gnuplot-py-users Mailing List for Gnuplot.py

Hi all

Anyone running python 2.3 ? To run the setup script,
I had to change one line in gp_java.py

I don't use Jython, so I can't test. But the installer
complains:

-----
byte-compiling /usr/lib/python2.3/site-packages/Gnuplot/gp_java.py to
gp_java.pyc
  File "/usr/lib/python2.3/site-packages/Gnuplot/gp_java.py", line 130
    self.process = Runtime.getRuntime().exec(command)
                                           ^
SyntaxError: invalid syntax
-----

Michael, should this be exec or eval() ? I am not sure, 
but it might need to be "eval(command)" according to
http://www.python.org/doc/current/lib/bltin-code-objects.html#l2h-259

Using eval byte-compiles ok, but who knows. Ideas ?

Best,
-- leo

Michael Haggerty wrote:
> Hi,
> 
> I just implemented some accumulated suggestions in the CVS version of 
> Gnuplot.py.
> 
> Fernando Perez wrote:
> 
> 
>>>Some quick comments regarding your specific ideas:
>>>
>>> > -plot(y) -> plot a NumPy array against its indices 
>>
> 
> Done.

Great!
>>> > -plot(x,y) -> plot y vs x
>>
> 
> I will not implement this (because of the ambiguity).  But since this 
> seems to be a popular feature, maybe I'll implement it with another name 
> (like 'qplot()', as suggested by Leo Milano).

Fair enough.  I agree that it's ambiguous.  I did it simply because _I_ use it 
a lot, and I know how to handle it.  But for a general public tool, it's much 
cleaner to have a different function.  However, qplot() should keep the same 
interface as plot() in regards to options, etc.

>>> > -plot(y1,None,y2) -> plot y1,y2 vs their indices (None is a 
>>>separator) 
>>
> 
> Ditto.

good.

>>> > -plot(y,filename='out.eps') -> make eps output reliably
>>> >  (there's a tricky timing issue which requires
>>> >  retrying the print command with a sleep()) 
>>
> 
> I still don't understand the timing issue.  But I see the need to be 
> able to plot directly to a "hardcopy" file (be it PS or EPS or PNG or 
> ...) without first plotting to a graphics window.  It would be very 
> nearly possible now, if Gnuplot._add_to_queue() were a documented part 
> of the interface.

Agreed on the auto-plotting to file without plotting window.  This was 
critical to me, but I only implemented eps output.

Re. the timing issue: honestly, I don't understand it either.  I only know I 
wasted an entire Saturday trying to find any pattern to it, and finally 
decided to brute-force a solution with the 20 tries approach.  Inelegant as 
hell, but so far it's done the job for me ;)

>>> > - Added the 'index' keyword arg to File, similar to Gnuplot's
>>> >   index option to select a dataset from a file with multiple
>>> >   datasets in it.
>>
> 
> Done.

Good.

As far as I can see, you've pretty much brought into the Gnuplot.py mainline 
almost everything which IPython's gnuplot support added.  This is great news, 
since I didn't want to fork your code, but simply offer functionality which I 
felt was needed for daily work.

Once you make a release (hopefully with hardcopy-without-window in, which I 
consider critical), perhaps we can comb over what's left and I could deprecate 
IPython's gnuplot stuff in favor of your mainline code.

Best regards,

Fernando.

Hi,

I just implemented some accumulated suggestions in the CVS version of 
Gnuplot.py.

Fernando Perez wrote:

>> Some quick comments regarding your specific ideas:
>>
>>  > -plot(y) -> plot a NumPy array against its indices 
>

Done.

>>  > -plot(x,y) -> plot y vs x
>

I will not implement this (because of the ambiguity).  But since this 
seems to be a popular feature, maybe I'll implement it with another name 
(like 'qplot()', as suggested by Leo Milano).

>>  > -plot(y1,None,y2) -> plot y1,y2 vs their indices (None is a 
>> separator) 
>

Ditto.

>>  > -plot(y,filename='out.eps') -> make eps output reliably
>>  >  (there's a tricky timing issue which requires
>>  >  retrying the print command with a sleep()) 
>

I still don't understand the timing issue.  But I see the need to be 
able to plot directly to a "hardcopy" file (be it PS or EPS or PNG or 
...) without first plotting to a graphics window.  It would be very 
nearly possible now, if Gnuplot._add_to_queue() were a documented part 
of the interface.

>>  > - Added the 'index' keyword arg to File, similar to Gnuplot's
>>  >   index option to select a dataset from a file with multiple
>>  >   datasets in it.
>

Done.

Michael

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

John,

I'm cc'ing the gnuplot-py-users mailing list...

[The original question was whether it is practical to use Gnuplot.py to 
generate graphics in a web server.]

John Draper wrote:

>>Hello,
>>
>>Sure, gnuplot + Gnuplot.py can do that.
>>
>>I would suggest you get the latest CVS version of Gnuplot.py; then you 
>>can do something like
>>
>>   g = Gnuplot.Gnuplot()
>>   ips = list of 4-tuples of ip octets from your database like 
>>[(192,168,1,1), (198,137,240,92),...]
>>   g.title('Spam sources')
>>   g.xlabel('First octet')
>>   g.ylabel('Second octet')
>>   g.plot(ips, cols=(0,1), with='dots')
>>   g.hardcopy('file.png', terminal='png', fontsize='large', color=1)
>>
>>This writes the output to a file called "file.png" in the current 
>>directory.  Then you just need to get your web server to insert the 
>>image into the web page and somehow arrange for old files to be cleaned 
>>up once in a while.
>>    
>>
>
>Wow - most excellent.
>  
>
>>Through some more magic it would be possible to create the output on the 
>>fly and pump it straight to the client, but that's a little bit 
>>trickier.  
>>    
>>
>
>Yes- I'm going to need to know how to do that.  So,  when you say "trickier" what do you mean by that.
>  
>

It might be possible (I'm not sure) to have gnuplot write its png output 
to stdout by using g("set term png ...") before generating the original 
plot.  I'm not sure if gnuplot outputs other stuff on standard out that 
would confuse the output.

Another alternative would be to create a named pipe and have gnuplot 
write its output to that "file".  Another thread or process could 
capture the output and pump it out of the web server.

>>It should also be possible to omit the initial output to the 
>>monitor but I haven't tried that before.
>>    
>>
>
>I have no "monitor" as there is no graphical device on the server at all.
>

Well then you would want to g("set term png [...]") before creating the 
initial plot.  It means you can't use the hardcopy() method (at least 
not without first making some minor changes) but that's not such a problem.

Michael

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