degrees) in a virtual 3-d coordinate system aligned with the screen such that initially (that is, before the rotations are performed) the screen horizontal axis is x, screen vertical axis is y, and the axis perpendicular to the screen is z. The first rotation applied is <rot_x> around the x axis. The second rotation applied is <rot_z> around the new z axis.

<rot_x> is bounded to the [0:180] range with a default of 60 degrees, while <rot_z> is bounded to the [0:360] range with a default of 30 degrees. <scale> controls the scaling of the entire splot, while <scale_z> scales the z axis only. Both scales default to 1.0.

Examples:

      set view 60, 30, 1, 1
      set view ,,0.5

The first sets all the four default values. The second changes only scale, to 0.5.

See also set ticslevel.

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 for details.

The set x2data command sets data on the x2 (top) axis to timeseries (dates/times). Please see set xdata.

The set x2dtics command changes tics on the x2 (top) axis to days of the week. Please see set xmtics for details.

The set x2label command sets the label for the x2 (top) axis. Please see set xlabel.

The set x2mtics command changes tics on the x2 (top) axis to months of the year. Please see set xmtics for details.

The set x2range command sets the horizontal range that will be displayed on the x2 (top) axis. Please see set xrange for details.

The set x2tics command controls major (labelled) tics on the x2 (top) axis. Please see set xtics for details.

The set x2zeroaxis command draws a line at the origin of the x2 (top) axis (x2 = 0). For details, please see set zeroaxis.

      set xdata {time}
      show xdata

The same syntax applies to ydata, zdata, x2data and y2data.

The time option signals that the datatype is indeed date/time. If the option is not specified, the datatype reverts to normal.

See set timefmt to tell gnuplot how to read date or time data. The date/time is converted to seconds from start of the century. There is currently only one timefmt, which implies that all the date/time columns must confirm to this format. Specification of ranges should be supplied as quoted strings according to this format to avoid interpretation of the date/time as an expression.

The function 'strftime' (type "man strftime" on unix to look it up) is used to print ticmark labels. gnuplot tries to figure out a reasonable format for this unless the set format x "string" has supplied something that does not look like a decimal format (more than one '%' or neither %f nor %g).

See also Time/date for more information.

      set xdtics
      set noxdtics
      show xdtics

The same syntax applies to ydtics, zdtics, x2dtics and y2dtics.

See also the set format command.

      set xlabel {"<label>"} {<xoff>}{,<yoff>} {"<font>{,<size>}"}
      show xlabel

The same syntax applies to x2label, ylabel, y2label and zlabel.

Specifying the constants <xoff> or <yoff> as optional offsets for a label will move it <xoff> or <yoff> character widths or heights. For example, " set xlabel -1" will change only the x offset of the xlabel, moving the label roughly one character width to the left. The size of a character depends on both the font and the terminal.

<font> is used to specify the font in which the label is written; the units of the font <size> depend upon which terminal is used.

To clear a label, put no options on the command line, e.g., "set y2label".

The default positions of the axis labels are as follows:

xlabel: The x-axis label is centered below the bottom axis.

ylabel: The position of the y-axis label depends on the terminal, and can be one of the following three positions:

1. Horizontal text flushed left at the top left of the plot. Terminals that cannot rotate text will probably use this method. If set x2tics is also in use, the ylabel may overwrite the left-most x2tic label. This may be remedied by adjusting the ylabel position or the left margin.

2. Vertical text centered vertically at the left of the plot. Terminals that can rotate text will probably use this method.

3. Horizontal text centered vertically at the left of the plot. The EEPIC, LaTeX and TPIC drivers use this method. The user must insert line breaks using \\ to prevent the ylabel from overwriting the plot. To produce a vertical row of characters, add \\ between every printing character (but this is ugly).

zlabel: The z-axis label is centered along the z axis and placed in the space above the grid level.

y2label: The y2-axis label is placed to the right of the y2 axis. The position is terminal-dependent in the same manner as is the y-axis label.

x2label: The x2-axis label is placed above the top axis but below the plot title. It is also possible to create an x2-axis label by using new-line characters to make a multi-line plot title, e.g.,

      set title "This is the title\n\nThis is the x2label"

Note that double quotes must be used. The same font will be used for both lines, of course.

