gh-142392: Update ExceptionGroups in docs to use a tuple of errors where possible

[dr-carlos]

This should skip news. Thanks!

• Issue: Update ExceptionGroups to use a tuple of errors where possible #142392

---

📚 Documentation preview 📚: https://cpython-previews--142393.org.readthedocs.build/

[picnixz]

Was it discussed in the PR and deemed a better way alternative? I personally prefer using [] because it's more natural to () on my keyboard but I wonder whether we really need to do mention this in the docs. It's fine to use a list right? I also think it has a nicer rendering as otherwise we have nested (.

[dr-carlos]

Was it discussed in the PR and deemed a better way alternative? I personally prefer using [] because it's more natural to () on my keyboard but I wonder whether we really need to do mention this in the docs. It's fine to use a list right? I also think it has a nicer rendering as otherwise we have nested (.

It certainly isn't vastly superior, just slightly more efficient. @iritkatriel asked me to open an issue to update the docs, so I opened an issue and PR for it since there aren't many changes.

When you run repr() on an ExceptionGroup, if the original sequence was a list, then we need to create a new list from the eg.exceptions tuple, which could be expensive for a large list. This step is unnecessary for tuples as we just use the repr for eg.exceptions straight-up. So it's slightly more efficient (though of minimal effect for small groups).

[picnixz]

Yes it could be expensive but exception handlers are usually rare paths, so performance usually aren't an issue here. Unless you have a list of millions of exception objects, it doesn't really matter IMO.

[iritkatriel]

I agree it's not a significant perf impact. My point is it's inconsistent to recommend in the docs to use tuples and then have examples using [].