python / python/cpython

Tuples should be immutable and safe in C, as well as in Python.

Open
#127,058 13 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

3.14 interpreter-core triaged type-bug
Dominant language
Python
Stars
77.2k
Forks
35.9k
PR merge metrics
PR metrics pending

Description

Bug report

Bug description:

[Apologies if this sounds a bit like a rant. I'm not blaming anyone. Just because something is the wrong choice now, doesn't mean it wasn't the right choice historically]

Tuples are immutable in Python, but we play all sorts of games in C with tuples, filling them will NULLs, mutating them and reusing them.

We do this in the mistaken belief that it improves performance.
But it doesn't. It makes the code base more complicated and fragile as we need to work around tuples that misbehave and do strange things. Any local performance gain is overwhelmed by slowdowns caused by the extra complexity in tuple code, the garbage collector and a few other places.

So let's fix this.

We need to:

  • Provide a new C API PyTuple_MakePair(). Pairs are by far the most common type of tuple that we play games with. By providing a fast way to create pairs, we can provide an upgrade path for C code that creates tuples in unsafe ways to do so safely and quickly.
  • Deprecate PyTuple_New. I don't know when we'll be able to remove it, but we should deprecate it ASAP.
  • Change PyTuple_New to fill the tuple with pointers to None instead of NULL. This doesn't fix the mutability issue, but it at least means the GC will only see valid objects. (This might break too much code, so we might just have to clearly document that tuples should be fully initialized in one go, before the tuple escapes the function it was created in)
  • Fix our own code to not use PyTuple_New() or perform tuple shenanigans. We can't reasonably expect third-party package authors to follow the rules if we don't.
CPython versions tested on:

CPython main branch

Operating systems tested on:

No response

Linked PRs
  • gh-127758

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the PyTuple_New and proposed PyTuple_MakePair C APIs, then inspect linked PR gh-127758 to understand the work already underway. Done means addressing the requested tuple-safety changes, including the API transition, initialization behavior, and removal of unsafe internal tuple usage.

Written by the indexing model from the issue text.

Assessment

Tech stack
c, python
Domain
backend
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.