Command line paramater to specify test command

[fdietze]

Thanks for creating aider!

Using it for the first time, I was missing a way to specify a test command. A test command would run automatically after every change and trigger a feedback loop until the exit code becomes 0.

I would use that for all kinds of testing and type checking.

[paul-gauthier]

Thanks for trying aider and filing this issue. You can use the in-chat /test command to launch a test loop like you describe.

So you might do:

/test pytest some_tests.py

It will run them and feed the errors back to the LLM until the tests all pass and return exit code 0.

You can use control-R or the up-arrow to quickly pull up that same /test command again later in the chat.

[fdietze]

Thanks. I'm aware of /test already. But since it's always the same tests which need to run for a project (especially type checks), I thought this could be set up to run automatically by specifying a command once.

[paul-gauthier]

Would it help if you could configure a test command, and then doing /test without any args would run that configured command?

[fdietze]

That would run the tests after the changes have already been commited, right? I thought it would be nicer to only have commits for changes, which have passed the test command.

[paul-gauthier]

In general, I'm not prioritizing making a "tidy" or "perfect" git commit history. The priority is using git to make sure it's easy to undo/revert/inspect the history of changes made by the AI. If a clean commit history is important to you, you can always use git rebase to tidy things up.

[fdietze]

Ok, I see, that makes sense. Then I like your suggestion of configuring a test command which you can invoke using /test.