Added general unittest + Python unittests

[ColinKennedy]

Closes: #173

Added plenary (busted-style lua tests). Also added about as many Python unittests as I could think of for now. Note that these tests uncovered several bugs. Just searching -- TODO: through this PR's diff will point them out. The worst so far is the "inline comment" bug where if you add an inline comment on the same line as the function you're trying to document, the generated docstring accidentally positions itself above the function and heavily indented. But that can be for another GitHub issue.

Anyway, I think this will do the repo a lot of good. People can of course add docstrings for their language + docstring style of choice. I added Python + Google style since that's my daily driver. What do you think?

P.S. I was hoping to run unittests on Neovim version 0.9 (currently stable) but it was failing on neogen.generate. Did Neogen change to be a "requires nightly" repository?)

[danymat]

Hello @ColinKennedy, thanks a lot for your PR proposal for the testing suites. Indeed, that will help a lot finding some edgy bugs, and making sure that everything passes before merging (and does not break anything). I still need to review it, but as soon as i get the time, i take this into matter. Anyways, can you explain a bit more (as i don't get the time to review atm) what the textmate.lua does ?

[ColinKennedy]

Sure thing. textmate.lua is a file for helper functions specifically for unittesting. Take this test for example

    it("works with methods + nested functions", function()
        local source = [[
        # Reference: https://github.com/danymat/neogen/pull/151
        class Thing(object):
            def foo(self, bar, fizz, buzz):|cursor|
                def another(inner, function):|cursor|
                    def inner_most(more, stuff):|cursor|
                        pass
        ]]

        local expected = [[
        # Reference: https://github.com/danymat/neogen/pull/151
        class Thing(object):
            def foo(self, bar, fizz, buzz):
                """[TODO:description]

                Args:
                    bar ([TODO:parameter]): [TODO:description]
                    fizz ([TODO:parameter]): [TODO:description]
                    buzz ([TODO:parameter]): [TODO:description]
                """
                def another(inner, function):
                    """[TODO:description]

                    Args:
                        inner ([TODO:parameter]): [TODO:description]
                        function ([TODO:parameter]): [TODO:description]
                    """
                    def inner_most(more, stuff):
                        """[TODO:description]

                        Args:
                            more ([TODO:parameter]): [TODO:description]
                            stuff ([TODO:parameter]): [TODO:description]
                        """
                        pass
        ]]

        local result = _make_python_docstring(source)

        assert.equal(expected, result)
    end)

As a person writing unittests, I want to be able to

• write the source code (almost) exactly as is
• describe to neogen exactly where in the source code the user's cursor should be before running :Neogen
• expand for multiple functions / classes at at time if needed

In the example above, |cursor| marks the user's cursor position in the source file. textmate scans the source, strips out each |cursor| marker, and gives you back the row + column instead. Of course people could always have written a real source code blob of text and manually provided each row + column for each function but textmate made it really easy. The unittests here now were created in about 10 minutes after textmate.extract_cursors was added. It's not critical for this PR but IMO it's a nice option.