Most changes to Pythonrequire a NEWS entry.

Please add it using theblurb_it web app or theblurb command-line tool.

There is something that Python core developers decide.

1. Supporting "TAB" symbolic character name for libedit in "cmd.py".

readline_doc = getattr(readline, '__doc__', '')if readline_doc is not None and 'libedit' in readline_doc:     if self.completekey is 'tab':        self.completekey = r'"\t"'

OR just

if self.completekey is 'tab':    self.completekey = r'"\t"'

If this is supported, then other third-party libraries can support Libedit tab completion without changing their code.
I don't want this support because it will give a false impression of common key sequences and the requirement to support other symbolic character names.

2. Removing editrc examples in readline.rst.

Because of the absence of the GNU inputrc example inreadline.rst, the editrc example may be just for hotfix to do a tab complete in cmd or pdb.

For Vim keybinding in the editrc example(python:bind -v), I think it's valid. For this reason, Vim keybinding for GNU inputrc example is needed.

3. Changing keybindings in related test files.
   There are^I andtab in related test files. If you need me to change them, I will do it.

Just so you know, if you want to debug this problem, please check what I found below.

Use one of these lines to enable tab completion in Python using~/.editrc for libedit.

python:bind "\t" rl_completepython:bind "\011" rl_completepython:bind \\t rl_completepython:bind "" rl_complete# "" is a tab.

To use them withreadline.parse_and_bind() in Python, remove'python:' prefix. To make clear, If you didimport readline; readline.parse_and_bind('bind "\t" rl_complete') in pdb interactively (e.g.,$python -m pdb somefile.py), it will autocomplete file names of the current directory. It's normal because the binding statement wasn't called in the pdb library.

The key sequences("\t","\011") of the first two lines in the code block above are documented inman editrc. The first one is called a backslashed escape sequence. The second one is called the ASCII character corresponding to the octal number(or octal escape character).

But the last two lines are not. I stumble across them.

To do another complete command in Libedit, this is possible.
import readline; readline.parse_and_bind('bind -s "\011" ee')
In this case, If you press a tab, 'ee' will be typed.
Other possible options are appeared inman editrc.

I found several ways to debug it. (these are examples.)

• In Python interpreter,import readline; readline.parse_and_bind('bind -s "\011" ee')
• Replace'tab' with'' in the pdb.py and cmd.py of your python directory in$PATH. e.g./usr/local/Cellar/python@3.11/3.11.4_1/Frameworks/Python.framework/Versions/3.11/lib/python3.11/pdb.py.

How to check the keybinding setting applied with Libedit.

import readlinereadline.parse_and_bind('bind')

This shows the user's current applied libedit readline setting.

I don't think we should change the default and hope the completekey passed in is correct. It could break users code, for example, if they compareself.completekey == tab. Also, if the user subclassed thecmd.Cmd class and did somethingpdb does like

classMyCmd(cmd.Cmd):def__init__(self,completekey='tab'):cmd.Cmd.__init__(self,completekey)

They will still have the issue.

We should convert the completekey internally and just do a different binding command withlibedit. Minimal code changes and fix things for good.

Also, why do you want to change thesite.py file? It works for now I believe. Python REPL has the complete feature withlibedit.

I believe the right way to do this is just add an extra check incmd.Cmd forlibedit, and do the correct thing (converttab to"\t" or the equivalent inlibedit, then use a different binding command).

Also, if you want to finish this, you'd need a regression test on at leastcmd.Cmd, it would be easier after#110945 is merged so you'll have a helper function for pty.

I can help you with the review and find the core-dev for approval if you want to work on this.

Yes, what you pointed is valid.
So you want Python should change the tab internally for portability.

Minimal code changes and fix things for good.

Because I was unsure about the decision I made. This is what I asked in number 1#107748 (comment).
In that case, a user who uses libedit with your custom object will experience when press 't' will be ignored sometimes.

At the time I wrote this PR, I thought, that if a user did something wrong, the wrong result must be thrown. And the docs must give a pledge. (You know when learning the linux thing there must be some reason I don't know and some result is unintuitive or counterintuitive for newcomer.).

Thank you for giving me direction for what I consider. I will do that within a week as soon as possible.

The second thing you mention(site.py) is just for a tone and manner thing(second and third thing that I asked in#107748 (comment)). I will revert it.

Thank you for helping me.

I believe you only need to changecmd.py, which would make this PR compact and easy to review. I'd recommend you to wait for#110945 so you'll have a test framework for completion. Actually enabling the test in that PR on libedit would just do.

BTW, you should not useis to compare strings. In this very specific case it actually works because of the python internal cache for intern strings, but that's definitely not what you should rely on. Always use== to check the string equality.

I have the hardest time believing that commit can cause the completion issue. The code should not be executed at all for completion. Are you sure you did the test correctly?

I'm sorry. You are right. Maybe my environment goes something wrong.

Hi@Constantin1489 , now that#111826 is merged, you have the utility to write tests for the feature. Could you:

• Add a test intest_cmd.py to do some basic testing for command completion? It does not need to be super fancy, refer togh-110944: Move pty helper to test.support and add basic pdb completion test #111826 for the usage ofrun_pty().
• Remove the check forlibedit ingh-110944: Move pty helper to test.support and add basic pdb completion test #111826 so it would be another confirmation that the feature works withlibedit? Do not remove theimport_module piece because that's the magic to skip the test ifreadline is not available.

And one more thing I always forget to point out (sorry for that)!

At the end of theCmd class documentation (before.. _cmd-objects:), add a versionchanged note like

   .. versionchanged:: 3.11   ``completekey='tab'`` is replaced by ``'^I'`` for ``editline``.

Also, please ignore the free-threaded CI failures; they're not your fault and they're allowed to fail for now.

🤖 New build scheduled with the buildbot fleet by@encukou for commit652100f 🤖

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

Thank you for your contribution!

Hi@gaogaotiantian, as libedit is now supported, is this part intest_pdb.py not needed anymore?https://github.com/python/cpython/blob/main/Lib/test/test_pdb.py#L4780-L4781

Can I make a PR to remove it?

@tanloong, I don't believe you can. If you read the full discussion (to be more specific -#107748 (comment)), you'll find that we tried to do that but failed. The problem islibedit has some issues for inputing spaces so all the multiline tests would fail. Feel free to give it a try (you'll need libedit so make surereadline.backend returnseditline before you start working on it). If you can solve the problem, I'd be happy to review the PR.

@gaogaotiantian, thanks so much for clarifying, I totally missed it from not reading carefully.

I want to be honest I'm not sure if I can fix the libedit space issue, but I'll still give it a try. Thanks again!