Menu
▾
▴
Re: [Gnuplot-info] Managing long/complicated plot commands?
|
From: Ethan M. <eam...@gm...> - 2026-03-29 23:35:06
|
On Sunday, 29 March 2026 10:25:16 PDT Bjorn Buckwalter wrote:
> Hi Philipp and all,
>
> So, trying to minimize a bit more, I realized that a key piece of
> information that was omitted seems to be that the plot command is inside a
> function block. Below is a case that reproduces the error for me that you
> could test. I realised that if I take the command out of the function it
> works properly.
>
> By the way I am on macos and `gnuplot --version` says `gnuplot 6.0
> patchlevel 4`.
>
> Here is the case that fails with the error in question. Since it fails
> early (at least for me) you will not actually need the data files:
>
> ```
> local lon(x) = x
> local lat(x) = x
> local filter_lat(x) = x
> local next_color(x) = "#00FF00"
> local next_color_t(x, y) = "#00FF00"
>
> local lat1 = 0
>
> function $plot_polar(orbits) <<EOF
> plot \
> 'world_110m.txt' using (lon($1)):(filter_lat($2)) linecolor "gray"
> notitle , \
> lat(0) black dt 2 title "Equator" , \
> lat(66.6) black dt 5 title "Arctic/Antarctic circle", \
> lat(-66.6) black dt 5 notitle , \
> for [orbit in orbits] \
> 'data/Report'.orbit.'.txt' using (lon($7)):(filter_lat($6)) \
> lc rgb next_color_t(1, 'cc') notitle, \
> for [orbit in orbits] \
> 'data/Report'.orbit.'.txt' using (lon($7)):($3 == 2030 ?
> filter_lat($6) : NaN) \
> lc rgb next_color(n) title orbit, \
> 0, \
> 1
> EOF
>
> _ = $plot_polar("A")
> ```
>
> The error message:
>
> ```
> plot 'world_110m.txt' using (lon($1)):(filter_lat($2)) linecolor
> "gray" notitle , lat(0) black dt 2 title "Equator" , lat(66.6)
> black dt 5 title "Arctic/Antarctic circle", lat(-66.6) black dt 5
> notitle , for [orbit in orbits] 'data/Report'.orbit.'.txt' using
> (lon($7)):(filter_lat($6)) lc rgb next_color_t(1, 'cc') notitle,
> for [orbit in orbits] 'data/Report'.orbit.'.txt' using
> (lon($7)):($3 == 2030 ? filter_lat($6) : NaN) lc rgb next_color(n)
> title orbit, 0, \ 1
>
> ^
> "$plot_polar" line 13: invalid character \
> ```
>
> If I change the `linecolor` in the first line of `plot` to `lc`, The error
> does not occur.
This is a strange one.
I believe I have tracked it down to a program bug that has has nothing in
particular to do with the details of your script, although as you noted it is
more likely to be triggered by a function block than by input from a file.
You just got unlucky.
The program has no fixed limit on input line length or command length.
The size of the command buffer is expanded if necessary to hold the
next bit of input. The new input is appended to the current content of
the buffer, except that if the current buffer ends in a backslash it is
appended on top of the backslash rather than after it.
The bug is that the check for a trailing backslash gets skipped if the next
chunk of input is the one that causes the command buffer to be expanded,
so the command buffer is left with a spurious backslash character.
I don't have an immediately work-around for existing binaries other than
to note that once the command buffer has been expanded, the bug will
not trigger again. So even though the first invocation of the function
block fails, a subsequent invocation will succeed because this time the
buffer does not need to be expanded and the check for a backslash is
made correctly.
