gh-102628: Fix sqlite3 CLI promt text on Windows

[erlend-aasland]

CTRL-Z is the Windows equivalent of CTRL-D.

• Issue: Prompt text in sqlite3 CLI (new in 3.12) is Linux oriented and doesn't apply to Windows #102628

[erlend-aasland]

@terryjreedy, will this work on IDLE on Windows, or does that only support CTRL-D?

[terryjreedy]

First, note that ^Z by itself does not work, which is confusing to *nix users who just learn 'Use ^Z instead of ^D". So maybe it should say "^Z + " or perhaps "^Z in a Windows console". But experienced Windows users 'should' know about -- although I managed to forget occasionally from disuse. What do you think?

Currently, IDLE only supports ^D, which is why I have not done this yet. About a decade ago, someone decided that IDLE should work same across platforms. Because of this issue, I was thinking maybe ^Z should have been left if possible. I don't know if we can make ^D work in CommandPrompt. I suspect not.

The current PR is an improvement, so merge if you want, or with either alternative.

[erlend-aasland]

Is there a stable way to detect if we are running in IDLE? I've seen people check if idlelib.run is in sys.modules, but I have no idea if that is too hacky or not.

[erlend-aasland]

I landing this, as it is an improvement over the current prompt. I'm slightly reluctant to expanding the prompt as hinted, since I think it is preferable to keep it as succinct as possible.

I still think we should try to fix the IDLE-on-Windows case, though.

[erlend-aasland]

Thanks for the insight, @terryjreedy. Thanks for the report, @Dutcho

[terryjreedy]

You merged after I decided to make the 'hack' official and wrote the following. Please do the followup or ask me to.

Good question. At least some people are checking idlelib.run' in sys.modules because I have recommend that probably on python list and certainly on stackoverflow. IDLE starts the user code execution subprocess with `python -c "import('idlelib.run')..." as it is the idlelib module that actually runs user code.

But what about when one starts IDLE with '-n' (no subprocess). This is 'deprecated' and not actively maintained and tested. (However, it has no removal date and I might reverse this.) It happens that other modules currently import common items from run, but these should be moved to util. To avoid breaking what I have suggested, I added a note to my draft issue for populating util that if run imports are removed, 'idlelib.run' must still be added to sys.modules. I should say something in the IDLE doc too.

[erlend-aasland]

I merged, since you said:

The current PR is an improvement, so merge if you want, or with either alternative.

Sorry if I misunderstood. Also, I do plan to follow up this; I'll request your review.

If the idlelib.run trick works in most cases, I think we should use that. The deprecated no-subprocess variant you mention seems too much of a corner case to support, in my opinion.

[erlend-aasland]

For the record, the follow-up PR is here:

• gh-102628: Fix sqlite3 CLI prompt in IDLE on Windows #103945

[terryjreedy]

No misunderstanding. Your question really was a good one. It is one thing to say on SO "Here's a hack that works now" and another to use it in the stdlib. So I reviewed the code to make sure I understood why it works. It turned out that the answer is completely different for the 2 different execution modes. Having it always work is trivial. pyshell already manipulates sys.modules for another back-compatibility issue.