Original report at SourceForge, opened Fri Oct 22 11:00:07 2010

In matplotlib-1.0.0: when trying to import utf-8 csv files, csv2rec messes up some characters like german umlauts.

My (probably stupid) fix does this:

--- /Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/matplotlib/mlab.py~ 2010-07-06 16:43:35.000000000 +0200
+++ /Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/matplotlib/mlab.py 2010-10-21 11:33:19.000000000 +0200
@@ -2207,7 +2207,7 @@
mydateparser = with_default_value(dateparser, datetime.date(1,1,1))
myfloat = with_default_value(float, np.nan)
myint = with_default_value(int, -1)

• mystr = with_default_value(str, '')
• mystr = with_default_value(unicode, u'')
  mybool = with_default_value(mybool, None)

def mydate(x):

SourceForge Comments

On Fri Oct 29 14:22:06 2010, mdboom wrote:

Maybe I'm missing something... Does the CSV format have a standard way to specify an encoding? In absence of that, I'm not sure how we could safely do the right thing here -- we can only make a guess as to the encoding of the file. I'd prefer it if there were a way to specify one in the file itself, but in the absense of that, I think csv2rec should grow a kwarg to choose one.