Pylab breaks locales with GTK

#1

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) :

···

##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49)
[GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.

import numpy as np
a=np.ma.array([1., 1.01, 2.11, 3.61])
for i in a:str(i)

...
'1.0'
'1.01'
'2.11'
'3.61'

import matplotlib
for i in a:str(i)

...
'1.0'
'1.01'
'2.11'
'3.61'

import pylab
for i in a:str(i)

...
'1.0'
'1,01'
'2,11'
'3,61'

np.__version__, matplotlib.__version__

('1.2.1', '0.98.3')
#############################

Any help welcomes!

--
Lionel Roubeyrie
chargé d'études
LIMAIR - La Surveillance de l'Air en Limousin
http://www.limair.asso.fr

#2

Lionel Roubeyrie wrote:

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) :
##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49)
[GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.

This has been reported before, and IIRC, it's a problem with PyGTK.

Ryan

···

--
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

#3

Hi Ryan,

Ryan May wrote:

Lionel Roubeyrie wrote:
  

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) :
##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49)
[GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
    
This has been reported before, and IIRC, it's a problem with PyGTK.
  
The fact that pygtk changes the local at import sounds buggy indeed, but
numpy should not be dependent on the locale anyway, so it also shows a
numpy bug I think:

http://scipy.org/scipy/numpy/ticket/884

Pauli and other have been working on those locale-related bugs in numpy,
but they are a bit fastidious to fix - hopefully, they will make their
way into numpy 1.3

David

#4

This seems to be a bug (or at least inconsistent behavior) in numpy when the locale is set (which happens when gtk is imported -- replace 'import pylab' with 'import gtk' and you'll see the same behavior).

We actually use a workaround for this in other parts of matplotlib, which is:

  for i in a: str(float(i))

By forcing the conversion to a Python float first, you can get around the localized number output.

I'll follow up with this on the numpy bug tracker or mailing list, since this has been a long-standing niggle of mine as well.

Cheers,
Mike

Lionel Roubeyrie wrote:

···

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) : ##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49) [GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
  

import numpy as np
a=np.ma.array([1., 1.01, 2.11, 3.61])
for i in a:str(i)
        

... '1.0'
'1.01'
'2.11'
'3.61'
  

import matplotlib
for i in a:str(i)
        

... '1.0'
'1.01'
'2.11'
'3.61'
  

import pylab
for i in a:str(i)
        

... '1.0'
'1,01'
'2,11'
'3,61'
  

np.__version__, matplotlib.__version__
        

('1.2.1', '0.98.3')
#############################

Any help welcomes!

--
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

#5

It's a bug with PyGTK in that merely importing it sets the locale.

But more seriously, it's also a bug in Numpy, in that its string formatting is dependent on locale (unlike standard floats in Python).

See this Numpy bug:

http://projects.scipy.org/scipy/numpy/ticket/902

···

*
*and this mailing list thread:

http://www.mail-archive.com/numpy-discussion@...177.../msg14563.html

Cheers,
Mike

Ryan May wrote:

Lionel Roubeyrie wrote:
  

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) : ##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49) [GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
    
This has been reported before, and IIRC, it's a problem with PyGTK.

Ryan

--
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

#6

Thanks for your responses,
Looking comments in the tickets, putting
import locale
locale.setlocale(locale.LC_NUMERIC, 'C')
after the pylab import resolves the problem (but not the bug...).

Have a happy new year

···

Le lundi 05 janvier 2009 à 15:37 +0100, Lionel Roubeyrie a écrit :

Hi all,
Trying to write to text files some plotted datas, we have a strange
behavour on masked arrays after importing pylab, with the dot decimal
separator replaced by a comma (but not all) :
##############################
Python 2.5.2 (r252:60911, Oct 5 2008, 19:24:49)
[GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy as np
>>> a=np.ma.array([1., 1.01, 2.11, 3.61])
>>> for i in a:str(i)
...
'1.0'
'1.01'
'2.11'
'3.61'
>>> import matplotlib
>>> for i in a:str(i)
...
'1.0'
'1.01'
'2.11'
'3.61'
>>> import pylab
>>> for i in a:str(i)
...
'1.0'
'1,01'
'2,11'
'3,61'
>>> np.__version__, matplotlib.__version__
('1.2.1', '0.98.3')
#############################

Any help welcomes!

--
Lionel Roubeyrie
chargé d'études
LIMAIR - La Surveillance de l'Air en Limousin
http://www.limair.asso.fr