Include end column / range in message output

[cdce8p]

Current problem

At the moment, pylint only include a column offset in the output. That works find for command line or CI uses but doesn't work too well in IDEs. They depend on us to output an error range so they can effectively highlight the issue.
microsoft/vscode-python#18027

Desired solution

We already have node information for most / all (?) errors. Those include end column information. It should be fairly strait forward to add a new formatting option to the pylint output.

A Todo list:

• Modify output formatter to allow outputting end column information -> Update reporters to (allow) use of end_line and end_column #5372
• Update documentation -> Update reporters to (allow) use of end_line and end_column #5372
• Test changes -> Update functional test expected output #5349 & Check node location attributes in unittests #5383
• Update test runner to check end column info -> Update functional test expected output #5349 & Check node location attributes in unittests #5383
• Update existing tests -> Update functional test expected output #5349 & Check node location attributes in unittests #5383
• Decide what to output if we don't have an end column. -> end of line, only one char, repeat column offset? -> Update reporters to (allow) use of end_line and end_column #5372

Additional context

Maybe we can release a beta version first. With that IDEs and extensions could be updated and then it's up to us.
For that first version we wouldn't need to update the existing tests. It should be enough to enable the output.

--
Side note: IMO, this should be completely optional. It doesn't make much sense to update the default output.

[cdce8p]

/CC: @Pierre-Sassoulas @DanielNoord This has come up recently with the last vscode update. I believe it would be worth prioritizing it and maybe even to include the beta version in 2.12.0. If we release it this weekend, there might be enough time to get it into vscode-python just in time for the last release of the year.

[DanielNoord]

Can we use end_col_offset and end_lineno?

I'm not sure what vscode (and other IDEs) need exactly and how they handle multi-line statements, but it seems ast might help us here.

[cdce8p]

Can we use end_col_offset and end_lineno?

That's what I was thinking

[Pierre-Sassoulas]

Decide what to output if we don't have an end column.

I would say only one char as it minimize visual disturbance when we don't know exactly ? There will be a lot of case where a full line would make more sense but it's hard to guess. We'd have to think about it for each message, right ?

[DanielNoord]

@cdce8p Are you working on this? If not I can start the process of adding end_lineno and end_col_offset to nodes.Statement in astroid. That should be the first step I think.

[cdce8p]

@DanielNoord I've started working on the astroid side. Hope to have a first draft ready tomorrow. Would you like to take a look at the pylint side? Expect end_lineno: int | None and end_col_offset: int | None as attributes on NodeNG.

[cdce8p]

Opened pylint-dev/astroid#1258. In theory that should already work for most nodes. Haven't tested it yet.

The change got a bit larger than I expected initially. I somehow though that end_lineno and end_col_offset had already been added to the astroid nodes. As it's larger, we might want to reconsider the timing. Does it make sense to move it to 2.13?