Add a feature to limit min and max phrase length + resolve issue #4

[LetianFeng]

This is a pull request to resolve the issue #4

The length limit should be applied before building the co-occurence matrix. Otherwise, we count meaningless phrases in the co-occurence matrix, and this could lead to a wrong ranked phrase list. Not only scores, order could also be influenced.

[LetianFeng]

Hi @csurfer could you please take a look at this pull request? Thx!

[csurfer]

Thank you for the pull request! 😄

[coveralls]

Coverage increased (+0.4%) to 89.773% when pulling 6eb828caa166f886a0939af01200885c06158d38 on LetianFeng:min_max_word_length into 0f6988223bed9fc1d91020f7691617edee36ca07 on csurfer:master.

[LetianFeng]

Hi @csurfer , I pushed a new commit according to your suggestions.

Since return [filter(lambda item: self.min_length <= len(item) <= self.max_length, phrases)] will return a list of filter object like [<filter object at 0x7f1990d12e10>], I used list(filter(lambda x: self.min_length <= len(x) <= self.max_length, phrases)) instead.

Personally, I don't have any preference between generator expression and lambda expression, but I feel a little bit weird when I put lambda expression in the second line:

        return list(filter(
            lambda x: self.min_length <= len(x) <= self.max_length, phrases))

Maybe you can read this disscussion list comprehension vs. lambda + filter if you have enough time.

Note: I know nothing about the README.rst, so please take a look at it before merging.

[LetianFeng]

Oops, already merged :sweat_smile: , then have a nice weekend and good luck with this project!

[csurfer]

I meant list(filter()) but mistyped it as [filtrt()]. I will go over the documentation and code and correct any minor changes I think are needed. No worries.