If you are not satisfied with the default position of an axis label, use set label instead--that command gives you much more control over where text is placed.

Please see set syntax for further information about backslash processing and the difference between single- and double-quoted strings.

      set xmtics
      set noxmtics
      show xmtics

The same syntax applies to x2mtics, ymtics, y2mtics, and zmtics.

See also the set format command.

      set xrange [{{<min>}:{<max>}}] {{no}reverse} {{no}writeback}
      show xrange

where <min> and <max> terms are constants, expressions or an asterisk to set autoscaling. If the data are date/time, you must give the range as a quoted string according to the set timefmt format. Any value omitted will not be changed.

The same syntax applies to yrange, zrange, x2range, y2range, rrange, trange, urange and vrange.

The reverse option reverses the direction of the axis, e.g., set xrange [0:1] reverse will produce an axis with 1 on the left and 0 on the right. This is identical to the axis produced by set xrange [1:0], of course. reverse is intended primarily for use with autoscale.

The writeback option essentially saves the range found by autoscale in the buffers that would be filled by set xrange. This is useful if you wish to plot several functions together but have the range determined by only some of them. The writeback operation is performed during the plot execution, so it must be specified before that command. For example,

      set xrange [-10:10]
      set yrange [] writeback
      plot sin(x)
      set noautoscale y
      replot x/2

results in a yrange of [-1:1] as found only from the range of sin(x); the [-5:5] range of x/2 is ignored. Executing show yrange after each command in the above example should help you understand what is going on.

In 2-d, xrange and yrange determine the extent of the axes, trange determines the range of the parametric variable in parametric mode or the range of the angle in polar mode. Similarly in parametric 3-d, xrange, yrange, and zrange govern the axes and urange and vrange govern the parametric variables.

In polar mode, rrange determines the radial range plotted. <rmin> acts as an additive constant to the radius, whereas <rmax> acts as a clip to the radius---no point with radius greater than <rmax> will be plotted. xrange and yrange are affected---the ranges can be set as if the graph was of r(t)-rmin, with rmin added to all the labels.

Any range may be partially or totally autoscaled, although it may not make sense to autoscale a parametric variable unless it is plotted with data.

Ranges may also be specified on the plot command line. A range given on the plot line will be used for that single plot command; a range given by a set command will be used for all subsequent plots that do not specify their own ranges. The same holds true for splot.

Examples:

To set the xrange to the default:

      set xrange [-10:10]

To set the yrange to increase downwards:

      set yrange [10:-10]

To change zmax to 10 without affecting zmin (which may still be autoscaled):

      set zrange [:10]

To autoscale xmin while leaving xmax unchanged:

      set xrange [*:]

      set xtics {axis | border} {{no}mirror} {{no}rotate}
                {  <incr>
                 | <start>, <incr> {,<end>}
                 | ({"<label>"} <pos> {,{"<label>"} <pos>}...) }
      set noxtics
      show xtics

The same syntax applies to ytics, ztics, x2tics and y2tics.

axis or border tells gnuplot to put the tics (both the tics themselves and the accompanying labels) along the axis or the border, respectively. mirror tells it to put unlabelled tics at the same positions on the opposite border. nomirror does what you think it does. rotate asks gnuplot to rotate the text through 90 degrees, if the underlying terminal driver supports text rotation. norotate cancels this. The defaults are border mirror norotate for tics on the x, y, x2, and y2 axes. For the z axis, the the {axis | border} option is not available and the default is nomirror. If you do want to mirror the z-axis tics, you might want to create a bit more room for them with set border.

