Closed JakobSteixner closed 5 years ago
StopIteration
)@thalelinh test_web.TestSearchEngine.test_search_twitter
failing under 2.7
Which tests?
Which tests?
test_web.py, TestSearchEngine.test_search_twitter
-- L503 with 0 results (1 expected), maybe an API limit issue? Nothing either of us touched, I'm skipping it for now.
Which tests?
test_web.py,
TestSearchEngine.test_search_twitter
-- L503 with 0 results (1 expected), maybe an API limit issue? Nothing either of us touched, I'm skipping it for now.
Yeah, I made a retry loop on my PR (which focused on fixing the CI). It worked there, not the best solution though :sweat_smile:
My guess now is that there's a seed(0)
missing after _test_classifier instantiate a Classifier, during the test_slp
.
If you look at shuffled
definition, which is called without the seed param at SLP constructor, you will see that it calls seed(None)
, which is equivalent of seed(time.now)
, that we obviously don't want to happen in a test case. :smile:
The Twitter test error was actually just causality. It's a query using a given string, sometimes there aren't any results.
The Twitter test error was actually just causality. It's a query using a given string, sometimes there aren't any results.
I thought so. Now failing with a version conflict in pytest setup though...
Appears to pass now!
Fix implementing logic to correctly identify the (irregular) base of prefixed verbs. Old:
Explanation: since 'vorgehen' is not found in the lexicon, a default regular inflection strategy applies. Even though the separable prefix is correctly identified, the base form thus extracted isn't checked against the lexicon and the available information about its irregular inflection thus lost.
New:
This fix is achieved with a second pass to
lemma
after stripping the prefix, to identify the known irregular inflection of the base form 'gehen'.Further, blacklists of verbs that look like they might be prefix verbs or latinate verbs with the suffix 'ier(en)' have been included to block the parser's exceptional treatment of those.