better csv modules and where have object-craft gone?

[David Fraser]

Skip Montanaro wrote:
>     Tim> I have been using object crafts csv module for quite a few
>     Tim> projects, mainly because I found the csv in python in it's current
>     Tim> incarnation is funtionally inferior to object crafts. The object
>     Tim> craft module for instance allowed you build up csv gradually (ie
>     Tim> field at a time rather the python csv module where the writer does
>     Tim> the work a record at a time) which isn't always the way I would
>     Tim> like to work, also I have always had encoding problems (specifcally
>     Tim> it doesn't support unicode as per the docs) everytime I used the
>     Tim> python module where as the object craft one always worked out of
>     Tim> the box.

> As for lack of Unicode support that's a known issue.  I suppose it hasn't
> been high enough on anyone's list of itches to have attracted any scratching
> yet.  Still, you might be able to get most of the way there with a subclass:
> 
>     class UnicodeWriter(csv.writer):
>         def __init__(self, *args, **kwds):
>             self.encoding = kwds.get('encoding', 'utf-8')
>             if 'encoding' in kwds: del kwds['encoding']
>             csv.writer.__init__(self, *args, **kwds)
> 
>         def writerow(self, row):
>             for (i,f) in enumerate(row):
>                 if isinstance(f, unicode):
>                     row[i] = f.encode(self.encoding)
> 
> I'm almost certain that reading data in multibyte encodings won't work
> though, as the low-level reader is byte-oriented instead of
> character-oriented.  Patches are welcome to resolve that deficiency.

Aha! so this is why my csv2po converter is getting strange encodings out...
What I'd really like to write/have is a pure Python implementation of 
the csv module. Presumably this would
1) Make Unicode trivial
2) Be slower
Anyhow it would be nice...

David