help CLI shows a traceback when import failed

[Eclips4]

Currently, help will show a traceback when we type name of something which doesn't exist:

help> 123
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<frozen _sitebuiltins>", line 103, in __call__
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 2004, in __call__
    self.interact()
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 2031, in interact
    self.help(request)
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 2057, in help
    elif request: doc(request, 'Help on %s:', output=self._output)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 1781, in doc
    pager(render_doc(thing, title, forceload))
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 1755, in render_doc
    object, name = resolve(thing, forceload)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\KIRILL-1\CLionProjects\cpython\Lib\pydoc.py", line 1741, in resolve
    raise ImportError('''\
ImportError: No Python documentation found for '123'.
Use help() to get the interactive help utility.
Use help(str) for help on the str class.
>>>

I think, this shouldn't be shown.
Is there any reason for this?
Can we show something like
No python documentation found for '123'.?

Reproducible on current main.

Linked PRs

• gh-102541: Hide traceback in help prompt #102614
• [3.12] gh-102541: Hide traceback in help prompt (gh-102614) #105778
• [3.11] gh-102541: Hide traceback in help prompt (gh-102614). #105830
• gh-102541: Fix Helper.help("mod") for non-existent mod #105934
• [3.12] gh-102541: Fix Helper.help("mod") for non-existent mod (GH-105934) #106322
• [3.11] gh-102541: Fix Helper.help("mod") for non-existent mod (GH-105934) #106323
• gh-102541: Add test case for help() for non_existent_module #106340
• [3.12] gh-102541: Add test case for help() for non_existent_module (GH-106340) #106639
• [3.11] gh-102541: Add test case for help() for non_existent_module (GH-106340) #106640

[terryjreedy]

I verified above in 3.11.2 and 3.12.0a6. 3.10.10 is better -- Gives import error message without traceback. But even better would be to give error message without ImportError and return to help> prompt. This should possibly be a separate PR as there might be more discussion that change. (It is possible that 3.11 change is intentional -- this should be checked.)

[Eclips4]

How I understand, it's introduced in #89130.

[Eclips4]

Yeah, i found right way to fix it. I I have slightly changed signature of some functions, which now have argument is_cli (set default to False).
That's doesn't break compatibility with issue which i mentioned above.
Currenly, in help prompt errors looks like that:

help> abd
No Python documentation found for 'abd'.
Use help() to get the interactive help utility.
Use help(str) for help on the str class.

help>

@terryjreedy What you think about this?
BTW, all tests are passed.
I think, error message we should change, any ideas? (this will make us to change some lines in test_pydoc, but it's not hard)

[Eclips4]

@terryjreedy
I think there's no need to backport it to 3.10 branch. I cannot reproduce it there

[terryjreedy]

In closed duplicate #105702, I noted that there was no bug in latest release of 3.8, 3.9, or 3.10. Patch should restore previous behavior, shown on that issue, which is same as shown above. I just installed 32-bit 3.11.0 and it does have the bug. Will look at PR.

[Eclips4]

In closed duplicate #105702, I noted that there was no bug in latest release of 3.8, 3.9, or 3.10. Patch should restore previous behavior, shown on that issue, which is same as shown above. I just installed 32-bit 3.11.0 and it does have the bug. Will look at PR.

Yeah, so, please, delete label 3.10 on this issue, and add 3.13 label, this still reproducible on current main

[terryjreedy]

@ambv Kirill believes this regression resulted from your #27868. Whichever PR is the cause, CI passed because help() interaction is not adequately tested. Can help Kirill add one to the fix? (AFAIK, not trivial.)

This also caused a regression in IDLE, which may or may not be independently fixable. #105689 It would be really nice to fix this before the next releases.

[corona10]

CI passed because help() interaction is not adequately tested.

Yeah, we have to add the test for interaction.

[Yhg1s]

Is the only thing left to do here improving test coverage?

[Eclips4]

Is the only thing left to do here improving test coverage?

Sadly, but no. When I tried to write a test for this case, I founded another bug. I have already sent PR to fix.

[terryjreedy]

This issue is at least a near duplicate of issue #98374 and PR #98450, about help('invalid'), closed and merged.

@Eclips4 The 2nd fix is now merged and backported. Can you now write a test for the first one?

[Yhg1s]

Okay, so now it's just about adding the last test case, right? 😅

[Eclips4]

Okay, so now it's just about adding the last test case, right? 😅

Yeah, you're right.
I already sent a PR to add this, and waiting for review.

[terryjreedy]

I reviewed and edited and intend to merge and close this after CI passes.

[Eclips4]

Thank you very much @terryjreedy for your help!

[Eclips4]

Now, all PR's merged and backported.
Everything that needed to be done was done.
Again, thanks Terry for his thorough help.