Re: [cclib-devel] GAMESS-US 2012

[Karol M. L. <kar...@gm...>]

On Nov 28 2012, Noel O'Boyle wrote:
> Sounds fine. We should never remove a test though, so I'd appreciate
> if you could move the old test to regressions.py with some sort of
> basic testing code (e.g. that it has the right number of etsecs).

What I did is call the appropriate unit test from a dynamically
generated function in the regression suite. This function is generated
as long as the logfile location is added to a list called 'old_tests'
inside the unit test class. It is clear from the code (I hope).

This makes archiving old logfiles as regressions easy and should scale
reasonably for doing such updates in the future. To withhold the unit
test for any reason for a particular regressed logfile (like in the
triplet TD case I brought up) just omit it from old_tests.

That being said, I will now update all GAMESS-US unit tests to the
newest 2012 version. Feel free to follow suit for other parsers, if you
like.

> Regarding parsing the version, I'm not too keen on parsing such
> metadata. That's going down a particular path we haven't gone down
> before. Maybe you can argue me around, but for sure we should not use
> version information during the parsing.

That's not what I had in mind. In any case, I don't wish or have the
free time to do this without a particular reason.

Cheers,
Karol

> On 28 November 2012 00:16, Karol M. Langner <kar...@gm...> wrote:
> > OK, I took another look at the GAMESS-US test files. It seems that
> > the files in basicGAMESS-US are from various different versions.
> > So, it seems more reasonable to just update the one output file
> > that gives more consistent results for dvb_td_triplet (that is, etsecs
> > sum up closer to 1), and possibly those that have changed substantially
> > in the new version of GAMESS-US and require parser work.
> >
> > That leaves the question, then, what to do with the old output files,
> > which are not from one set version. Rename with a version postfix and
> > transfer to regressions?
> >
> > Another option would be to just add additional logfiles for the outputs
> > that have changes, by adding _a or _b to the names like was done in the
> > case of several other parser.
> >
> > Let me know what you think,
> > Karol
> >
> > P.S. It also seems now a good idea to me to parse the version of a
> > program for information purposes, and it should be quite easy.
> >
> > On Nov 28 2012, Karol M. Langner wrote:
> >> Hi guys,
> >>
> >> I propose to update the GAMESS-US standard tests. The main motivation
> >> for this is that the dvb_td_triplet test in the 2012 version actually
> >> passes all the unit tests we have (the 2010 fails in one case). Also,
> >> there are some formatting changes in the output files, so some
> >> straightforward update to the parser is in order. I have all the output
> >> files ready.
> >>
> >> Let me know what you think. And, what do we do with the current output
> >> files in basicGAMESS-US? We should still make sure they are parsed
> >> correctly. Shall I move them to, say, basicGAMESS-US-2005 as we
> >> discussed some time ago, or rather to the regression suite?
> >>
> >> - Karol

-- 
written by Karol M. Langner
Fri Nov 30 00:00:09 CET 2012