Add end_line and end_column to the output of the JSONReporter

[DanielNoord]

Current problem

These attributes were added in astroid 2.9.0 to most nodes and should be accessible from the output of the JSONReporter.

Desired solution

Basically all changes in this reverted commit should be committed to 3.0:

6e91225

Additional context

No response

[cdce8p]

I'm wondering if we could do this change earlier.

The last few days I've been working on the VSCode python extension to enable error ranges for pylint. The issue is that the --msg-template argument isn't (or wasn't until 2.12.0) backwards compatible. I.e. --msg-template={end_line} will crash on pylint < 2.12. Thus we can't use it version independently to also support ranges.

As I see it, the JSONReporter would be our best bet to implement it in a backwards compatible manner. With the existing text reporters, it's reasonable to assume that if user parse the output, any changes to the format will break it. In contrast, for JSON I would expect that most already use a json parser. So as long as we don't add any invalid values, adding two additional fields won't break this.

As the VSCode error range update depends on this (or an alternative), I would like to add a solution in time for 2.12.2.

What do you think @Pierre-Sassoulas and @DanielNoord? I'm also open to other ideas how to resolve this.

[DanielNoord]

I think it makes sense. The only issue I foresee now is when people use **output.items() on the output of the JSONReporter, but that is such an anti-pattern that I don't think we should really worry about supporting it.

I do wonder however how this relates to the breaking changes introduced with #3514 and #4741. I think it might make sense to see if we can prepare support for these changes in vscode now, or comment on those issue based on the experience you have with vscode about how to do that in the most non-breaking way.

[cdce8p]

I do wonder however how this relates to the breaking changes introduced with #3514 and #4741. I think it might make sense to see if we can prepare support for these changes in vscode now, or comment on those issue based on the experience you have with vscode about how to do that in the most non-breaking way.

Left a comment there: #4741 (comment). Basically I believe that it could be implemented without breaking existing use cases.