Am 06.07.2010, 12:23 Uhr, schrieb Martijn Faassen <faassen@startifact.com>:
While I acknowledge its fragility, I do like the way doctests express exceptions quite a lot, so I'd prefer something that kept them like that. We've used normalization in the past, so if core Python exception messages changed I'd imagine more normalization could help again? Anyway, since we've only had a few test failures the problems with changing message formatting seems to be limited, right?
I think this has come up before especially in reference to Lennart's port to Python 3.x. Any test that depends upon exception formatting is likely to be a problem. In the specific zope.formlib test we have a doctest embedded within another: >>> MyForm(None, request)() # doctest: +NORMALIZE_WHITESPACE +ELLIPSIS There were errors: (u'Invalid floating point data', <exceptions.ValueError instance at ...>) I think this is bound to cause problems sooner or later in any environment. The test here is whether, and not how, the constraint is fulfilled and the nature of the exception is irrelevant. The more common
u'Invalid floating point data' in browser.contents
would be a better solution, I think. Furthermore, while it's great that form.txt actually runs I wasn't aware that it contained any additional tests that are not already run and I've always treated it as testable documentation not as an integral part of the formlib tests. But I'm ready to be believe this is a large misconception on my part. Charlie -- Charlie Clark Managing Director Clark Consulting & Research German Office Helmholtzstr. 20 Düsseldorf D- 40215 Tel: +49-211-600-3657 Mobile: +49-178-782-6226