gh-140870: PyREPL auto-complete module attributes in import statements #140871

loic-simon

Add the ability for _pyrepl._module_completer.ModuleCompleter to suggest attributes, in addition to modules.

If the module to import from is not imported yet, ask the user if they want to import it.
This is done by adding the notion of completion action to _pyrepl.completing_reader.CompletingReader:

get_completions can return a completion action, in addtion to the suggestions list
A completion action is a prompt to show + a callback to call if the user press TAB again
The callback can return a message to show instead (here, if import failed)
After the callback is fired, get_completions is called again, so it can propose new results (here, module attributes)
If the user does any other input than TAB when the prompt is shown, the action is discarded

Other changes:

I had to tweak a little the code used to display a message below the prompt, to handle the case when it doesn't fit on the console
~~Some unrelated tests to get the coverage rate of _module_completer to 100% (in a separate commit, I can remove it if needed)~~ -> gh-140870: Full coverage for _pyrepl._module_completer #143244

Interactive WebAssembly demo (up to date): https://pyrepl-attributes-import-completion.pages.dev

Possible improvement: make modules printing stuff when imported not mess with the prompt ; see this Discourse message

Issue: PyREPL: auto-complete module attributes in import statements #140870

tomasr8

Some unrelated tests to get the coverage rate of _module_completer to 100% (in a separate commit, I can remove it if needed)

Could you extract those into a separate PR? I think those are nice to have in any case and we can merge it quickly :)

loic-simon

Could you extract those into a separate PR? I think those are nice to have in any case and we can merge it quickly :)

Sure, #143244

tomasr8

Quick initial question, can we disable the prompt for stdlib modules? I don't think there should be any concerns with auto-importing from the stdlib

loic-simon

We could, yes. I don't think we can decide to auto-import solely on module name (to take care of potential shadowing by first/third-party modules), so we would need to look the module spec location, but since we already have self._is_stdlib_module and self._global_cache that should be easy and fast!

(I'll won't be to get my hands on the code until early January tho)

tomasr8

Yup, we definitely need to look at the module location, not just the name

loic-simon

The new test is quite platform-dependant (I just fixed an iOS + Andoid only fail), it may be a good idea to trigger buildbots on this PR!

StanFromIreland

Thanks!

trigger buildbots on this PR

I can run the buildbots, do you want them now?

loic-simon

I can run the buildbots, do you want them now?

Yes please! I don't plan any more change on my side.

bedevere-bot

🤖 New build scheduled with the buildbot fleet by @StanFromIreland for commit 9c80367 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F140871%2Fmerge

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

loic-simon

Okay that was useful, buildbot failures analysis:

A lot of fails on compression.ztsd not suggestingcompress attribute, which is expected since it is an optional module, I'll use a better test case 😅
All Reafleaks buildbots fail on collections.abc not suggesting Buffer, I don't know why but I suppose it's expected too, I'll use another test case
iOS ARM64 Simulator has a lot of other fails, I don't quite understand what's going on 🤔
- maybe an importer cache issue like last time? I'll add some invalidation

Details

Missing compression.ztsd:

AMD64 Debian root PR
AMD64 RHEL8 FIPS Only Blake2 Builtin Hash PR
AMD64 RHEL8 LTO + PGO PR
AMD64 RHEL8 LTO PR
AMD64 RHEL8 PR
AMD64 RHEL8 Refleaks PR
AMD64 Ubuntu Shared PR
ARM64 MacOS M1 NoGIL PR
ARM64 MacOS M1 Refleaks NoGIL PR
ARM64 Raspbian Debug PR
ARM64 Raspbian PR
PPC64LE RHEL8 LTO + PGO PR
PPC64LE RHEL8 LTO PR
PPC64LE RHEL8 PR
PPC64LE RHEL8 Refleaks PR
s390x Fedora Stable Clang Installed PR
s390x Fedora Stable Clang PR
s390x Fedora Stable LTO + PGO PR
s390x Fedora Stable LTO PR
s390x Fedora Stable PR
s390x Fedora Stable Refleaks PR
s390x RHEL8 LTO + PGO PR
s390x RHEL8 LTO PR
s390x RHEL8 PR
s390x RHEL8 Refleaks PR
s390x RHEL9 LTO + PGO PR
s390x RHEL9 LTO PR
s390x RHEL9 PR
s390x RHEL9 Refleaks PR
x86-64 MacOS Intel ASAN NoGIL PR
x86-64 MacOS Intel NoGIL PR

