GH-103319: Fix inspect.getsourcelines() to return 1-based line numbers

[artemmukhin]

The problem is described in GH-103225.

I did not find any other usages of a line number returned by inspect.getsourcelines() except for pdb. However, I am not sure how many external usages of inspect.getsourcelines() already rely on 0-based line numbers for modules.

• Issue: inspect.{getsourcelines, findsource} return 0 as the starting line number for modules #103319

[ghost]

All commit authors signed the Contributor License Agreement.

[artemmukhin]

@iritkatriel Could you please take a look, as this is a follow-up to #101674 you already reviewed?

Cc @gaogaotiantian

[gaogaotiantian]

1-based does seem more reasonable. But if we are fixing this, how about the code in inspect.findsource()? It's also 0-based.

[iritkatriel]

@iritkatriel Could you please take a look, as this is a follow-up to #101674 you already reviewed?

Cc @gaogaotiantian

The bug in pdb that you describe in the issue may be new, but you are changing behaviour of inspect in the PR. Question is whether this is a bug in inspect.

[gaogaotiantian]

The doc of inspect says this function Return a list of source lines and starting line number for an object. In my opinion, starting line number should not be 0. How is "starting from 0" and "starting from 1" different? They all include line 1 and there's no line 0. The fundamental of 0-based is that [0] means something and that's not the case in line number.

I would think this as a bug, but it's true that this is a breaking behavior. If we kept it as it is in inspect, I'd like to fix it in pdb.

[iritkatriel]

If there's a fix in pdb that you can do so we don't release a new bug there, then let's do it. Changing inspect probably would require more discussion.

FYI @Yhg1s .

[artemmukhin]

@iritkatriel Thank you for the quick response. That sounds reasonable. I can create a separate issue regarding inspect and then relink this PR, leaving #103225 unassigned for now. Is that OK?

@gaogaotiantian Thanks for pointing inspect.findsource() out. I will investigate it further and provide a summary in the new issue.

[artemmukhin]

@iritkatriel @gaogaotiantian I have opened a new issue regarding inspect #103319

[George-Ogden]

A solution may be to change the documentation, rather than the result, so that existing code does not break.