[Numpy-discussion] npz load on numpy 3

[Pauli Virtanen]

Tue, 29 Mar 2011 08:27:52 +0000, Pauli Virtanen wrote:
> Tue, 29 Mar 2011 04:16:00 -0400, josef.pktd wrote:
>> Traceback (most recent call last):
>>   File "<stdin>", line 1, in <module>
>>   File "C:\Programs\Python32\lib\site-packages\numpy\lib\npyio.py",
>> line 222, in __getitem__
>>     return format.read_array(value)
>>   File "C:\Programs\Python32\lib\site-packages\numpy\lib\format.py",
>> line 449, in read_array
>>     array = pickle.load(fp)
>> UnicodeDecodeError: 'ascii' codec can't decode byte 0xf0 in position 6:
>> ordinal not in range(128)
> [clip]
>> Any ideas ?
> 
> There's some open() call that opens the file in text mode rather than in
> binary mode, I guess.

Ah, that's not it. The problem is that pickled Numpy arrays are not 
backward compatible between Python 2 and 3 because of the string vs. 
unicode change --- the pickle.load() call should specify an encoding eg. 
pickle.load(fp, encoding='latin1').

This needs to be wrapped in a try-expect block so that it tries to load 
it with encoding='latin1' if the first attempt fails.

-- 
Pauli Virtanen