gnuplot-beta — Developers and beta testers of new gnuplot code

Re: files for plotting given by string function

[Lutz M. <ma...@uc...>]

On Friday 17 June 2005 13:14, Ethan Merritt wrote:
> On Friday 17 June 2005 12:48 pm, Lutz Maibaum wrote:
> > The problem with this is that it produces a plot with only a single item,
> > i.e., all the different files have the same (point|line)style. I was
> > hoping for a shortcut to automatically plot many files, each with a
> > different style.
>
> Huh?  No it doesn't.  It expands to
>       plot "a.1.dat", "a.2.dat", "a.3.dat", ...

My apologies, I wrote my email before actually trying it out. My bad. This 
might afterall be exactly what I'm looking for.

  Lutz

Re: files for plotting given by string function

[Juergen W. <wie...@fr...>]

Am Freitag, 17. Juni 2005 21:48 schrieb Lutz Maibaum:
> On Thursday 16 June 2005 00:23, J=FCrgen Wieferink wrote:
> > In this special case, backtic substition will already do the job:
> >
> >   plot `ls -1 a.*.dat | dirlist.pl`
>
> The problem with this is that it produces a plot with only a single item,
> i.e., all the different files have the same (point|line)style. I was hopi=
ng
> for a shortcut to automatically plot many files, each with a different
> style.

Have you actually tried?  I get each file with its own line- and
pointstyle.

It may be possible to get the "dirlist.pl" simpler than the
following, though...

  #!/usr/bin/perl -w
  my $save =3D "\"";
  while (<>) {
      print $save;
      s/\n/", "/;
      $save =3D $_;
  }
  $save =3D~ s/, \"$//;
  print $save;


Juergen

Re: files for plotting given by string function

[Ethan M. <merritt@u.washington.edu>]

On Friday 17 June 2005 12:48 pm, Lutz Maibaum wrote:
> On Thursday 16 June 2005 00:23, J=FCrgen Wieferink wrote:
> > On Wednesdy, 15 June 2005 23:36, Ethan Merritt wrote:
> > >   set macros
> > >   file_list =3D system("ls -1 a.*.dat | dirlist.pl")
> > >   plot @file_list
> >
> > In this special case, backtic substition will already do the job:
> >
> >   plot `ls -1 a.*.dat | dirlist.pl`
>=20
> The problem with this is that it produces a plot with only a single item,=
=20
> i.e., all the different files have the same (point|line)style. I was hopi=
ng=20
> for a shortcut to automatically plot many files, each with a different st=
yle.

Huh?  No it doesn't.  It expands to
      plot "a.1.dat", "a.2.dat", "a.3.dat", ...

So each plot gets a new line/point style.
I know it works, because I actually tested it using the short script
in my previous Email. An even simpler demo of the principle is

gnuplot> plot `demo.pl`

where=20

#~/usr/bin/perl
# demo.pl
  print " '1.dat', '2.dat', '3.dat' "




=2D-=20
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: files for plotting given by string function

[Lutz M. <ma...@uc...>]

On Thursday 16 June 2005 00:23, J=FCrgen Wieferink wrote:
> On Wednesdy, 15 June 2005 23:36, Ethan Merritt wrote:
> >   set macros
> >   file_list =3D system("ls -1 a.*.dat | dirlist.pl")
> >   plot @file_list
>
> In this special case, backtic substition will already do the job:
>
>   plot `ls -1 a.*.dat | dirlist.pl`

The problem with this is that it produces a plot with only a single item,=20
i.e., all the different files have the same (point|line)style. I was hoping=
=20
for a shortcut to automatically plot many files, each with a different styl=
e.

  Lutz

Re: linetype palette

[Petr M. <mi...@ph...>]

> > release, about a year ago, there was a general discussion of which
> > patches to include.  The ones you see on SourceForge dated prior
> > to that point were all discussed and rejected for inclusion.
>
> Yes, that was what I meant.  I'm sorry I judged the process little
> hastily, but one certainly can get an impression that old patches are
> just abandoned there.

There is a discussion for each patch. You may have read that some patches
have final questions like "is somebody using this patch"?

> I didn't realise that they are left there for a purpose.  However, I'd
> still say some cleaning could take place there.

Then, contribute your impression to the discussion for each particular
patch.

> > you would like to see included, by all means feel free to advocate
> How do you do that?  Or do you mean by discussing the patches on this
> list?

