issue_comments

okay, thanks, we can take it from here.

Note that we have a CI to automatically run doctests, so you could in theory push to verify changes.

I have done an update to the docstring https://github.com/pydata/xarray/pull/6140

But note it is a minimal change rather than converting to doctest-style because I don't actually download or install xarray locally - I'm a bot that runs on AWS lambda to do static analysis checks on codebases for bugs. This repo was one of the 666 repos we run against when validating testing our new checkers.

As I do not have xarray setup locally I dont want to push larger code changes that have not been manually ran before I push, as there would be no confidence in the quality of the change :)

I updated the title of the issue to reflect the new state after #6139.

I will make a PR to remove the paren to make the intent clearer :+1:

Regarding black: worth also considering something like this to automatically highlight issues like this during PR :)

this does indeed look like a typo (even though as @mathause says that value will not be checked because the absence of data is checked first) but in any case it is valid to pass a str: dims="t", dims=["t"] and dims=("t",) mean the same thing to the DataArray constructor for 1D data. Maybe we should just remove the parens to make this clear? I would have hoped that black takes care of something like that...

We also have the same issue in the docstring of DataArray.from_dict, which by the way would really benefit from converting the code block to a doctest-style example (it will fail right now).

This tests checks if the appropriate error is raised if the passed dict misses the key "data". So it does IMHO not really matter what is passed as dims.