Refuse to quit cli if a transaction is ongoing.

[dbaty]

Description

If a transaction is ongoing, quitting the cli (with "ctrl-d" or "quit") is not possible anymore. An error message is displayed. The user is forced to commit or rollback the transaction before quitting.

Fixes #1071.

Notes

1. I did not add a new configuration option. This commit thus changes the current behaviour of pgcli (for good, I believe). This is debatable, I could add a configuration option (off by default) if preferable.
2. I added a "behave" test. This is my first one, feel free to comment if it's not the right way to do it.
3. Demo:

Checklist

• I've added this contribution to the changelog.rst.
• [ ] I've added my name to the AUTHORS file (or it's already there). (I am already in this file.)
• I installed pre-commit hooks (pip install pre-commit && pre-commit install), and ran black on my code.
• Please squash merge this pull request (uncheck if you'd like us to merge as multiple commits)

[codecov-commenter]

Codecov Report

Patch coverage: 52.62% and project coverage change: -4.74% ⚠️

Comparison is base (6884c29) 84.15% compared to head (5f38269) 79.42%.
Report is 62 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1400      +/-   ##
==========================================
- Coverage   84.15%   79.42%   -4.74%     
==========================================
  Files          21       25       +4     
  Lines        2720     3008     +288     
==========================================
+ Hits         2289     2389     +100     
- Misses        431      619     +188     

Files Changed	Coverage Δ
pgcli/completion_refresher.py	91.76% <ø> (+1.17%)	⬆️
pgcli/magic.py	0.00% <0.00%> (ø)
pgcli/packages/parseutils/tables.py	97.67% <ø> (ø)
pgcli/packages/sqlcompletion.py	97.67% <ø> (ø)
pgcli/pgcompleter.py	96.95% <ø> (ø)
pgcli/pgstyle.py	64.00% <ø> (ø)
pgcli/pyev.py	15.38% <15.38%> (ø)
pgcli/pgtoolbar.py	25.71% <20.00%> (-5.72%)	⬇️
pgcli/explain_output_formatter.py	46.15% <46.15%> (ø)
pgcli/key_bindings.py	52.94% <66.66%> (-0.19%)	⬇️
... and 8 more

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[j-bennet]

@dbaty Thank you for the PR! Very nice work, you went above and beyond, what with the behave test, and the demo.

I think I would do things a little differently. If the user tries to quit and a transaction is in progress, I would output a message as follows:

A transaction is ongoing. Choose c to COMMIT or r to ROLLBACK Default: [r] >

If the user presses ENTER, transaction is rolled back, and pgcli exits.
If the user types r and presses ENTER, same.
If the user types c and presses ENTER, transaction is rolled back, and pgcli exits.
If the user types anything else besides r or c, same message is repeated.

It's slightly more complicated to implement, but nicer user experience, because if the user is ok with losing their changes, they can just press ENTER once more to quit, without typing anything.

What do you think?

[dbaty]

That makes sense. I'll try to implement that next week.

[dbaty]

I'll try to implement that next week.

Yeah, right. By "next week", I actually meant "within 6 months"... Anyway, it's done.

Notes:

• I struggled a bit to add Behave tests but they pass. Suggestions are welcome.
• @j-bennet : I slightly adapted your proposal to add a way to cancel the exit. Use case: I have an ongoing transaction, I type "Ctrl-D" because I forgot what I was doing (or because I have very clumsy fingers), but I don't want to commit nor rollback yet. I can "cancel" the exit and do more stuff before a commit or rollback.

I recorded a new demo:

[j-bennet]

Hi @dbaty, this looks great, and works as expected. Thanks a lot, great feature! Merging. 🍫

[matt0089]

Hi everyone 👋. Thank you for this great tool, not sure how I'd get work done without it tbh.

I have an issue where pgcli thinks I've started a transaction periodically, when I haven't. Then when I try to exit, it doesn't let me, so I have to do a pkill in another window.

It usually happens after I've cancelled a query (with ctrl-c).

Is this a known issue? I've searched Github issues and haven't found any mentions of this behavior. If not, I may open one (with more details). I'll see if I can come up with a fix also.

I'm using pgcli v4.1.0

[dbaty]

Hi. Could you please open an issue that details how to reproduce it? Thanks!

[matt0089]

Thanks @dbaty! I've been really busy this past week. I'll open a separate issue with details to reproduce. I'm on pgcli version 4.1.0, so that at least looks up to date.