From help-octave-request at bevo dot che dot wisc dot edu  Wed Nov 12 13:08:03 1997
Subject: RE: Temp files created by octave
From: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
To: efh at nessie dot mcc dot ac dot uk
Cc: "Juan I. Arribas" <nacho at pluton dot tel dot uva dot es>, help-octave@bevo.che.wisc.edu
Date: Wed, 12 Nov 1997 13:08:29 -0600

On 12-Nov-1997, Ted Harding <Ted dot Harding at nessie dot mcc dot ac dot uk> wrote:

| These are temporary files created by octave which hold data which will be
| sent to gnuplot. Keeping them there means that the "replot" command works:
| 
| *** though (John?) surely one doesn't need to keep them all for this
| *** purpose?

| If your "for" loop doesn't use "replot", and the plot is created anew each
| time round the loop, then it's probably safe to do "purge_tmp_files" each
| time (this shouldn't affect things you have "set", such a title, key,
| xlabel, ylabel, etc).

Octave can't delete them automatically because it has no way of
knowing when gnuplot is finished with them, even when there are no
replot commands involved.  The reason for this is that Octave
communicates with gnuplot through a pipe, and depending on the amount
of data and the complexity of the plot (simple x-y, 3d with hidden
lines, etc.) gnuplot may not be finished with the first set of data by
the time Octave executes the next plot command.  For example, on my
system,

  for i = 1:10; plot (rand (10,1)); purge_tmp_files; end

nearly always works and displays a `movie' of random lines, but

  for i = 1:10; plot (rand (1000,1)); purge_tmp_files; end

I typically see 2-3 failures.

I think gnuplot 3.6 allows data to be included directly with the plot
commands, so this problem can probably be avoided in future releases,
but I have not had time to fix it yet, and no one else has contributed
patches.

jwe