Open tedim52 opened 12 months ago
I dug into this more over the weekend. The problem comes from https://github.com/kurtosis-tech/starlark-lsp , and particularly the way that it does docstring parsing. The issue is in part because the current parsing ALWAYS treats the first non-empty line as the description, regardless of its contents. Therefore, if I do this:
"""
Args:
some_arg: Some arg description
"""
then the parser will think my description is actually:
Args:
some_arg: Some arg description
with some_arg: Some arg description
being a continuation line of the "description" of Args:
.
This happens because the docstring is first passed through normalizeLines
and then passed through the args-extraction. The solution is to make the parser smarter, and intelligently identify when the first line is a description vs an Args
descriptor block.
I've pushed some non-functioning work to this branch, but it was a naive approach and it will need more work.
Context: https://kurtosistech.slack.com/archives/C05DZ4UABDZ/p1699571012055539