gh-130472: Integrate fancycompleter with the new repl, to get colored tab completions by antocuni · Pull Request #130473 · python/cpython · GitHub

As the title says, this is (WIP) to integrate fancycompleter into the new REPL.

Issue: Use colors in tab-completion in PyREPL #130472

Exciting! Just some nits.

Assuming the tests are going to pass (last famous words), this is ready to review.
So far I have tested it only on my linux machine, if anybody has access to windows and macos it would be nice to do a "visual check" that things are going as intended.
This is how it looks like on my machine:

Compared to the initial version, I'm now using _colorize.get_theme() to get the colors.
I'm not 100% sure about _colorize.FancyCompleter though: for other use cases using a dataclass seems a good idea because the set of keys is known in advance, but here you potentially have an open ended set of types (ideally, users should be able to add their own types to the config).
So maybe we could turn the dataclass into a plain dict? This would also solve the ugly workaround to allow coloring method-wrapper here:
https://github.com/antocuni/cpython/blob/13a269807d24c4df0ec7bdca7b04369171c0db16/Lib/_pyrepl/fancycompleter.py#L153-L160

/cc @ambv as I think he's the one responsible for themes.

Also, I'm getting a warning from the doc builder but I'm not sure how to fix:

Warning: py:mod reference target not found: _pyrepl.fancycompleter [ref.mod]
/home/runner/work/cpython/cpython/Doc/using/cmdline.rst:1288: WARNING: py:mod reference target not found: _pyrepl.fancycompleter [ref.mod]

_pyrepl.fancycompleter is a module, do I need to add it somewhere else?

_pyrepl.fancycompleter is a module, do I need to add it somewhere else?

Yes, but it is not documented. You can add an exclamation mark at the start to remove the link.

So far I have tested it only on my linux machine, if anybody has access to windows and macos it would be nice to do a "visual check" that things are going as intended.

I tested it on Windows 10 using the Windows Terminal app.
This is using Powershell with Campbell theme:

Co-authored-by: Stan Ulbrych <89152624+StanFromIreland@users.noreply.github.com>

…y this fixes the failures on Android CI

…d color. This is both more robust and more correct

Co-authored-by: Pieter Eendebak <pieter.eendebak@gmail.com>

…ections

Co-authored-by: Pieter Eendebak <pieter.eendebak@gmail.com>

…vance, let's add a comment to explain why

Keep pyrepl completion logic working on the uncolored completion text. The reader now strips ANSI escapes before comparing the typed stem, inserting a sole completion, computing the shared prefix, and filtering an open completion menu. This fixes colored completions that would stop refining correctly once more characters were typed. Restore readline's callable postfix behavior for attribute completions by routing single attribute matches through rlcompleter's callable postfix logic while keeping the full expr.attr stem for menu refinement. Global completion also treats soft keywords as keywords instead of trying to evaluate them. Avoid side effects while probing attribute values for coloring by not forcing property access and by preserving lazy module imports. Also make the fake color-sorting escape prefix round-trip cleanly once the match index grows past three digits. Only honor PYTHON_BASIC_COMPLETER when the environment is enabled, so pyrepl setup now respects -E / sys.flags.ignore_environment. Add regression tests for the reader behavior, callable attribute completion, property and lazy-import safety, large color-sort prefixes, and the -E setup path.

Thanks for the great work on this, @antocuni . I rebased the branch and added a handful of follow-up fixes around the fancycompleter integration, including some edge cases and extra regression coverage.

This is in good shape now, and I’m planning to land it soon so it makes the next release.

LGTM after my fixes

I think this has some issues...:

Screencast.From.2026-04-06.10-41-01.mp4

These are in fact not always "fake," leading to the issues in my comment above. Fixing that (ensuring they are greater than 100, which IIRC is the limit of valid sequences), I still have contrast issues, so maybe there is more to it:

Let's open an issue and fix this in a separate PR

…olored tab completions (python#130473) Co-authored-by: Pablo Galindo <pablogsal@gmail.com>

bedevere-app Bot mentioned this pull request Feb 22, 2025

Use colors in tab-completion in PyREPL #130472

Closed

hugovk reviewed Feb 22, 2025

View reviewed changes

tomasr8 reviewed Feb 22, 2025

View reviewed changes

AA-Turner reviewed Feb 26, 2025

View reviewed changes

antocuni marked this pull request as ready for review September 19, 2025 14:15

antocuni requested review from ambv, lysnikolaou and pablogsal as code owners September 19, 2025 14:15

bedevere-app Bot added the awaiting review label Sep 19, 2025

antocuni mentioned this pull request Sep 19, 2025

Use pyrepl+fanycompleter as the default readline for pdb #139167

Open

StanFromIreland reviewed Sep 19, 2025

View reviewed changes

Comment thread Doc/using/cmdline.rst Outdated

StanFromIreland reviewed Sep 19, 2025

View reviewed changes

Comment thread Doc/using/cmdline.rst Outdated

eendebakpt reviewed Sep 21, 2025

View reviewed changes

Comment thread Lib/_pyrepl/fancycompleter.py

Comment thread Lib/_pyrepl/fancycompleter.py Outdated

hugovk reviewed Sep 21, 2025

View reviewed changes

Comment thread Lib/_colorize.py Outdated

eendebakpt reviewed Sep 26, 2025

View reviewed changes