bpo-42877: add the 'compact' param to TracebackException's __init__

[iritkatriel]

…re is cause

https://bugs.python.org/issue42877

[iritkatriel]

I don't know whether a change like this needs news.

[gvanrossum]

Hm, I don't think you can make this change at all -- __cause__ and __context__ are documented and hence public APIs.

Maybe a different way to save space/effort is to set __context__ to None if there is a __cause__ -- but even that seems to violate the documented value for this attribute.

[iritkatriel]

I see. I didn't realise all these fields are in the doc.

This class is probably mostly used from print_exception etc, we could add this as an opt-in optimisation to use from there. Is that worth doing?

[gvanrossum]

This class is probably mostly used from print_exception etc, we could add this as an opt-in optimisation to use from there. Is that worth doing?

I don't know -- I guess you ran into this when solving the recursion error issue? I guess you could add a new keyword parameter skip_context=False which you can set to something like exc is not None and exc.__cause_ is not None_ at the call site? (Not sure if the exc is not None guard is needed.)

[iritkatriel]

I'd pass in a bool (like compact or save_space) and leave the logic of whether to skip internet he class itself (note that there is also __suppress_context__).

Yes, this came up re the recursion issue. I'm still not sure what to do about that.

[gvanrossum]

I'd pass in a bool (like compact or save_space) and leave the logic of whether to skip internet he class itself (note that there is also __suppress_context__).

Since the latter is set on the original exception we can't change it. Agreed to put as much logic in the constructor.

[gvanrossum]

(This would also require a news item.)

[iritkatriel]

Yes, and some new tests.

Since this won't simplify the code, it won't help deal with the recursion issue so it can come after PR 24158, which I've now refactored to use the fact that the chained exceptions are a list and not a tree.

[gvanrossum]

Looks like you have a merge conflict due to your own #24158.

[gvanrossum]

Maybe clarify that this suppresses context if cause is not None?

[iritkatriel]

Yes, though I'll avoid "suppresses" so as not to confuse with __suppress_context__ which is applied in format().

I also don't need to mention format_exception_only because it's called from format.

[gvanrossum]

Suggested change

	exc_value.__suppress_context__ if exc_value else False
	exc_value.__suppress_context__ if exc_value is not None else False

(Some joker could add a __bool__ method to their exception that sometimes makes it falsey.)

[gvanrossum]

Yup!

[bedevere-bot]

@gvanrossum: Please replace # with GH- in the commit message next time. Thanks!