issue-261 Fixed missing parts of argument descriptions

[MichaelCG8]

issue-261 Fixed missing parts of argument descriptions for Google and Numpy style docstrings.

[dbieber]

Belated thanks for this PR!

If I'm understanding correctly:

There are two behavior changes within:

1. We now use arg_pattern = r'^[a-zA-Z_]\w*$' for determining valid argument names.
2. We now set state.current_arg = arg when we encounter a Google arg-and-type.

Change 1. doesn't seem so important -- it prevents the parser from treating some invalid arg names as args, but as I understand it this increased strictness hasn't come up in practice.

Change 2. is important; it fixes a bug where previously we were losing some information about multiline arg descriptions.

Can you confirm my understanding is correct?

Thanks.

Thanks also for the test cases 👍

[MichaelCG8]

With similar belatedness, thank you for reviewing!

Change 1
The old code would treat a line as an arg name if it is one word long and doesn't contain a :. The case that fails is any multiline description where the last line consists of one word only, as in foo3 #261 (comment)

This could have been fixed by also checking for a . in the string, so that any text ended with a period would be parsed as the plaintext description. The regex is more general and fixes the behaviour for any string that isn't a valid argument pattern, for example if the docstring ends with a reference to a type by surrounding it in back ticks.

The case where there is a single word followed by a period is something I've hit a few times if I reach my line length limit and wrap the last word onto a new line.

Change 2
Spot on.

[dbieber]

Closed with c39de6a