bpo-40077: Convert itertools types to heap types

[erlend-aasland]

https://bugs.python.org/issue40077

[erlend-aasland]

@rhettinger, let me know how large chunks you want me to split this PR in.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @pablogsal for commit 55cee70 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[rhettinger]

• I can't review this because I don't fully understand the changes
• I'm unclear why this is necessary.
• The code churn is enormous.
• There are measurable performance hits.
• There are no doc changes or tests, so I'm unclear what the user benefit would be.

[erlend-aasland]

• I can't review this because I don't fully understand the changes

The changes are described in the PR subject: "Convert itertools types to heap types"
As an improvement, and in order to make the change more explicit, I could alter the subject to "Convert itertools types to heap allocated types".

In order to fully understand the changes, I would recommend taking a look at PEP 384, PEP 573, PyType_Spec, and PyType_Slot.

I guess @vstinner or @corona10 could assist in reviewing this, if they don't mind.

• I'm unclear why this is necessary.

• issue1635741
• Why Isolate Subinterpreters?
• PEP 630

• The code churn is enormous.

#24065 (comment)

• There are measurable performance hits.

https://mail.python.org/archives/list/python-dev@python.org/thread/S5GZZCEREZLA2PEMTVFBCDM52H4JSENR/#RIK75U3ROEHWZL4VENQSQECB4F4GDELV

• There are no doc changes

The itertools docs does not contain information about how its types are implemented. I see no reason to add this information.

or tests

A module should not alter its behaviour depending on how its types are implemented. I see no reason to add type implementation specific tests for itertools.

, so I'm unclear what the user benefit would be.

• Why Isolate Subinterpreters?
• PEP 630

[vstinner]

I prefer to close the PR for now. I'm working on a PEP with @corona10 to explain the rationale. We can wait until this (future) PEP is accepted before changing itertools.

[erlend-aasland]

I prefer to close the PR for now. I'm working on a PEP with @corona10 to explain the rationale. We can wait until this (future) PEP is accepted before changing itertools.

All right!

I'll keep the branch around, then :)

[erlend-aasland]

FYI, PEP-687 was just accepted.