gnuplot-py-users Mailing List for Gnuplot.py

Hi all,

here's the promised patch to make Gnuplot properly support mouse.  Please note 
that I only have access to linux boxes, so this hasn't been tested under any 
other platform.  I have, however, tried to make it portable by using only 
python code and no system commands.

The Gnuplot() constructor now has a new keyword option, 'mouse'.  If not 
given, it will try to auto-detect mouse support and enable it if present.  You 
may want to disable it explicitly for batch processing of very large datasets, 
because the mouse disables fifos, so it will be slower.

I hope this is useful to some, and after review Michael may choose to include it.

Regards,

f

Hi all,

I found out that the 'fix' I suggested yesterday for Unix wasn't quite enough 
to get the mouse properly working with Gnuplot.py 1.7.  While it works for 
plots of string-specified functions (like in gp.plot('x**2')), it fails when 
you supply arrays and try to zoom in.  The resulting error is something like:

In [2]: x=frange(0,2*pi,npts=200)

In [3]: plot x,sin(x)
------> plot(x,sin(x))

In [4]:
gnuplot> plot '/tmp/@14248.1' notitle
               ^
          can't read data file "/tmp/@14248.1"
          line 0: util.c: No such file or directory

where after input [3] I tried to zoom into the plot with the mouse.  I think 
the problem stems from the use of pipes, in fact I had some old notes (from 
2002) in ipython's changelog about this.

If I change line 87 of gp_unix.py from:

     prefer_fifo_data = 1

to
     prefer_fifo_data = 0

the above problem goes away.  It may be in fact possible to make the mouse and 
FIFOs happily coexist, but I've never been able to.

Now that Gnuplot 4.0 with mouse support is officially out, this may become a 
more common issue for users, so I figured it would be worth at least providing 
  people with a temporary solution.  Perhaps Michael can find a cleaner, 
permanent one, which maintains the advantages of FIFOs while keeping the mouse 
features active (which are essential for those of us used to them: once you 
get them, you don't want to go back :).

In summary, for users of Gnuplot.py v1.7: if you want full mouse support to 
work under python as it does in a regular gnuplot session, a hackish fix is to:

1. Comment out line 188 of _Gnuplot.py
2. Set prefer_fifo_data=0 in line 87 of gp_unix.py

Best regards,

F.

Michael Haggerty wrote:
> One quick-and-dirty fix would be to defer the initialization of the 
> terminal until the first plot (or splot) invocation.  But I suppose that 
> a cleaner solution should be possible.  However, I'm at a loss regarding 
> the best solution, since I've still never even used a version of gnuplot 
> with mouse support!

Since Gnuplot 4.0 is now officially out, and it has this feature, I suspect 
you are going to get more and more reports about this (it only used to be 
available to CVS users, which is a far smaller group).

I'll try to whip up a decent patch against 1.7 tonight and send it your way, 
hopefully covering all the bases.

Cheers,

f

Hello,

Fernando Perez wrote:

> Nadav had reported in
>
> http://sourceforge.net/mailarchive/forum.php?thread_id=3592388&forum_id=11272 
>
>
> having problems with mouse support with v1.7.  After following Michael's
> debugging suggestion, I found the culprit to be line 188 of _Gnuplot.py:
>
>         self('set terminal %s' % (gp.GnuplotOpts.default_term,)) 

[...]

> I fixed this for myself by just commenting out that line.  It wasn't 
> there in
> v1.5 which I've used for a long time without any problems, so I'm not 
> too worried.


Coincidentally, this line was added at the suggestion of another user, 
Chris Barker, who wrote:

 > [...] I found  (and fixed) a small bug in _Gnuplot.py. I was 
installing it on Mac OS-X,
 > using the Python2.2 that Apple delivered with OS-X 10.2. The terminal 
should be
 > set to "aqua" by default, but that was not happening. What I 
discovered is that the
 > Gnuplot object was not initializing the terminal to the default. The 
correction is:
 >
 > In _Gnuplot.py
 >
 > At the end of: Gnuplot.__init__ (line 196)
 >
 > add:
 >     self('set terminal %s' % gp.GnuplotOpts.default_term)
 >
 > This sets the terminal to the default, and all is well.