If you strongly want to support applying or closing a particular patch,
write about it here, and it will get an immediate attention.

---
PM

Re: linetype palette

[Aapo L. <aap...@gm...>]

On Thu, 2005-06-16 at 10:44 -0700, Ethan Merritt wrote:
> Or perhaps you are referring to the fact that there are some very
> old patchsets still marked "Open".  At the time of the version 4
> release, about a year ago, there was a general discussion of which
> patches to include.  The ones you see on SourceForge dated prior
> to that point were all discussed and rejected for inclusion.

Yes, that was what I meant.  I'm sorry I judged the process little
hastily, but one certainly can get an impression that old patches are
just abandoned there.
I didn't realise that they are left there for a purpose.  However, I'd
still say some cleaning could take place there.

BTW, don't take my criticism too heavily.  I think you are doing a great
job; especially taking into account that Gnuplot is free.  :-)


> you would like to see included, by all means feel free to advocate

How do you do that?  Or do you mean by discussing the patches on this
list?

						Aapo

-- 
Aapo Lankinen <aap...@gm...>

Re: linetype palette

[Ethan M. <merritt@u.washington.edu>]

On Thursday 16 June 2005 03:42 am, Aapo Lankinen wrote:

> the "patch" section in Sourceforge seems not  to work very well.

How so?

I see 16 patchsets that have been applied or incorporated into
other work just since 1 Jan 2005. That's about 2 a month, and 
doesn't include things incorporated via direct Email to the
developers or to the mailing list. There are another dozen or so
patchsets on SourceForge that are still in active development
and will probably go into cvs over the next few months.

If you feel this rate of progress is too slow, you might consider
contributing some of your own time towards the process.

Or perhaps you are referring to the fact that there are some very
old patchsets still marked "Open".  At the time of the version 4
release, about a year ago, there was a general discussion of which
patches to include.  The ones you see on SourceForge dated prior
to that point were all discussed and rejected for inclusion.
Some contain good ideas, though, so it seems useful to leave them
as inspiration for other people to pick up the good parts and
develop them further.  If there is something in particular that
you would like to see included, by all means feel free to advocate
for it, submit an improved patch, or upload a better one.

-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: Plotting in real time

[Matt R. <tra...@ho...>]

Sig Lange <sig.lange <at> gmail.com> writes:

> 
> I would like to plot data in real time. I am not sure if this is
> possible but I would like to have time on my X axis, and various
> values on Y. I know this could be possible by tailing the last "20
> entries" of a file and then feeding that into gnuplot, but was
> wondering if there was a better way.
> 

> Has anyone any other alternative than feeding gnuplot a "plot
> 'myfile.txt' with lines" every update interval?
> 
> I tried feeding data into gnuplot via a fifo but graphing didn't start
> until the fifo was closed on the  writing end. Is there a way around
> that?
> 
> I understand that gnuplot would have to shift the graph or redraw the
> entire thing, this may not have been considered when designing.
> 
> TIA,
> Sig

Hey Sig-

    I have been trying to do the same thing with plotting in real time using
Gnuplot and was wondering if you have gotten anywhere with this?  I have seen a
program that does it, kinda, but i have not been able to completely disect how
it does it as of yet.  If you are still trying to figure it out as well, let me
know and if i get anywhere with this other program, ill let ya know

   -Matt

Re: linetype palette

[Aapo L. <aap...@gm...>]

On Thu, 2005-06-16 at 11:32 +0200, Johannes Zellner wrote:
> The problem was a small bug in my depth patch, see the patches section
> at sourceforge for the actual patch.

Checked your patch.  I'd say it is a cool addition that makes the
scansforward, scansbackward and scansautomatic options unnecessary.  I
think the functionality it provides is very important; especially
because it seems to be impossible to draw e.g. the two toruses properly
with the current CVS version.  And I think the patch is also reasonably
fast, I didn't have any problems with the interactive mouse mode either.
Thanks for the patch; I added it to my own Gnuplot.  :-)

Few thing I noticed; even though your coding is quite "clean", I'd
personally like some more comments in your code.  Especially the various
"pm3d_depth_queue" functions would require some commentation for easier
understanding.  Furthermore, I didn't understand following oddity, at
least for portability I suppose things like

quadrangles = (quadrangle*)0;  should be  quadrangles = NULL;

or am I missing something?


> Btw.: how should I proceed to get the depth patch into gnuplot?

