[SciPy-Dev] test output - io and stats

[josef.pktd at gmail.com]

On Sun, May 16, 2010 at 1:49 AM, Ralf Gommers
<ralf.gommers at googlemail.com> wrote:
>
>
> On Sat, May 15, 2010 at 7:12 PM, <josef.pktd at gmail.com> wrote:
>>
>> On Sat, May 15, 2010 at 12:04 AM, Ralf Gommers
>> <ralf.gommers at googlemail.com> wrote:
>> > Hi,
>> >
>> > In two weeks we'll create the 0.8 branch, and it would be good to have
>> > the
>> > test output cleaned up before then. It is very noisy, the main offenders
>> > are
>> > the io and stats modules. Output pasted below, most of it is
>> > deprecations
>> > and future warnings.
>> >
>> > Matthew and Josef, will you have time for some spring cleaning in the
>> > next
>> > few weeks?
>>
>> I won't have much time, however in your output, I only see
>> depreciation warnings for scipy stats and, I think, (almost) all of
>> them are intentional. So the only way to reduce this noise would be to
>> turn off depreciation warnings in the test run.
>>
> They can be filtered out. I think the test output for a release should be
> silent, and this thread seems to confirm that:
> http://thread.gmane.org/gmane.comp.python.numeric.general/21818/focus=21865
>

I think these are mainly other type of warnings, for example we
wouldn't want to have Zerodivision error turned on, because dividing
by zero is supposed to be handled correctly.

But Depreciation warnings and numerical precision warnings,
non-convergence of an optimization or integral are quite useful for
the user to know.

If a user upgrades and sees DepreciationWarnings while running the
tests, it's a good reminder to check their own code for them.

But, I don't really care either way, a clean test run looks nice.

Josef

>
>>
>> I think, if you run the "full" tests there are more warnings about
>> about numerical problems with integrate or similar.
>>
>  Yes, there's other modules besides io and stats that are generating some
> warnings, I will look at those.
>
> Cheers,
> Ralf
>
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-dev
>
>