Or I suppose you could prefix the script with a useless bit of code that
constructs and executes a very long command so that the command
buffer gets expanded:
junk = "garbage garbage garbage "
do for [i=1:8] { junk = junk . junk }
command = 'print "' . junk . '" [1:8]'
print "reference a string of length ", strlen(junk)
eval(command)
...
start your real script here
- Ethan
> If you test, you may instead get:
>
> ```
> "$plot_polar" line 13: warning: Cannot find or open file "world_110m.txt"
> "$plot_polar" line 13: warning: Cannot find or open file "data/ReportA.txt"
> "$plot_polar" line 13: warning: Cannot find or open file "data/ReportA.txt"
> "$plot_polar" line 13: undefined variable: orbit
> ```
>
> By the way, the last message is rather curious. I have had trouble where
> this pops up sometimes and I haven't been able to nail down the exact
> circumstances. I believe in this case that it refers to `title orbit`. If I
> change that to `notitle` it disappears. And changing from `notitle` to
> `title orbit` in the preceding line does not trigger the message. This
> seems also somehow related to the length of the command.
>
> I guess Hernán's suggestion to generate a “flat” gnuplot script with a
> non-gnuplot script could allow me to avoid using function blocks. (At least
> that is how I interpreted the suggestion).
>
> Thanks,
> Bjorn
>
>
> On Sun, Mar 29, 2026 at 2:24 PM Philipp K. Janert <ph...@ja...> wrote:
>
> > On Sun, 29 Mar 2026 11:49:01 +0200
> > Bjorn Buckwalter <bjo...@gm...> wrote:
> >
> > > Thanks Jesper and Philpp. Jespers suggestion with variables can
> > > likely help – I will try it!
> >
> > It's not so easy for me to reproduce your
> > problem, because I don't have access to the
> > data files, etc.
> >
> > The cmd does not seem overly long. I still
> > suspect that there is a problem with one of
> > the CRs not being properly escaped - such
> > as having a blank after a backslash, but
> > before the actual CR. The error msg in your
> > original email points to that, too.
> >
> > I did not dig into the code to see whether
> > there is an actual limit to the length of
> > the cmd line. Ethan would know.
> >
> > Do you enter the cmd interactively, or do
> > you have it in a file, which you load?
> >
> > >
> > > FYI here is an example plot command that fails:
> > >
> > > ```
> > > plot \
> > > 'world_110m.txt' using (lon($1)):(filter_lat($2)) with lines
> > > linecolor "gray" notitle , \
> > > lat(0) with lines black dashtype 2 title "Equator" , \
> > > lat(66.6) with lines black dashtype 5 title "Arctic/Antarctic
> > > circle", \ lat(-66.6) with lines black dashtype 5 notitle , \
> > > for [orbit in orbits] \
> > > 'data/Report'.orbit.'.txt' using (lon($7)):(filter_lat($6)) \
> > > with lines linecolor rgb next_color_t(n, 'cc') notitle, \
> > > for [orbit in orbits] \
> > > 'data/Report'.orbit.'.txt' using (lon($7)):($3 == 2030 ?
> > > filter_lat($6) : NaN) \
> > > with lines linecolor rgb next_color(n) title orbit, \
> > > lat(lat1) black notitle # edge of plot
> > > ```
> > >
> > > There is A LOT going on there with various custom functions, etc. I
> > > tried to make a minimal and “dependency-free” example but failed to
> > > reproduce the behavior. I could maybe try harder.
> > >
> > > Here is in any case the error from gnuplot. The command has been
> > > concatenated at all baskslashes except the last:
> > >
> > > ```
> > > plot 'world_110m.txt' using (lon($1)):(filter_lat($2)) with
> > > lines linecolor "gray" notitle , lat(0) with lines black dashtype
> > > 2 title "Equator" , lat(66.6) with lines black dashtype 5 title
> > > "Arctic/Antarctic circle", lat(-66.6) with lines black dashtype 5
> > > notitle , for [orbit in orbits] 'data/Report'.orbit.'.txt'
> > > using (lon($7)):(filter_lat($6)) with lines linecolor rgb
> > > next_color_t(n, 'cc') notitle, for [orbit in orbits]
> > > 'data/Report'.orbit.'.txt' using (lon($7)):($3 == 2030 ?
