How to plot data without a separate file by specifying all points inside the Gnuplot script?

My program generates bash scripts that call gnuplot. I don't want to have to make an extra file to store the data; is there any way I can explicitly call all of the values? Or possibly having bash make a temporary file.

Something like

plot {(1,5),(2,10),(开发者_如何学Python3,1)}

is what I am looking for.

---

You can use the syntax for inline data - filename '-'.

The following example produces a simple plot in a GIF image (bash script):

gnuplot << EOF
set terminal gif
set output 'plot1.gif'
plot '-' using 1:2
        1 10
        2 20
        3 32
        4 40
        5 50
        e
EOF

---

Gnuplot 5.0.1 datablocks

main.gnuplot

$data << EOD
1 1
2 4
3 9
4 16
EOD

plot "$data" \
  with linespoints \
  title "my data"

Convert to PNG:

gnuplot -e 'set terminal png' -e 'set output "main.png"' main.gnuplot

Output:

This method is a bit more versatile than '-' as it makes it easier to reuse the same data multiple times, including on the same plot command: How to embed multiple datasets in a gnuplot command script for a single plot command?

Version 5 is available on Ubuntu 15.04, or compile from source with: https://askubuntu.com/a/684136/52975

You may also be interested in the + and ++ special file names when plotting with functions.

Tested on Ubuntu 18.10, gnuplot 5.2.

---

Example from using shell with pipeline,

gnuplot -p <(echo -e 'plot "-"\n1 1\ne')

---

bash One line ping graph with gnuplot

Sorry, it's not light (361 characters):

gnuplot -p -e "set xdata time;set timefmt '%s';set xrange [ '$(date +%s)' : '$(date -d 'now +30 seconds' +%s)' ];plot '-' using 1:2 with line title 'ping google';" < <(( ping -c 30 -n google.com| sed -u 's/^64.*time=\([0-9.]\+\) .*$/\1/p;d' | tee >(sed -u 's/.*/now/'| stdbuf -oL date -f - +d%s)) | sed -u 'N;s/\n/ /;s/\([0-9.]\+\) d\([0-9]\+\) */\2 \1/;s/d//')

Running this line will hold your terminal for 30 seconds, than plot on screen a graph presenting ping delay to google.com on last 30 seconds.

The same line could be splitted like this (workable too):

gnuplot -p -e "
    set xdata time;
    set timefmt '%s';
    set xrange [ '$(
        date +%s
      )' : '$(
        date -d 'now +30 seconds' +%s
      )' ];
    plot '-' using 1:2 with line title 'ping google';
    " < <((
    ping -c 30 -n google.com |
        sed -u 's/^64.*time=\([0-9.]\+\) .*$/\1/p;d' |
        tee >(sed -u 's/.*/now/'| stdbuf -oL date -f - +d%s)) |
    sed -u 'N;s/\n/ /;s/\([0-9.]\+\) d\([0-9]\+\) */\2 \1/;s/d//'
)

But this don't print values!

So i've added some few bytes:

gnuplot -p -e "set xdata time;set timefmt '%s';set xrange [ '$(date +%s)' : '$(date -d 'now +30 seconds' +%s)' ];plot '-' using 1:2 with line title 'ping google';" < <(( ping -c 30 -n google.com| sed -u 's/^64.*time=\([0-9.]\+\) .*$/\1/p;d' | tee >(sed -u 's/.*/now/'| stdbuf -oL date -f - +d%s)  ) | sed -u 'N;s/\n/ /;s/\([0-9.]\+\) d\([0-9]\+\) */\2 \1/;s/d//' | tee >(printf "%(%T)T %s\n" $(</dev/stdin) | column -c $COLUMNS >&2 ))

This may create a window like this:

And simultaneously print on terminal:

17:58:53 19.6   17:59:00 124    17:59:07 159    17:59:13 194    17:59:19 17.1
17:58:54 18.7   17:59:02 19.4   17:59:08 20.3   17:59:14 16.8   17:59:20 20.0
17:58:55 17.9   17:59:03 180    17:59:09 76.4   17:59:15 48.9   17:59:21 192
17:58:57 115    17:59:04 186    17:59:10 226    17:59:16 221    17:59:22 17.1
17:58:58 18.5   17:59:05 16.8   17:59:11 109    17:59:17 19.0
17:58:59 17.0   17:59:06 184    17:59:12 18.8   17:59:18 18.7

Re-written splitted:

gnuplot -p -e "
    set xdata time;
    set timefmt '%s';
    set xrange [ '$(date +%s)' : '$(date -d 'now +30 seconds' +%s)' ];
    plot '-' using 1:2 with line title 'ping google';" < <(
  (
      ping -c 30 -n google.com |
        sed -u 's/^64.*time=\([0-9.]\+\) .*$/\1/p;d' |
        tee >(sed -u 's/.*/now/'| stdbuf -oL date -f - +d%s)
  ) | sed -u 'N;s/\n/ /;s/\([0-9.]\+\) d\([0-9]\+\) */\2 \1/;s/d//' |
  tee >(printf "%(%T)T %s\n" $(</dev/stdin) |
  column -c $COLUMNS >&2 )
)

---

Solution is using arrays and parametric lines.

X="0.1 0.2 0.3 0.4 0.5"
Y="-1  0   2   4   8"

set parametric
set trange [1:words(X)]; set samples words(X)
plot (0+word(X,int(t))),(0+word(Y,int(t))) 

output

source: https://groups.google.com/d/msg/comp.graphics.apps.gnuplot/UdiiC2cBQNo/xEyj6i7Y910J

---

10 years later...

You can write something like this directly in gnuplot:

plot "< ./format.sh '{(1,5),(2,10),(3,1)}'"

where format.sh is a pretty simple script in a local file (remember to give it exec permissions):

#!/bin/bash
echo "$1" | sed 's:^..\(.*\)..$:\1:' | sed 's:),(:\n:g' | sed 's:,: :g'

How does it works?

The three sed commands do the following:

1. drops the first and last two characters, respectively {( and })
2. changes the sequence ),( to a newline
3. changes the remaining , to spaces

So:

$ ./format.sh '{(1,5),(2,10),(3,1)}'
1 5
2 10
3 1

The gnuplot plot "< command" syntax levers on popen to run command in a shell, capture the output and plot it on the fly. It can be pretty useful also in other circumstances.