[SciPy-dev] Matlab io bug; request for advice

[Stéfan van der Walt]

2009/2/19 Matthew Brett <matthew.brett at gmail.com>:
>> It may be undocumented, but I think it is safe to assume it will break a
>> lot of code. If it changes, I think it is better to raise a warning
>> before changing it. Because of the mismatch mentioned above, there may
>> not be such as thing as best solution; one choice shall be made, and
>> then we should stick to it.
>
> OK - I propose then that I do this:
>
> Leave the current behavior as it is (with the bug fixed) for 0.7.x
> Add a parameter at some point to the call allowing this behavior to be changed
> Raise a deprecation warning in 0.8 for the change in default behavior
> Change for 0.9

Personally, I would have changed this behaviour ASAP.  By putting it
off to 0.9 we are forcing users to produce more of these "broken"
MATLAB files.

With 0.7 just out the door, we could fix the bug in 0.7.1 with the
proper documentation (changelog and docstring).  If that is not
acceptable, at least propose the fix for 0.8, rather than 0.9.

Cheers
Stéfan