Awesome! I found a couple issues:

This shows variables, but the "variables" header is at the top of the output, not above the variables:

just <TAB> 

This completes VARIABLE_NAME:

just --set V<TAB> 

But it adds a = after VARIABLE_NAME, but = is only used for the just VARIABLE_NAME=value shorthand, not the --set form, which uses just --set VARIABLE_NAME value.

This produces an error:

just --set VALUE foo <TAB>

It looks like perhaps 'foo' isn't being passed to the underlying invocation?

The tests were failing because a new version of clippy introduced a new lint, and because the generated completion scripts were behind the code in subcommands.rs. I pushed diffs that fix both errors.

But it adds a = after VARIABLE_NAME, but = is only used for the just VARIABLE_NAME=value shorthand, not the --set form, which uses just --set VARIABLE_NAME value

By convention, one argument can be followed by at most one flag asaik. It's out of my knowledge to pass two argument to one flag. I can remove the completion part of --set option if you think it is not suitable here.

This shows variables, but the "variables" header is at the top of the output, not above the variables

Can I have a look at your zshrc?

I pushed diffs that fix both errors.

Thanks :)

By convention, one argument can be followed by at most one flag asaik. It's out of my knowledge to pass two argument to one flag. I can remove the completion part of --set option if you think it is not suitable here.

Ah, okay. I think in that case it's probably better not to try to complete --set in that case.

Can I have a look at your zshrc?

Yup, it's here: https://github.com/casey/local/blob/master/etc/zshrc

It refers to other files, which are also in that repo, and I think they're all in this folder: https://github.com/casey/local/tree/master/etc/zsh

Thanks :)

This shows variables, but the "variables" header is at the top of the output, not above the variables

This issue doesn't exist on my side. I'll have a look at your code once I have time.

This issue doesn't exist on my side. I'll have a look at your code once I have time.

Hmm, interesting. I don't want to hold up this diff because of a problem that might only be on my end.

I pushed a diff that makes the test pass, but when testing locally I ran into the following issue:

just --set V<TAB>

Completes to:

just --set VARIABLE_NAME=

The = can only be used with sets of the form just VARIABLE_NAME=value, and not --set VARIABLE_NAME value, which doesn't take a =. I tried having a look at the code to see if I could fix this myself, but I have absolutely no idea how zsh completion scripts work 😅

I noticed that this PR introduces a numargs variable, that maps flags to the number of arguments they take. Will this have to be kept in sync with new flags, and what will happen if a new flag is added that takes arguments, but this map isn't updated?

I noticed that this PR introduces a numargs variable, that maps flags to the number of arguments they take. Will this have to be kept in sync with new flags, and what will happen if a new flag is added that takes arguments, but this map isn't updated?

I'm afraid not. Probably checking the first argument that matches the name of an existing recipe should work if you'd like to avoid the args checking hack. I'll try that once I have time.

I'm afraid not. Probably checking the first argument that matches the name of an existing recipe should work if you'd like to avoid the args checking hack. I'll try that once I have time.

That would be great. My expectation is that I'll forget to update this script when I add new command line options. If the completion script has fewer features, like not completing certain things that it could, I think that worth it if it doesn't get out of date.

Hi @casey, about:

The = can only be used with sets of the form just VARIABLE_NAME=value, and not --set VARIABLE_NAME value, which doesn't take a =.

Do you think just could be changed to allow a space and a = when using --set?
i.e. to have both '--set VARIABLE_NAME value' and '--set VARIABLE_NAME=value' be valid syntax.
It seems pretty common for other commands to allow both.
And doing this should allow completing the variable name too?

This is definitely a good idea, but I don't think it can be implemented without undesirable side-effects with the argument parser, clap, that Just uses.

When I set the "value delimiter" for --set to =, this does work as you'd expect --set foo=bar. However, it will error if you do --set foo=bar=baz, i.e. if you use = in the value that you're trying to set. I think this would be undesirable, since it would mean that setting variables would fail if the value happened to contain a =.

(The reason it fails is because it interprets foo=bar=baz as three arguments to --set, which is like if you did --set foo bar --set baz, the second set only has the key and no value.)

I think this has an issue, if any flags or variables are given, recipe name tab completion doesn't work, and it instead shows the usage string for the first recipe:

$ just JUST_LOG=bar <TAB>
# add git log messages to changelog
changes:
    git log --pretty=format: >> CHANGELOG.md

It seems that I've accidentally removed one line from the source code. It should be fine now.

It looks like there's some state that's persisting between calls to the completion script.

$ just JUST_LOG=foo <TAB> # shows recipes and variables
$ just JUST_LOG=foo chang<TAB> # completes to `changes`
$ just JUST_LOG=foo <TAB> # now shows the text of the `changes recipe`

It looks like there's some state that's persisting between calls to the completion script.

Thanks for the effort on debugging. Fixed now.

Awesome, looks good!