The positions of the tics may be specified in either of two forms:

The <start>, <incr>, <end> form specifies that a series of tics will be plotted on the axis between the values <start> and <end> with an increment of <incr>. If <end> is not given, it is assumed to be infinity. The increment may be negative. If neither <start> nor <end> is given, <start> is assumed to be negative infinity, <end> is assumed to be positive infinity, and the tics will be drawn at multiples of <step>---there will be a tic at zero (if it is within the plotted range). If the axis is logarithmic, the increment will be used as a multiplicative factor.

Examples:

Make tics at 0, 0.5, 1, 1.5, ..., 9.5, 10.

      set xtics 0,.5,10

Make tics at ..., -10, -5, 0, 5, 10, ...

      set xtics 5

Make tics at 1, 100, 1e4, 1e6, 1e8.

      set logscale x; set xtics 1,100,10e8

The ("<label>" <pos>, ...) form allows arbitrary tic positions or non-numeric tic labels. A set of tics is a set of positions, each with its own optional label. Note that the label is a string enclosed by quotes, and may be a constant string, such as "hello", or contain formatting information for the tic number (which is the same as the position), such as "%3f clients". See set format for more information about this case. The label may be made empty by specifying it as an empty string. If no string is given, the default label (numerical) is used. In this form, the tics do not need to be listed in numerical order.

Examples:

      set xtics ("low" 0, "medium" 50, "high" 100)
      set xtics (1,2,4,8,16,32,64,128,256,512,1024)
      set ytics ("bottom" 0, "" 10, "top" 20)

In the second example, all tics are labelled. In the third, only the end tics are labelled.

Tics will only be plotted when in range.

Minor (unlabelled) tics can be added by the set mxtics command.

In case of timeseries data, position values must be given as quoted dates or times according to the format timefmt. If the <start>, <incr>, <end> form is used, <start> and <end> must be given according to timefmt, but <incr> must be in seconds. Times will be written out according to the format given on "set format", however.

Examples:

      set xdata time
      set timefmt "%d/%m"
      set format x "%b %d"
      set xrange ["00/12":"06/12"]
      set xtics "01/12", 172800, "05/12"

      set xdata time
      set timefmt "%d/%m"
      set format x "%b %d"
      set xrange ["00/12":"06/12"]
      set xtics ("01/12", "" "03/12", "05/12")

Both of these will produce tics "Dec 1", "Dec 3", and "Dec 5", but in the second example the tic at "Dec 3" will be unlabelled.

The set xzeroaxis command draws a line at x = 0. For details, please see set zeroaxis.

The set y2data command sets y2 (right-hand) axis data to timeseries (dates/times). Please see set xdata.

The set y2dtics command changes tics on the y2 (right-hand) axis to days of the week. Please see set xmtics for details.

The set y2dtics command sets the label for the y2 (right-hand) axis. Please see set xlabel.

The set y2mtics command changes tics on the y2 (right-hand) axis to months of the year. Please see set xmtics for details.

The set y2range command sets the vertical range that will be displayed on the y2 (right-hand) axis. Please see set xrange for details.

The set y2tics command controls major (labelled) tics on the y2 (right-hand) axis. Please see set xtics for details.

The set y2zeroaxis command draws a line at the origin of the y2 (right-hand) axis (y2 = 0). For details, please see set zeroaxis.

Sets y-axis data to timeseries (dates/times). Please see set xdata.

The set ydtics command changes tics on the y axis to days of the week. Please see set xmtics for details.

This command sets the label for the y axis. Please see set xlabel.

The set ymtics command changes tics on the y axis to months of the year. Please see set xmtics for details.

The set yrange command sets the vertical range that will be displayed on the y axis. Please see set xrange for details.

The set ytics command controls major (labelled) tics on the y axis. Please see set xtics for details.

The set yzeroaxis command draws a line at y = 0. For details, please see set zeroaxis.

Set zaxis date to timeseries (dates/times). Please see set xdata.