> > > filter_lat($6) : NaN) with lines linecolor rgb next_color(n)
> > > title orbit, \ lat(lat1) black notitle # edge of plot
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > > ^
> > > "$plot_polar" line 15: invalid character \
> > > ```
> > >
> > > There is nothing wrong with the backslash (no trailing spaces, etc.).
> > >
> > > If I change it as so (just shortening style definitions) it works:
> > >
> > > ```
> > > set style data lines
> > > plot \
> > > 'world_110m.txt' using (lon($1)):(filter_lat($2)) lc "gray"
> > > notitle , \ lat(0) black dt 2 title "Equator" , \
> > > lat(66.6) black dt 5 title "Arctic/Antarctic circle", \
> > > lat(-66.6) black dt 5 notitle , \
> > > for [orbit in orbits] \
> > > 'data/Report'.orbit.'.txt' using (lon($7)):(filter_lat($6)) \
> > > lc rgb next_color_t(n, 'cc') notitle, \
> > > for [orbit in orbits] \
> > > 'data/Report'.orbit.'.txt' using (lon($7)):($3 == 2030 ?
> > > filter_lat($6) : NaN) \
> > > lc rgb next_color(n) title orbit, \
> > > lat(lat1) black notitle # edge of plot
> > > ```
> > >
> > > All I did was remove `with lines` (using `set style data lines`
> > > instead and replace `linecolor` and `dashtype` with `lc` and `dt`,
> > > respectively. This is why I say it seems to be somehow related to the
> > > length/number of characters of a command.
> > >
> > > Thanks,
> > > Bjorn
> > >
> > >
> > > On Sun, Mar 29, 2026 at 5:28 AM Philipp K. Janert <ph...@ja...>
> > > wrote:
> > >
> > > > On Sat, 28 Mar 2026 22:13:40 +0100
> > > > Bjorn Buckwalter <bjo...@gm...> wrote:
> > > >
> > > > > HI all,
> > > >
> > > > How long (in characters) are the command lines that
> > > > you are using? Mine frequently run long, and it has
> > > > never been a problem.
> > > >
> > > > Is it possible that you have problems with embedded
> > > > newlines, or that something goes awry if your command
> > > > line runs into a new line in your terminal?
> > > >
> > > > Best,
> > > >
> > > > Ph.
> > > >
> > > > >
> > > > > I rather often run into a problem of my plot commands being too
> > > > > long, typically resulting in a not-very-helpful error message
> > > > > such as:
> > > > >
> > > > > "$plot_polar" line 21: invalid character \
> > > > >
> > > > > It took a lot of head scratching to realize the root of the
> > > > > problem. I assume there are strong reasons for both the
> > > > > limitation on command length and the poor error message.
> > > > >
> > > > > I cannot be the only one running into this problem. The situation
> > > > > being what it is, what are the rest of you doing to manage when
> > > > > you run up against this limit? Replacing `linecolor` with `lc`,
> > > > > etc, will only get me so far and at the expense of readability.
> > > > >
> > > > > Many thanks,
> > > > > Bjorn
> > > > >
> > > > > _______________________________________________
> > > > > gnuplot-info mailing list
> > > > > gnu...@li...
> > > > > Membership management via:
> > > > > https://lists.sourceforge.net/lists/listinfo/gnuplot-info
> > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Philipp K. Janert
> > > > www.janert.me
> > > >
> > > >
> > > > _______________________________________________
> > > > gnuplot-info mailing list
> > > > gnu...@li...
> > > > Membership management via:
> > > > https://lists.sourceforge.net/lists/listinfo/gnuplot-info
> > > >
> >
> >
> >
> > --
> >
> > Philipp K. Janert
> > www.janert.me
> >
> >
> > _______________________________________________
> > gnuplot-info mailing list
> > gnu...@li...
> > Membership management via:
> > https://lists.sourceforge.net/lists/listinfo/gnuplot-info
> >
>
> _______________________________________________
> gnuplot-info mailing list
> gnu...@li...
> Membership management via: https://lists.sourceforge.net/lists/listinfo/gnuplot-info
>
|