Missing collections.abc.Buffer:

Missing AMD64 CentOS9 NoGIL Refleaks PR
AMD64 FreeBSD Refleaks PR
AMD64 RHEL8 Refleaks PR
ARM64 MacOS M1 Refleaks NoGIL PR
PPC64LE RHEL8 Refleaks PR
s390x Fedora Stable Refleaks PR
s390x RHEL8 Refleaks PR
s390x RHEL9 Refleaks PR

A lot of fails:

iOS ARM64 Simulator PR

loic-simon

@StanFromIreland if you could trigger a builbots run again, that would be great! Thanks

bedevere-bot

🤖 New build scheduled with the buildbot fleet by @StanFromIreland for commit da6cf7f 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F140871%2Fmerge

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

loic-simon

Ok, good news: iOS Siumlator is green!

Bad news: all Refleaks builds now fail ~all attribute completion tests 😞

Good news: I reproduced in local (python -m test pass, -m test -R: fails), on it!

loic-simon

That check fail doesn't look related to this PR (Error: The template is not valid. .github/workflows/build.yml (Line: 720, Col: 24): Error reading JToken from JsonReader. Path '', line 0, position 0.)

Refleak tests pass locally, may I ask for a last run with 🔨 test-with-refleak-buildbots @StanFromIreland, just to be sure? Thanks a lot!

StanFromIreland

That check fail doesn't look related to this PR (Error: The template is not valid. .github/workflows/build.yml (Line: 720, Col: 24): Error reading JToken from JsonReader. Path '', line 0, position 0.)

However, it is. You need to update the branch, the current version is incompatible with our CI.

Tighten the pyrepl completer around a few import-completion edge cases. Relative imports now share one resolver, find_spec failures no longer bubble out, failed-import tracking resets when sys.path changes, and double-tab import actions are cleared once consumed. The patch also keeps empty message lines intact in reader output and updates the targeted tests and comments to match the intended behavior.

pablogsal

Thanks a lot @loic-simon! I made a small commit fixing a bunch of small things and adding fixes for edge cases and related test cleanups. I also rebased the branch onto the latest main and reran the relevant smoke/tests locally.

With this I am happy to land :)

pablogsal

LGTM

loic-simon

Thanks a lot Pablo! If don't want to abuse your time, but while you have PyREPL import suggestions in mind, I have this one-line bugfix PR that could use a merge too 😄

pablogsal

Thanks a lot Pablo! If don't want to abuse your time, but while you have PyREPL import suggestions in mind, I have this one-line bugfix PR that could use a merge too 😄

Landed! Thank you for your time :)

loic-simon

Heh, of course these two would conflict 😅 fixing it

pablogsal

Thanks a lot for your contribution @loic-simon 🚀

…tements (python#140871) Co-authored-by: Pablo Galindo Salgado <pablogsal@gmail.com>

loic-simon requested review from ambv, lysnikolaou and pablogsal as code owners November 1, 2025 00:53

bedevere-app Bot added the awaiting review label Nov 1, 2025

loic-simon changed the title ~~gh-14870: PyREPL auto-complete module attributes in import statements~~ Nov 1, 2025

bedevere-app Bot mentioned this pull request Nov 1, 2025

PyREPL: auto-complete module attributes in import statements #140870

Closed

tomasr8 self-requested a review December 28, 2025 11:53

loic-simon mentioned this pull request Dec 28, 2025

gh-140870: Full coverage for _pyrepl._module_completer #143244

Merged

loic-simon mentioned this pull request Jan 1, 2026

gh-69605: Check for already imported modules in PyREPL module completion #139461

Merged

StanFromIreland reviewed Jan 11, 2026

View reviewed changes

StanFromIreland added the 🔨 test-with-buildbots Test PR w/ buildbots; report in status section label Jan 11, 2026