The set zdtics command changes tics on the z axis to days of the week. Please see set xmtics for details.

      set zero <expression>
      show zero

gnuplot will not plot a point if its imaginary part is greater in magnitude than the zero threshold. Axis ranges cannot be less than zero. The default zero value is 1e-8.

The x axis may be drawn by set xzeroaxis and removed by set noxzeroaxis. Similar commands behave similarly for the y, x2, and y2 axes.

Syntax:

      set zeroaxis {<linetype>}
      set xzeroaxis {<linetype>}
      set yzeroaxis {<linetype>}
      set x2zeroaxis {<linetype>}
      set y2zeroaxis {<linetype>}
      set nozeroaxis
      set noxzeroaxis
      etc.
      show zeroaxis
      show xzeroaxis
      etc.

By default, these options are off. The selected zero axis is drawn with a line of type <linetype> from the default linetype list provided by the terminal; user-defined linetypes (via the set linestyle command) are not accessible for these axes. If <linetype> is not specified, any zero axes selected will be drawn using the axis linetype (linetype 0).

set zeroaxis l is equivalent to set xzeroaxis l; set yzeroaxis l. set nozeroaxis is equivalent to set noxzeroaxis; set noyzeroaxis.

This command sets the label for the z axis. Please see set xlabel.

The set zmtics command changes tics on the z axis to months of the year. Please see set xmtics for details.

The set zrange command sets the range that will be displayed on the z axis. The zrange is used only by splot and is ignored by plot. Please see set xrange for details.

The set ztics command controls major (labelled) tics on the z axis. Please see set xtics for details.

The shell command spawns an interactive shell. To return to gnuplot, type logout if using VMS, exit or the END-OF-FILE character if using Unix, endcli if using AmigaDOS, or exit if using MS-DOS or OS/2.

A single shell command may be spawned by preceding it with the ! character ($ if using VMS) at the beginning of a command line. Control will return immediately to gnuplot after this command is executed. For example, in Unix, AmigaDOS, MS-DOS or OS/2,

      ! dir

prints a directory listing and then returns to gnuplot.

On an Atari, the ! command first checks whether a shell is already loaded and uses it, if available. This is practical if gnuplot is run from gulam, for example.

splot is the primary command for drawing 3-d plots (well, actually projections on a 2-d surface, but you knew that). It can create a plot from functions or data in a manner very similar to the plot command.

Please see plot for features common to the plot command; only differences are discussed in detail here.

Syntax:

      splot {<ranges>}
            {<function> | {"<datafile>" {datafile-modifiers}}}
            {<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 triple (splot) of mathematical expressions in parametric mode.

By default splot draws the xy plane completely below the plotted data. The offset between the lowest ztic and the xy plane can be changed by set ticslevel. The orientation of a 'splot' is controlled by set view. See set view and set ticslevel for more information.

The syntax for setting ranges on the splot command is the same as for plot. In non-parametric mode, the order in which ranges must be given is xrange, yrange, and zrange. In parametric mode, the order is urange, vrange, xrange, yrange, and zrange.

The title option is the same as in plot. The operation of with is also the same as in plot, except that the plotting styles available to splot are limited to lines, points, linespoints, dots, and impulses; the error-bar capabilities of plot are not available for splot.

The datafile options have more differences.

data-file

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

Syntax:

      splot '<file_name>' {binary | matrix}
                          {index <index list>}
                          {every <every list>}
                          {using <using list>}

The special filenames "" and "-" are permitted, as in plot.

In brief, binary indicates that the file is binary, matrix indicates that the data are in matrix form, index selects which data sets in a multi-data-set file are to be plotted, every specifies which datalines within a single data set are to be plotted, and using determines how the columns within a single record are to be interpreted.

The options index and every behave the same way as with plot. using also does, with the obvious difference that the using list must provide three entries instead of two.

The plot options thru and smooth are not available for splot.