Unfortunately I don't know.  I suppose you must bug one of the devs to
add the patch into the CVS; the "patch" section in Sourceforge seems not
to work very well.

Best Regards,
Aapo Lankinen


PS. I know it's not your fault, but the newest version of your patch
didn't go cleanly against the current CVS.  There was a problem with
indentation changes or something like that in the CVS...

-- 
Aapo Lankinen <aap...@gm...>

Re: linetype palette

[Johannes Z. <joh...@ze...>]

On Tue, Jun 14, 2005 at 07:22:16PM +0300, Aapo Lankinen wrote:
> On Tue, 2005-06-14 at 17:53 +0200, Johannes Zellner wrote:
> > something bad happened with "linetype palette", see the attached files.
> > The good one was generated with gnuplot 4.0 and something like
> > "plot 'tmp.dat' u 1:2:3 w p lt pal". The bad one was generated by the
> > latest gnuplot from cvs with the same command. Apparently "linetype
> > palette" doesn't work any more :-(
> 
> Hmm, I cannot reproduce the problem, "linetype palette" works just fine.
> BTW, what do you mean by 'plot' command?  You should use 'splot', at
> least my Gnuplot says
> 
> ---- clip ----
> gnuplot> plot 'tmp.dat' u 1:2:3 w p lt pal
>          2D plots cannot color by Z value; please use splot instead
> ---- clap ----
> 
> Why doesn't your Gnuplot say the same?  What kind of output does
> "splot sin(0.01*x*y) w p lt pal" give to you?


well, actually I /meant/ splot.
The problem was a small bug in my depth patch, see the patches section
at sourceforge for the actual patch.

Btw.: how should I proceed to get the depth patch into gnuplot?

-- 
Johannes

Re: files for plotting given by string function

[<wie...@we...>]

On Wednesdy, 15 June 2005 23:36, Ethan Merritt wrote:
> With the bleeding-edge cvs patch from Juergen Wieferink to implement a
> system() call, you could use a similar perlscript to do something along
> the lines of:
>
>   set macros
>   file_list = system("ls -1 a.*.dat | dirlist.pl")
>   plot @file_list

In this special case, backtic substition will already do the job:

  plot `ls -1 a.*.dat | dirlist.pl`

But feel free to try the patch and report bugs or suggest
enhancements. :-)


Juergen

Re: files for plotting given by string function

[Petr M. <mi...@ph...>]

> > BTW, in the recent addition: I would prefer "substr" function name (like in
> > awk) instead of the longer "substring".
>
> But I think awk uses substr(string, beg, length) rather than
> substr(string, beg, end). Is that a problem?

I see, then I propose it to use the length, not end. (Further, awk starts
counting from 1...?)

---
PM

Re: wxwidgets terminal

[Daniel J S. <dan...@ie...>]

Ethan Merritt wrote:
> On Tuesday 14 June 2005 10:46 pm, Daniel J Sebald wrote:
> 
>>I've updated things on my system.  The demo works as originally designed except 
>>for one problem.  The buttons and list items need to be pressed several times to 
>>cause the actions of starting a new demo or clearing a demo.  It seems to be a 
>>random number of times, but usually two or three times.
> 
> 
> Hmm. Does this mechanism only work from tcl?
> I had forgotten about tcl, and just tried speficying X-window ids for windows
> already open on my desktop, including the gnuplot command line window.
> It didn't work.

It should work once I put the patch on SourceForge.  (I was going to wait until 
I had tcl working.)  Many hunks were rejected and needed fixing.

Should I put an intermediate patch there for you to test?

Dan

Re: files for plotting given by string function

[Ethan M. <merritt@u.washington.edu>]

On Wednesday 15 June 2005 06:41 am, Petr Mikulik wrote:
> 
> BTW, in the recent addition: I would prefer "substr" function name (like in
> awk) instead of the longer "substring".

OK by me.  I'll change it.

But I think awk uses substr(string, beg, length) rather than
substr(string, beg, end). Is that a problem?

-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: wxwidgets terminal

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 14 June 2005 10:46 pm, Daniel J Sebald wrote:
> 
> I've updated things on my system.  The demo works as originally designed except 
> for one problem.  The buttons and list items need to be pressed several times to 
> cause the actions of starting a new demo or clearing a demo.  It seems to be a 
> random number of times, but usually two or three times.

Hmm. Does this mechanism only work from tcl?
I had forgotten about tcl, and just tried speficying X-window ids for windows
already open on my desktop, including the gnuplot command line window.
It didn't work.