bedevere-bot removed the 🔨 test-with-buildbots Test PR w/ buildbots; report in status section label Jan 11, 2026

loic-simon requested a review from StanFromIreland January 24, 2026 14:30

StanFromIreland added the 🔨 test-with-buildbots label Jan 24, 2026

bedevere-bot removed the 🔨 test-with-buildbots Test PR w/ buildbots; report in status section label Jan 24, 2026

loic-simon and others added 13 commits April 5, 2026 18:54

pythongh-140870 PyREPL: add import completion for attributes

099cf71

Improve behavior when both submodules + prompt

42e734c

Auto-import stdlib modules

78e0ce6

Blacklist some stdlib modules from auto-import

4af851b

Fix tests post rebase

61e797a

Improve & test auto-import blacklist

f02fc8f

Add _ios_support to auto import blacklist

17599ed

Fix tests requiring tkinter

5125a0d

Avoid charged terminology

3bb88d1

Use always available modules in tests + invalidate caches

1013912

Don't test with html module either

0d1105f

Fix refleaks in test_pyrepl.py

7b9a26d

pablogsal force-pushed the pyrepl-import-completion-for-attributes branch from a7116a1 to d57de37 Compare April 5, 2026 18:13

pablogsal approved these changes Apr 5, 2026

View reviewed changes

bedevere-app Bot added awaiting merge and removed awaiting review labels Apr 5, 2026

pablogsal enabled auto-merge (squash) April 5, 2026 18:25

pablogsal disabled auto-merge April 5, 2026 18:32

Merge branch 'main' into pyrepl-import-completion-for-attributes

d7aa3b5

pablogsal enabled auto-merge (squash) April 5, 2026 18:46

bedevere-app Bot removed the awaiting merge label Apr 5, 2026

python deleted a comment from bedevere-bot Apr 5, 2026

ljfp pushed a commit to ljfp/cpython that referenced this pull request Apr 25, 2026

pythongh-140870: PyREPL auto-complete module attributes in import sta… …

597be3e

…tements (python#140871) Co-authored-by: Pablo Galindo Salgado <pablogsal@gmail.com>

Conversation

loic-simon commented Nov 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tomasr8 commented Dec 28, 2025

Uh oh!

loic-simon commented Dec 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tomasr8 commented Dec 28, 2025

Uh oh!

loic-simon commented Dec 28, 2025

Uh oh!

tomasr8 commented Dec 29, 2025

Uh oh!

loic-simon commented Jan 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

StanFromIreland commented Jan 11, 2026

Uh oh!

loic-simon commented Jan 11, 2026

Uh oh!

bedevere-bot commented Jan 11, 2026

Uh oh!

loic-simon commented Jan 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

loic-simon commented Jan 24, 2026

Uh oh!

bedevere-bot commented Jan 24, 2026

Uh oh!

loic-simon commented Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

loic-simon commented Jan 24, 2026

Uh oh!

StanFromIreland commented Jan 24, 2026

Uh oh!

pablogsal commented Apr 5, 2026

Uh oh!

pablogsal left a comment

Choose a reason for hiding this comment

Uh oh!

loic-simon commented Apr 5, 2026

Uh oh!

pablogsal commented Apr 5, 2026

Uh oh!

loic-simon commented Apr 5, 2026

Uh oh!

Uh oh!

pablogsal commented Apr 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

loic-simon commented Nov 1, 2025 •

edited

Loading

loic-simon commented Dec 28, 2025 •

edited

Loading

loic-simon commented Jan 10, 2026 •

edited

Loading

loic-simon commented Jan 11, 2026 •

edited

Loading

loic-simon commented Jan 24, 2026 •

edited

Loading