> Michael may want to do something a bit better, perhaps including an 
> option for
> 'set mouse' in the constructor, or mouse support auto-detection.  If 
> you want
> to go this route, let me know and I can send you the mouse support 
> code I have
> in IPython's Gnuplot routines.


One quick-and-dirty fix would be to defer the initialization of the 
terminal until the first plot (or splot) invocation.  But I suppose that 
a cleaner solution should be possible.  However, I'm at a loss regarding 
the best solution, since I've still never even used a version of gnuplot 
with mouse support!

Michael
the absentee landlord

-- 
Michael Haggerty
mh...@al...

Hi all,

Nadav had reported in

http://sourceforge.net/mailarchive/forum.php?thread_id=3592388&forum_id=11272

having problems with mouse support with v1.7.  After following Michael's
debugging suggestion, I found the culprit to be line 188 of _Gnuplot.py:

         self('set terminal %s' % (gp.GnuplotOpts.default_term,))

The problem is that now (v.1.6 and earlier did NOT do this), each Gnuplot
instance gets issued a 'set terminal FOO' command immediately upon starting.
Unfortunately, the 'set mouse' command (for reasons I don't understand), can
ONLY be issued as the very first command to gnuplot.  Otherwise it has no effect.

So in v1.7, by the time the Gnuplot() instance is returned, it's already
impossible to activate the mouse for it in any way.

I fixed this for myself by just commenting out that line.  It wasn't there in
v1.5 which I've used for a long time without any problems, so I'm not too worried.

Michael may want to do something a bit better, perhaps including an option for
'set mouse' in the constructor, or mouse support auto-detection.  If you want
to go this route, let me know and I can send you the mouse support code I have
in IPython's Gnuplot routines.

Regards,

f

Michael Haggerty wrote:

> Hi,
>
> The "matrix" option is currently not supported by Gnuplot.py.  If you 
> want to add it, start from the example of the "binary" option to 
> _FileItem.
>
> It is of course possible to read the file into a Numeric array and 
> then plot the array as a "Data()" PlotItem.
>
> Michael
>
> Eric Fernandez wrote:
>
>> I would like to plot 3d graph from data in a file, but reading data 
>> in "matrix" mode.
>> [...]
>> Unfortunately, it reads the file as 'binary', not 'matrix'. I did not 
>> find out how to pass the matrix option. Under gnuplot, it works with :
>>
>> splot('data_file') matrix
>>
>> is it possible to use it with gnuplot-py ?
>
>
Hi Michael

Thanks for your answer. I have added the option to the code, and it works !
I am actually working on MacOSX, so gp_macosx.py was also changed. 
Please find the three files I have changed. Sorry to not make a real 
patch for it, I prefer to send you the whole files, because I am not 
really confident with diff.
Could you please check if this is correct (it works on MacOSX and Linux) 
and add it to the CVS ? If there is another way to submit a patch, 
please tell me, I shall try to send it more formally.

Have a nice Easter week-end.
Eric Fernandez

Hi,

The "matrix" option is currently not supported by Gnuplot.py.  If you 
want to add it, start from the example of the "binary" option to _FileItem.

It is of course possible to read the file into a Numeric array and then 
plot the array as a "Data()" PlotItem.

Michael

Eric Fernandez wrote:

> I would like to plot 3d graph from data in a file, but reading data in 
> "matrix" mode.
> [...]
> Unfortunately, it reads the file as 'binary', not 'matrix'. I did not 
> find out how to pass the matrix option. Under gnuplot, it works with :
>
> splot('data_file') matrix
>
> is it possible to use it with gnuplot-py ?

-- 
Michael Haggerty
mh...@al...

Hello

I would like to plot 3d graph from data in a file, but reading data in 
"matrix" mode.

Here is the code I use :

g = Gnuplot.Gnuplot()
plot_volt = Gnuplot.PlotItems._FileItem('data_file')
g.splot(plot_volt)


Unfortunately, it reads the file as 'binary', not 'matrix'. I did not 
find out how to pass the matrix option. Under gnuplot, it works with :

splot('data_file') matrix

is it possible to use it with gnuplot-py ?

Thanks a lot for your help.
Eric