-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: files for plotting given by string function

[Ethan M. <merritt@u.washington.edu>]

On Wednesday 15 June 2005 09:43 am, Lutz Maibaum wrote:
> 
> I was wondering if it would be feasible to include shell-like wildcard 
> expansion into gnuplot. For example
>   plot "a.*.dat"
> 
> Is it possible to do this in a platform-independent way? I currently use the 
> little script "pl" (http://www.phys.uni-paderborn.de/~stern/pl/) to use the 
> globbing features of the shell, but there are quite a few things that would 
> be much easier to do in an interactive gnuplot session.

With the bleeding-edge cvs patch from Juergen Wieferink to implement a
system() call, you could use a similar perlscript to do something along
the lines of:

  set macros
  file_list = system("ls -1 a.*.dat | dirlist.pl")
  plot @file_list


where dirlist.pl is a cleaned up version of this quick perl hack
(which leaves a missing " at the start of the line and an extra ," at the end,
but you get the idea)

#!/usr/bin/perl -w
  while (<>) {
	s/\n/", "/;
	print;
  }


In fact, if you use the equally bleeding-edge substring manipulation code
you can do the cleanup inside gnuplot. The following actually works
when run in the source directory using yesterday's cvs version with
Juergen's patch:

  set macros
  file_list = system("ls -1 ../demo/?.dat | ./dirlist.pl")
  file_list = "\"".file_list
  file_list = file_list[0:strlen(file_list)-4]

  plot @file_list


-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: files for plotting given by string function

[Lutz M. <ma...@uc...>]

On Wednesday 15 June 2005 05:51, Petr Mikulik wrote:
> f(n)='a'.n.'.dat'
> plot f(1), f(2)
>
> => it should be equivalent to plot 'a1.dat', 'a2.dat'. However, gnuplot

I was wondering if it would be feasible to include shell-like wildcard 
expansion into gnuplot. For example, if there a two files a.1.dat and a.2.dat 
in the current working directory,
  plot "a.*.dat"
should be interpreted as
  plot "a.1.dat", "a.2.dat"

Is it possible to do this in a platform-independent way? I currently use the 
little script "pl" (http://www.phys.uni-paderborn.de/~stern/pl/) to use the 
globbing features of the shell, but there are quite a few things that would 
be much easier to do in an interactive gnuplot session.

  Lutz

Re: files for plotting given by string function

[Ethan M. <merritt@u.washington.edu>]

On Wednesday 15 June 2005 06:41 am, Petr Mikulik wrote:
>>> f(n)='a'.n.'.dat'
>>> plot f(1), f(2)
>>> NB: you cannot plot a string-valued function

> But it is clear whether the function returns string or scalar. 

Well no, it is not clear. That is the heart of the problem.
There is nothing about a user-defined function that indicates
what type of value it returns. The only way to tell is to try
to evaluate it, and at the point the code must choose between
"plot function" and "plot data" it is not in general possible
to evaluate the function at all. 

Part of the problem is that if the function is really a function
for plotting then it has dummy variables. But at the time the plot
command is parsed,  dummy variables have not been assigned values
and the evaluation returns an error.

Another part of the problem is that gnuplot's syntax allows function
definitions like:
   f(x) = x ? x : "NULL"
which *have* no intrinsic return type.

> If string, then it is filename for (s)plot, otherwise scalar function its
> values to be drawn. Otherwise, would it make sense draw a curve/surface
> with strings, like  f(x,y)='point '.x.','.y; splot f(x,y) ? I don't
> think so.

I agree with that part, but I do not see a way to test what type of value
is returned by a function before actually plotting it, by which time it
is too late. If you can figure out how to make that test, then I support
the resulting plot behavior you propose.

-- 
Ethan A Merritt
Biomolecular Structure Center
University of Washington 98195-7742

Re: files for plotting given by string function

[Petr M. <mi...@ph...>]

> A workaround is the following:
>    plot ''.f(1), ''.f(2)

I wanted to simplify my life when plotting files a333.dat, a334.dat, ... not
to make it more difficult via ''.

> clear that a data file is being plotted, whereas `plot f(1)` could be a
> constant-valued function.

But it is clear whether the function returns string or scalar. If string,
then it is filename for (s)plot, otherwise scalar function its values to be
drawn. Otherwise, would it make sense draw a curve/surface with strings,
like  f(x,y)='point '.x.','.y; splot f(x,y) ? I don't think so.

