[Biopython-dev] Upcoming release

[Jeffrey Chang]

Hi Brad,

Thanks for doing this!  I've started running pyunit_testing, and it looks
really nice.  A few things to do before we incorporate it:

- it should respect command line arguments and run only the tests
specified.  For example, If I do "python pyunit_test.py test_translate,"
it should only run the test_translate test.

- in compare_output:
        # normalize the newlines in the two lines
        expected_line = string.strip(expected_line)
        output_line = string.strip(output_line)

It looks like this is stripping the whitespace from the lines before
they're being compared.  This may cause problems, because it won't catch
whitespace related errors, such as formatting.  Instead, it should convert
the ending newlines into some canonical form, e.g. '\n'.


Are you going to have a chance to fix these *real soon*?  If so, then
perhaps we can squeeze it into this release.

Jeff





On Sun, 25 Feb 2001, Brad Chapman wrote:

> [I was working on a PyUnit framework for integrating the tests]
>
> Jeff:
> > I'm glad someone's looking seriously into this!  It sounds like
> > something for the next release, though...
>
> Okay, well I completely ignored your message and worked more on this
> :-). In the shower this morning I thought of some ways to fix the
> problems we've been having, using the PyUnit framework I posted
> yesterday.
>
> It seems like I've got the regression comparisons working now so I
> implemented a "replacement" for br_regrtest.py that uses PyUnit. The
> only downside of the comparisons now is that it reads the entire
> output into a string, and then does the comparison, but I can't ever
> imagine that an output would be so incredibly huge this would be a
> problem (otherwise the test should probably be split up!).
>
> I used the fancy pyunit GUI stuff, so now the tests run by default
> with a little Tk GUI (should be nicer on Windows, and especially
> nicer for Macs).
>
> This all works for me okay on both Unix and Windows.
>
> What do people think? Does anyone have time to look at this before
> next weeks release, or do you all want to put it off until after?
>
> BTW, I noticed some problems with the tests while doing this, which I
> can now attribute to actual problems:
>
> o test_NCBIWWW is failing right now, due to problems in comparing the
> output (and these are not due to newline problems). I looked in the
> logs to see what had changed, and it looks like Thomas checked in an
> output change, but there wasn't a corresponding change to the
> tests.
>
> o test_SubsMat -- this seems to be failing on windows due to the fact
> that Windows prints -0.00 and the output is 0.00. I guess this is a
> Windows/UNIX difference. It's probably not worth worrying about since
> -0.00 and 0.00 are the same thing (as far as I know :-).
>
> Brad
>
>