[SciPy-User] Loading a CSV with loadtxt

[Bruce Southey]

On Sat, Oct 30, 2010 at 8:06 PM, Esko Lehtonen <esko.lehtonen at iki.fi> wrote:
> 30.10.2010 0:19, Athanasios Anastasiou kirjoitti:
>> I have also tried with a simple one line file containing dummy entries like:
>> "Άλφα", "Βήτα", 1.0, 0.5, 1200
>>
>> It still finds that buffer object unreadable :-)
> It seems to be something with greek letters:
>
> I made two files, one containing
>
> "Άλφα", "Βήτα", 1.0, 0.5, 1200
>
> and another containing
>
> "abcd", "bcde", 1.0, 0.5, 1200
>
> three times. Both files were encoded in UTF-8.
>
> Only the file containing greek letters gave the error. Does not sound like correct behavior.
>
> My numpy version was the same, 1.3.0.
>
> - Esko
>
>

First this is a numpy issue that should be on the numpy list - a small
self-contained example will help.

About 6 months ago Pierre added Changeset 8416 that had a fix for
ticket 1473 (http://projects.scipy.org/numpy/ticket/1473). This should
be in numpy 1.5. If that does not help, please either add to that
ticket or create a new one.

Patches are welcome that allow numpy to read unicode file in Python
2.4+ (as I recognize that it is easier in Python 3).

Bruce