June 20, 20242 min read

GNUplot Cheatsheet

Plotting a column against its row index, and building normalized histograms from a single-column data file.

#gnuplot#plotting

Plot a column against its row index

plot "data.txt" u ($0):4 title "Plotting with x innumeration"

$0 is the line number, so this plots column 4 against its position in the file.

Column 0 is the pseudo-column $0 — the zero-based row number — so plot 'file' using 1 is equivalent to plot 'file' using 0:1.

Plot a histogram

set terminal png
set output 'histogram.png'
set style data histogram
set style fill solid
set xlabel 'Angles(°)'
set ylabel 'Normalized Frequency'
set title 'Angles distribution for α'
bin_width = 1
bin(x, width) = width * floor(x / width) + bin_width / 2.0
stats 'angles.txt' nooutput
total_points = STATS_records
plot 'angles.txt' using (bin($1, bin_width)):(1.0/total_points) smooth freq with boxes lc rgb 'blue' notitle

bin_width controls the sampling density (bin width).
Dividing by total_points normalizes the histogram — drop that division for an unnormalized count.
angles.txt is assumed to be a single column of values.

For grouped or categorical bars, use the histograms style instead of manual binning:

set style data histograms
set style histogram clustered
plot 'file.dat' using 2:xtic(1)

using 2:xtic(1) reads the bar heights from column 2 and labels each bar with the text in column 1.

Saving to a file

set terminal pngcairo size 800,600
set output 'fig.png'
 
# switch to PDF instead
set terminal pdfcairo
set output 'fig.pdf'

After changing terminal and output, re-run plot (or replot) so the file actually gets written.

Multiple panels

set multiplot layout 2,2   # 2x2 grid, filled row-major
plot 'a.dat' using 1:2
plot 'b.dat' using 1:2
plot 'c.dat' using 1:2
plot 'd.dat' using 1:2
unset multiplot

Log-scale axes

set logscale x    # log x only
set logscale y
set logscale xy   # both