Make npm run-script completion faster with jq

[thalesmello]

Description

When jq is available, it's actually faster to invoke jq and parse the package.json
invoking the npm command.

Also, prior to this commit, both __fish_complete_npm and __fish_npm_run were being run
whenever completions for npm run subcommand was being used, which was actually making
repetitive work (invoking npm command twice). This pull request is supposed to make completion
without jq faster as well

TODOs:

• Changes to fish usage are reflected in user documentation/manpages.
• Tests have been added for regressions fixed
• User-visible changes noted in CHANGELOG.md

[thalesmello]

I'm actually not sure. Should there be tests for completions?

Also, should I check one of the other checkboxes?

[thalesmello]

This is actually repeated code from the while loop below, as I'm not proficient in fish scripting.

How could I improve this? Suggestions would be appreciated.

[faho]

Currently, the best you could do is make a function that calls either jq -r [...] or command npm run | string match [...], and then only do the while read stuff here.

[krader1961]

Should there be tests for completions?

Ideally, yes. But at this time we do not have any unit tests specifically for completions. If you think you can implement unit tests for your completions we would be happy to consider how that could be expanded to all completion scripts.

[thalesmello]

@krader1961 Is there any API that could be used to:

1. Type something to the terminal
2. Invoke the completion for the current command
3. Get the completions in a variable for assertion?

[faho]

Currently, the best you could do is make a function that calls either jq -r [...] or command npm run | string match [...], and then only do the while read stuff here.

[faho]

set value (string sub -l $trim -- $value)

[faho]

Type something to the terminal

We do that via expect in our tests. It's a bit weird and can have some issues with timing, but in general it works. However, it's not necessary in this case because

Invoke the completion for the current command

complete -C"$commandline", e.g. complete -C"npm " can be used to print the completions. Note that the argument parsing is idiosyncratic here - there can be no space between the "-C" and the commandline (IIRC).

Get the completions in a variable for assertion?

complete -C is pretty much a regular command and can be used in a command substitution or similar.

---

The real problem with tests for completions is a different one:

• We cannot guarantee that every single tool for every single completion is installed, so we need to allow for skipping tests (which we currently don't)

• Completions often depend on a bunch of data (in fact that's the most interesting ones), so we'd need to ship that (e.g. a fake git repository for the git completions)

• Tools often change their output based on configuration (which is generally okay), so we'd need to override that

Currently, this hasn't been done. I'd appreciate any good ideas.

[thalesmello]

@faho I've refactor my original pull request. Would you be kind to take a look?

We cannot guarantee that every single tool for every single completion is installed, so we need to allow for skipping tests (which we currently don't)

Regarding completion testing, would it be reasonable to use docker as a dependency for testing the project?

By doing so, it would be possible to many different test environments, so it would be possible to test completions.

Also, I think it would be pretty straightforward to execute only one test script using docker, or maybe skip completion testing in the main test suite. What do you think?

Completions often depend on a bunch of data (in fact that's the most interesting ones), so we'd need to ship that (e.g. a fake git repository for the git completions)

That could be accomplished with a fixtures directory. Would be complicated, but doable, I think.

Tools often change their output based on configuration (which is generally okay), so we'd need to override that

By using docker, we could guarantee we're always testing the same environment.

Anyhow, all of these look like a lot of changes for a simple pull request regarding a small completion improvement. Maybe we could foward this discussion to an issue?

[faho]

Anyhow, all of these look like a lot of changes for a simple pull request regarding a small completion improvement. Maybe we could foward this discussion to an issue?

Yes, please. See #4249.