gnuplot-py-users Mailing List for Gnuplot.py

Hi,
thanks for the quick answers, I was thinking that it might be some 
float-precision problem. But, being a newbie and as gnuplot.py apparently 
doesn't have a documentation, I couldn't figure it out.

You're right, this type of  precision is probably only interesting for 
physicists and us chemists ;-) And I could probably do with less precision, I 
just haven't bothered to think about if rounding is always a good idea. But 
couldn't there be an auto-detection for the float type or sth? I mean, if 
someone's passing highly precise numbers, he'll probably want to plot them 
correctly...

Could you maybe point me to some information on how to use that workaround? 
I've tried reading the list values as 'np.float64' or converting the list 
via 'np.asarray(x).astype(np.float64)' but that doesn't seem to change 
anything...

Regards,
X.W.

Am Sunday 11 January 2009 17:47:07 schrieb Alan G Isaac:
> On 1/11/2009 11:34 AM Michael Haggerty apparently wrote:
>  > Not Python arrays, but numpy arrays.  The reason this is done is to save
>  > space for huge data sets.  My assumption was that it is unusual to plot
>  > values that vary only in the ninth decimal place.  And since there is a
>  > workaround, it doesn't seem like a bad compromise.
>
> How about letting dtype be an optional argument?
> The default can be np.float32, but the user could
> specify any dtype.
>
> Alan Isaac
>
>
> ---------------------------------------------------------------------------
>--- Check out the new SourceForge.net Marketplace.
> It is the best place to buy or sell services for
> just about anything Open Source.
> http://p.sf.net/sfu/Xq1LFB
> _______________________________________________
> Gnuplot-py-users mailing list
> Gnu...@li...
> https://lists.sourceforge.net/lists/listinfo/gnuplot-py-users



-- 
Xaver Wurzenberger
Guardinistr. 89
81375 München

Tel.: 089/32163661
Mobil: 0179/4478756

On 1/11/2009 11:34 AM Michael Haggerty apparently wrote:
 > Not Python arrays, but numpy arrays.  The reason this is done is to save
 > space for huge data sets.  My assumption was that it is unusual to plot
 > values that vary only in the ninth decimal place.  And since there is a
 > workaround, it doesn't seem like a bad compromise.


How about letting dtype be an optional argument?
The default can be np.float32, but the user could
specify any dtype.

Alan Isaac

Alan G Isaac wrote:
> On 1/11/2009 9:53 AM Michael Haggerty apparently wrote:
>> The Gnuplot.py library uses float (i.e., single-precision) arrays to
>> store values internally.  When your value is rounded to
>> single-precision, it results in the number that is actually plotted.
> 
> Are you saying it uses the Python array
> module to do this?  This seems a very
> odd decision.  Why not just change that
> to 'double'?

Not Python arrays, but numpy arrays.  The reason this is done is to save
space for huge data sets.  My assumption was that it is unusual to plot
values that vary only in the ninth decimal place.  And since there is a
workaround, it doesn't seem like a bad compromise.

But feel free to change it in your copy :-)

Michael

On 1/11/2009 9:53 AM Michael Haggerty apparently wrote:
> The Gnuplot.py library uses float (i.e., single-precision) arrays to
> store values internally.  When your value is rounded to
> single-precision, it results in the number that is actually plotted.

Are you saying it uses the Python array
module to do this?  This seems a very
odd decision.  Why not just change that
to 'double'?

Alan Isaac

Xaver Wurzenberger wrote:
> I am trying to plot large negative numbers, with 9 digits after the .,  via 
> python (version 2.5.2-3) and python-gnuplot, version 1.8-1 (both Debian Sid 
> packages), on terminal 'dumb', so ascii output.
> But it occurs to me that whenever I plot them from a (float or string) python 
> list (g.plot(Gnuplot.Data(list)), gnuplot is plotting a wrong number. 
> [...]
> 
> list_scf[0][1] is -902.904164954

The Gnuplot.py library uses float (i.e., single-precision) arrays to
store values internally.  When your value is rounded to
single-precision, it results in the number that is actually plotted.

IIRC you can get around this by passing a numpy double array to the
plot() method; I think that the library leaves the values as double
precision in this case.