Data file organization is essentially the same as for plot, except that each point is an (x,y,z) triple. If only a single value is provided, it will be used for z, the data point number will be used for x, and the y-isoline number will be used for y; thus "splot 'file' using 1" is identical to "splot 'file' using 0:-1:1". If two values are provided, gnuplot gives you an error message. Three values are interpreted as an (x,y,z) triple. Additional values are generally used as errors, which can be used by fit.

Single blank records separate datalines (which are interpreted as y-isolines) in a splot datafile. No line will join points separated by a blank record. If all datalines contain the same number of points,gnuplot will draw cross-isolines in the opposite direction. This is termed "grid data", and is required for drawing a surface, for contouring (set contour) and hidden-line removal (set hidden3d).

It is no longer necessary to specify parametric mode for three-column splots.

binary
example datafile
matrix

In previous versions, gnuplot dynamically detected binary data files. It is now necessary to specify the keyword binary directly after the filename.

Single precision floats are stored in a binary file as follows:

      <N+1>  <y0>   <y1>   <y2>  ...  <yN>
       <x0> <z0,0> <z0,1> <z0,2> ... <z0,N>
       <x1> <z1,0> <z1,1> <z1,2> ... <z1,N>
        :      :      :      :   ...    :

which are converted into triplets:

      <x0> <y0> <z0,0>
      <x0> <y1> <z0,1>
      <x0> <y2> <z0,2>
       :    :     :
      <x0> <yN> <z0,N>

      <x1> <y0> <z1,0>
      <x1> <y1> <z1,1>
       :    :     :

These triplets are then converted into gnuplot iso-curves and then gnuplot proceeds in the usual manner to do the rest of the plotting.

A collection of matrix and vector manipulation routines (in C) is provided in binary.c. The routine to write binary data is

      int fwrite_matrix(file,m,nrl,nrl,ncl,nch,row_title,column_title)

An example of using these routines is provided in the file bf_test.c, which generates binary files for the demo file demo/binary.dem.

The index keyword is not supported, since the file format allows only one surface per file. The every and using filters are supported. using operates as if the data were read in the above triplet form. Binary File Splot Demo.

      splot 'datafile.dat'

      # The valley of the Gnu.
         0 0 10
         0 1 10
         0 2 10

         1 0 10
         1 1 5
         1 2 10

         2 0 10
         2 1 1
         2 2 10

         3 0 10
         3 1 0
         3 2 10

Actually for grid data it is not necessary to keep the x values constant within an dataline, nor is it necessary to keep the y values the same along the perpendicular datalines. gnuplot requires only that the number of points be the same for each dataline.

      z11 z12 z13 z14 ...
      z21 z22 z23 z24 ...
      z31 z32 z33 z34 ...

      test

This command writes the current values of the fit parameters into the given file, which is formatted as an initial-value file (as described in the fit section). This is useful for saving the current values for later use or for restarting a converged or stopped fit.

Syntax:

      update <filename> {<filename>}

If the file already exists, gnuplot first renames it by appending .old and then opens a new file. That is, "update 'fred'" behaves the same way as "!rename fred fred.old; update 'fred.old' 'fred'". [On DOS and other systems that use the twelve-character "filename.ext" naming convention, "ext" will be "old" and "filename" will be related (hopefully recognizably) to the initial name. Renaming is not done at all on VMS systems, since they use file-versioning.]

If a second filename is supplied, the updated values are written to this file instead, and the original parameter file is left unmodified.

Please see fit for more information.

Several graphical user interfaces have been written for gnuplot and one for win32 is included in this distribution. In addition, there is a Macintosh interface at

       ftp://ftp.ee.gatech.edu/pub/mac/gnuplot

and several X11 interfaces include three Tcl/Tk located at the usual Tcl/Tk repositories.

There is a bug in the stdio library for old Sun operating systems (SunOS Sys4-3.2). The "%g" format for 'printf' sometimes incorrectly prints numbers (e.g., 200000.0 as "2"). Thus, tic mark labels may be incorrect on a Sun4 version of gnuplot. A work-around is to rescale the data or use the set format command to change the tic mark format to "%7.0f" or some other appropriate format. This appears to have been fixed in SunOS 4.0.

