[SciPy-dev] Add bz2 support to io.mmio [not tested] <--

Jarrod Millman millman@berkeley....
Thu Apr 10 12:45:36 CDT 2008


On Thu, Apr 10, 2008 at 8:41 AM, Pearu Peterson <pearu@cens.ioc.ee> wrote:
>  > I appreciate that you labelled the test deficiency, but what now?  Are
>  > you going to add the test, or was the message meant as a reminder to
>  > someone else?
>
>  Have you followed the thread
>
>   http://projects.scipy.org/pipermail/scipy-dev/2008-April/008804.html
>
>  ? The changeset is there because a user asked a feature. Adding
>  this feature was trivial, testing it is not so because it requires certain
>  input that I don't have at hand. There is a request for the user to check
>  if the feature works.
>
>  I don't see harm in such patches, it does not break anything,
>  and there is potential user who can confirm if the feature works or not.

I haven't talked with Stefan about this, but I believe that his point
(or at least this is my point) is not whether this code is fully
tested in the sense that the code is correct.  I think the issue is
that we are requiring unittests; that is, there should be a short
little unittest added here:
http://projects.scipy.org/scipy/scipy/browser/trunk/scipy/io/tests/test_mmio.py

I also agree that adding simple features such as this to satisfy user
requests is very good.  I spoke up about this because I think this a
really important general discussion and isn't exclusively related to
this specific check-in.  Obviously, adding features will often require
having user's test these features to ensure that they perform as
specified.  But it is essential that we have extensive unittests so
that when we edit and change code, there is some mechanism available
to determine whether old code is broken (hopefully, with a
semi-automated mechanism).

The solution to this problem shouldn't be that Stefan monitors add new
code to determine whether it comes with unittests or not.  He
shouldn't have to write the unittests himself and he shouldn't have to
create a ticket every time someone checks in un(unit)tested code.

Thanks,

-- 
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/


More information about the Scipy-dev mailing list