Michael

Hi,
I'm a newbie to python programming, so I hope this is not a dumb mistake of 
mine ;-)

I am trying to plot large negative numbers, with 9 digits after the .,  via 
python (version 2.5.2-3) and python-gnuplot, version 1.8-1 (both Debian Sid 
packages), on terminal 'dumb', so ascii output.
But it occurs to me that whenever I plot them from a (float or string) python 
list (g.plot(Gnuplot.Data(list)), gnuplot is plotting a wrong number. 
Plotting a file (g.plot(Gnuplot.File('file.plt')) however works. Look at this 
example: I "zoomed in" to the first value here (the ascii formatting is lost 
due to the email, but I think you can see the problem; look at the asterisk 
psoition and y axis numbers!):

(12:55:03)xaver@siduxbox:~/Programme/python$./gw-minimal2.py test_real.log
39 cycles and -902.904164954  Hartree written to list of energies.
5 cycles and -902.912462667  Hartree written to list of energies.
6 cycles and -902.921650947  Hartree written to list of energies.
6 cycles and -902.927426487  Hartree written to list of energies.
6 cycles and -902.930537069  Hartree written to list of energies.
6 cycles and -902.933026922  Hartree written to list of energies.
21 cycles and -902.935555041  Hartree written to list of energies.
6 cycles and -902.937610935  Hartree written to list of energies.
6 cycles and -902.939101838  Hartree written to list of energies.
6 cycles and -902.94003734  Hartree written to list of energies.
6 cycles and -902.94119645  Hartree written to list of energies.
6 cycles and -902.942518852  Hartree written to list of energies.
15 cycles and -902.944932326  Hartree written to list of energies.
6 cycles and -902.947591728  Hartree written to list of energies.
6 cycles and -902.949241052  Hartree written to list of energies.
6 cycles and -902.950658896  Hartree written to list of energies.
6 cycles and -902.951400948  Hartree written to list of energies.
6 cycles and -902.952198923  Hartree written to list of energies.

list_scf is [[39, -902.90416495399995], [44, -902.912462667], 
[50, -902.92165094699999], [56, -902.92742648700005], 
[62, -902.93053706900002], [68, -902.93302692199995], 
[89, -902.93555504100004], [95, -902.93761093499995], 
[101, -902.93910183800006], [107, -902.94003734], [113, -902.94119645000001], 
[119, -902.94251885200003], [134, -902.94493232599996], 
[140, -902.94759172800002], [146, -902.94924105200005], 
[152, -902.95065889600005], [158, -902.95140094800001], 
[164, -902.95219892299997]]

list_scf[0][1] is -902.904164954

                                    Plotting float values:
  Energy/Hartree
   -902.904174810 ++-----+-------+------+-------+------+------+-------+-----++
                  +      +       +      +       +    SCF Optimization +.*..+ +
                  |      :                                                   |
                  |      :                                                   |
   -902.904174808 ++     :                                                  ++
                  |      :                                                   |
                  |      :                                                   |
   -902.904174806 ++     :                                                  ++
                  |      :                                                   |
                  |      *                                                   |
                  |                                                          |
   -902.904174804 ++                                                        ++
                  |                                                          |
                  |                                                          |
   -902.904174802 ++                                                        ++
                  |                                                          |
                  |                                                          |
                  +      +       +      +       +      +      +       +      +
   -902.904174800 ++-----+-------+------+-------+------+------+-------+-----++
                  20     40      60     80     100    120    140     160    
180
                                            Cycles


                                      Plotting plt file:
  Energy/Hartree
   -902.904164960 ++-----+-------+------+-------+------+------+-------+-----++
                  +      +       +      "test_real.log.plt" using 1:2 +.*..+ +
                  |      :                                                   |
                  |      :                                                   |
   -902.904164958 ++     :                                                  ++
                  |      :                                                   |
                  |      :                                                   |
   -902.904164956 ++     :                                                  ++
                  |      :                                                   |
                  |      :                                                   |
                  |      :                                                   |
   -902.904164954 ++     *                                                  ++
                  |                                                          |
                  |                                                          |
   -902.904164952 ++                                                        ++
                  |                                                          |
                  |                                                          |
                  +      +       +      +       +      +      +       +      +
   -902.904164950 ++-----+-------+------+-------+------+------+-------+-----++
                  20     40      60     80     100    120    140     160    
180

Does anyone know what's wrong here?
I can put all files on the net somewhere, if that helps.
Thanks in advance,
X. W.