Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

_values with controllable duplicates suppression?

My completion-fu is failing me once more and I would like some guidance
on how to handle supplying completions, where currently I use
`_values -w`.

I have a custom `git grab` which has various heuristics and ultimately
lets me do `git grab -o orgname repo1 repo2 repo3`.  I have
tab-completion for this and it mostly works, and has for a long time.

The completion caches the results of using the GitHub API for a given
org/user, using appropriate keying into an associative array var, so
that if I hit tab multiple times there's only one remote API call until
a half hour passes and the cached data is invalidated.

I will be creating entries in the current directory, so have:


and then call, when the cache is already populated:

  _values -w "uncloned github repositories for $identifier" "${(@)candidates:|already}"

and ultimately the flow for when the cache needs to be populated ends up
with the same _values call.

The top level is:

  _arguments -C -s [options stuff] '*:REPO:_github_repos_for'

I really like the way that `_values -w` means that if I have already put
a given repo onto the command-line, it's not offered again.

My problem is that sometimes there's an organization with the same name
as one of the repos inside that organization.  Eg, "foo/foo" as well as
"foo/bar", "foo/baz", so if I do:

  git grab -o foo <TAB>

then the "foo" result is suppressed because `_values -w` sees that `foo`
is already on the command-line, and I'm only offered `bar` and `baz`.

Is there any way (context shift, something else?) to track from
_arguments which word is the start of the `*:REPO:...` completion point
and constrain _values (or equivalent) to only suppress duplicate words
which appear from that point onwards?


Messages sorted by: Reverse Date, Date, Thread, Author