s-singular fix

[jonmcoe]

Few notes:

1) Two tests failed for me both before, and after, my changes

This is probably something with my setup, but not sure what. I am using I am using python 2.7.10 and followed the instructions in https://github.com/sloria/TextBlob/blob/dev/CONTRIBUTING.rst with the only addition being the download_corpora. I don't think I experienced these failures when working on #96 , but I am getting them now even when I checkout far into the past (8904336 for example) so that's why I feel confident it is something in my setup. Any ideas?

======================================================================
FAIL: test_correct (tests.test_blob.SentenceTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/jon/TextBlob/tests/test_blob.py", line 217, in test_correct
    assert_equal(blob.correct(), tb.Sentence("I have bad spelling."))
AssertionError: Sentence("") != Sentence("I have bad spelling.")

======================================================================
FAIL: test_correct (tests.test_blob.TextBlobTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/jon/TextBlob/tests/test_blob.py", line 778, in test_correct
    assert_equal(blob.correct(), tb.TextBlob("I have bad spelling."))
AssertionError: TextBlob("") != TextBlob("I have bad spelling.")

----------------------------------------------------------------------
Ran 222 tests in 43.071s

FAILED (failures=2)

2) Why aren't plural_categories['uninflected']/['uncountable'] and singular_uninflected/uncountable the exact same lists?

I looked into the differences, the only difference is that there are two singular_uninflected not in its counterpart. I'm guessing this is an artifact of http://www.bermi.org/inflector/ vs http://www.csse.monash.edu.au/~damian/papers/HTML/Plurals.html

>>> set(plural_categories['uninflected']) - set(singular_uninflected)
set([])
>>> set(singular_uninflected) - set(plural_categories['uninflected']) 
set(['swiss', 'georgia'])
>>> set(plural_categories['uncountable']) - set(singular_uncountable)
set([])
>>> set(singular_uncountable) - set(plural_categories['uncountable'])
set([])

3) high-jinks pluralizes wrong

>>> pluralize('high-jinks')
'high-jinkss'

I'm guessing this is due to the hyphen, and something about the treatment of rules in the for loop around inflect.py:277. I didn't really dig yet, and it seemed a little bit heavy of an issue to just lump into this pull request anyway.

---

Anyway, love your library. Been happily using the classifiers and sentiment analysis for a little while now. Just found these few issues with pluralization/singularization and am happy to contribute.

[jonmcoe]

so those tests also failed on travis, and for me when I tried on master. eager to hear your thoughts

[jonmcoe]

i figured out the unittest failures. they are due to nltk version 3.1. runs fine when i downgrade 3.0.5.
https://github.com/nltk/nltk/blob/develop/ChangeLog

sorry to dump all of this here. can write up an issue if you'd like me to. don't know enough about about nltk and the spelling functions to fix it, but i could set it up to pin to nltk==3.0.5 if you think that's the best move

[sloria]

I'm -0 on adding things like this to project gitignores. This should really be in a global gitignore. See https://help.github.com/articles/ignoring-files/#create-a-global-gitignore

[sloria]

Thanks @jonmcoe for the patch!

I'll try to look more closely at the test failure within the next day or two. I would rather not pin to older versions of NLTK.

Re: your questions about inflection--TextBlob uses pattern's implementation from https://github.com/clips/pattern/blob/master/pattern/text/en/inflect.py with just a few modifications to support Python 3. I recommend directing your questions on that repo, and possibly sending this patch there if you're feeling generous. =).

Just a few changes to make above. Once those and the failing tests are addressed, this should be good to merge.

[jonmcoe]

sounds good! thanks for explaining

[sloria]

The failing tests have been addressed on dev. This is good to merge. Thanks again @jonmcoe!