BTW, in the recent addition: I would prefer "substr" function name (like in
awk) instead of the longer "substring".

---
PM

files for plotting given by string function

[Petr M. <mi...@ph...>]

I tried to generate filename via a string function, and plot them (sometimes
filenames are very long). For example:

f(n)='a'.n.'.dat'
plot f(1), f(2)

=> it should be equivalent to plot 'a1.dat', 'a2.dat'. However, gnuplot
complains:

	warning: encountered a string when expecting a number
	NB: you cannot plot a string-valued function

I wish it behaves like I expect, i.e. string means filename.

Ethan, could you change this behaviour?

---
PM

Re: Gnuplot logarithmic contours; "cntrparam level incremental" bug revisited

[Aapo L. <aap...@gm...>]

On Wed, 2005-06-15 at 14:38 +0300, Aapo Lankinen wrote:
> I wrote a very small patch so that '5,2,200' delivers tics at
> (5,10,20,40,80,160), as should be the case. 

Oops, forgot to attach the patch; sorry.	

-- 
Aapo Lankinen <aap...@gm...>

Gnuplot logarithmic contours; "cntrparam level incremental" bug revisited

[Aapo L. <aap...@gm...>]

Hi!

After discussing the cntrparm problem on this list, the following script
still doesn't work as expected with the newest Gnuplot CVS:

---- clip ----
unset surface
set contour
set log z
set palette model HSV functions gray, gray, 0.8-0.5*gray
set log cb
unset colorbox
set view map
set cntrparam level incremental 5,2,200
splot x**2+y**2+1.0 palette
---- clap ----

Nothing is plotted.  The problem is in the "cntrparam level incremental"
definition; Gnuplot recognizes the values as linear increments in
logarithmic axis, which leads in HUGE steps in the actual z-value.  The
steps go like (10^5, 10^7, 10^9, ... , 10^200), which is obviously very
wrong.

I wrote a very small patch so that '5,2,200' delivers tics at
(5,10,20,40,80,160), as should be the case. 

Best Regards,
Aapo Lankinen

-- 
Aapo Lankinen <aap...@gm...>

Re: wxwidgets terminal

[Daniel J S. <dan...@ie...>]

Ethan Merritt wrote:
> On Tuesday 14 June 2005 03:09 pm, mi...@ph... wrote:
> 
>>If it was under X11, have you used the new feature of an x11 handle from
>>sourceforge patches? This patch waits for confirmation that it works.

I've updated things on my system.  The demo works as originally designed except 
for one problem.  The buttons and list items need to be pressed several times to 
cause the actions of starting a new demo or clearing a demo.  It seems to be a 
random number of times, but usually two or three times.

Can anyone think of any changes in gnuplot that might cause problems with the 
following tcl lines?

     # Send an ctrl-C to gnuplot in case a different demo is running.
     exec kill -INT [pid $gpfd]
     set file [file join $dirName [.sel.f.list get @$x,$y]]

     # Some demos may have set terminal and failed to set it back.
     gnuplot "set term x11; reset; load \"$file\""

Or similarly, this line?

button .plt.buttons.bstop -text Reset -command {exec kill -INT [pid $gpfd]; 
gnuplot "reset; clear"}

The idea in both cases is to issue a CNTRL-C to the gnuplot process to break out 
of any currently running demo, then issue "reset", then proceed.

This used to work fine.  Could something have changed where the CNTRL-C doesn't 
occur until the "reset; clear" is partially through the pipe?

Dan

Re: wxwidgets terminal

[Daniel J S. <dan...@ie...>]

Ethan Merritt wrote:
> On Tuesday 14 June 2005 03:09 pm, mi...@ph... wrote:
> 
>>If it was under X11, have you used the new feature of an x11 handle from
>>sourceforge patches? This patch waits for confirmation that it works.
> 
> 
> I think that patchset may have broken due to subsequent changes to the
> cvs code.  I can't get it to work here any more.  Daniel?

I'll have a look at that one later this evening.  Should be fairly straightforward.

Dan

Re: wxwidgets terminal

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 14 June 2005 03:09 pm, mi...@ph... wrote:
> 
> If it was under X11, have you used the new feature of an x11 handle from
> sourceforge patches? This patch waits for confirmation that it works.

I think that patchset may have broken due to subsequent changes to the
cvs code.  I can't get it to work here any more.  Daniel?


-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195