word-count: Add multiple apostrophes test case

[c4llmeco4ch]

This is a localized version of a change I discussed with the main problem-specs repo. They told me that, due to their lockdown, I should bring this to the language-specific repositories.

Effectively, the problem states that contractions ("can't", "won't", etc.) should treat apostrophes as part of the word, but rule 3 states:

Other than the apostrophe in a contraction all forms of punctuation are ignored
As such, I've added a very simple case testing that normal apostrophes are removed. The following code (while bad) should do a good job of showcasing a solution that passes all the live tests but fails the new one:

from string import punctuation
def count_words(sentence): 
    dict = {}
    punc = punctuation.replace("'",'')
    for val in punc:
        if val in sentence:
            sentence = sentence.replace(val, ' ')
    for word in sentence.split():
        if (w := word.lower())[0] == "'":
            w = w[1:]
        if w[-1] == "'":
            w = w[:-1]
        if w not in dict:
            dict[w] = 0
        dict[w] += 1
    return dict

Thanks!

[cmccandless]

Ref: exercism/problem-specifications#1628

[yawpitch]

Slight divergence, but what the heck should happen with a plural possessive? For instance, what should the word count be for the phrase "lawyers accept both lawyer's fees and lawyers' fees?"?

[yawpitch]

I'm okay with this test as is, though we are getting to the point where only a regex-based solution will work.

[cmccandless]

Please check out the docs on adding track specific tests and regenerating the test suite. CI will not pass if word_count_test.py is edited directly.

[c4llmeco4ch]

Sounds good, I'll get on that later today. I appreciate the help from everyone

[c4llmeco4ch]

Really sorry for possibly blowing up everyone's feed. After generating tests from the automated script, I'm still getting failures from Travis. Do I need to revert the test script back to the original passing version then push without the test added? Thanks and sorry again for the mess.

[c4llmeco4ch]

So I double checked and the reason I was pushing two versions of the test was due to my local version of problem-specs contained the test from my discussion with the main repo. I updated that but I'm still having the same failures from Travis.

[yawpitch]

Did you actually set the branch in your local copy of problem-specifications to master, rather than just remove the test in your working branch? The version number shouldn't be bumped in the generated tests, since the canonical-data.json it's pulling from hasn't (in master) been versioned.

[c4llmeco4ch]

I'll take a look Monday. That could be the issue.

[c4llmeco4ch]

Good call @yawpitch