Another bug: On a Sun3 under SunOS 4.0, and on Sun4's under Sys4-3.2 and SunOS 4.0, the 'sscanf' routine incorrectly parses "00 12" with the format "%f %f" and reads 0 and 0 instead of 0 and 12. This affects data input. If the data file contains x coordinates that are zero but are specified like '00', '000', etc, then you will read the wrong y values. Check any data files or upgrade the SunOS. It appears to have been fixed in SunOS 4.1.1.

Suns appear to overflow when calculating exp(-x) for large x, so gnuplot gets an undefined result. One work-around is to make a user-defined function like e(x) = x<-500 ? 0 : exp(x). This affects plots of Gaussians (exp(-x*x)) in particular, since x*x grows quite rapidly.

Microsoft C 5.1 has a nasty bug associated with the %g format for 'printf'. When any of the formats "%.2g", "%.1g", "%.0g", "%.g" are used, 'printf' will incorrectly print numbers in the range 1e-4 to 1e-1. Numbers that should be printed in the %e format are incorrectly printed in the %f format, with the wrong number of zeros after the decimal point. To work around this problem, use the %e or %f formats explicitly.

gnuplot, when compiled with Microsoft C, did not work correctly on two VGA displays that were tested. The CGA, EGA and VGA drivers should probably be rewritten to use the Microsoft C graphics library. gnuplot compiled with Borland C++ uses the Turbo C graphics drivers and does work correctly with VGA displays.

VAX/VMS 4.7 C compiler release 2.4 also has a poorly implemented %g format for 'printf'. The numbers are printed numerically correct, but may not be in the requested format. The K&R second edition says that for the %g format, %e is used if the exponent is less than -4 or greater than or equal to the precision. The VAX uses %e format if the exponent is less than -1. The VAX appears to take no notice of the precision when deciding whether to use %e or %f for numbers less than 1. To work around this problem, use the %e or %f formats explicitly. From the VAX C 2.4 release notes: e,E,f,F,g,G Result will always contain a decimal point. For g and G, trailing zeros will not be removed from the result.

VAX/VMS 5.2 C compiler release 3.0 has a slightly better implemented %g format than release 2.4, but not much. Trailing decimal points are now removed, but trailing zeros are still not removed from %g numbers in exponential format.

The two preceding problems are actually in the libraries rather than in the compilers. Thus the problems will occur whether gnuplot is built using either the DEC compiler or some other one (e.g. the latest gcc).

ULTRIX X11R3 has a bug that causes the X11 driver to display "every other" graph. The bug seems to be fixed in DEC's release of X11R4 so newer releases of ULTRIX don't seem to have the problem. Solutions for older sites include upgrading the X11 libraries (from DEC or direct from MIT) or defining ULTRIX_KLUDGE when compiling the x11.trm file. Note that the kludge is not an ideal fix, however.

The constant HUGE was incorrectly defined in the NeXT OS 2.0 operating system. HUGE should be set to 1e38 in plot.h. This error has been corrected in the 2.1 version of NeXT OS.

Some older models of HP plotters do not have a page eject command 'PG'. The current HPGL driver uses this command in HPGL_reset. This may need to be removed for these plotters. The current PCL5 driver uses HPGL/2 for text as well as graphics. This should be modified to use scalable PCL fonts.

On the Atari version, it is not possible to send output directly to the printer (using /dev/lp as output file), since CRs are added to LFs in binary output. As a work-around, write the output to a file and copy it to the printer afterwards using a shell command.

On AIX 4, the literal 'NaNq' in a datafile causes the special internal value 'not-a-number' to be stored, rather than setting an internal 'undefined' flag. A workaround is to use set missing 'NaNq'.

There may be an up-to-date list of bugs since the release on the WWW page:

      http://www.cs.dartmouth.edu/gnuplot

Please report any bugs to bug-gnuplot@dartmouth.edu.