Complete bug IDs where appropriate #531

krobelus · 2021-01-24T13:16:00Z

This completes bug IDs in bash and fish. I haven't gotten completions to work in zsh.
In fish, the bug titles are shown as completion label, and can be searched for.
I don't know if/how other shells can show the completion label.
Maybe they don't use "\t" as separator but it seems to do no harm.

The interface does not yet seem ideal because sometimes subcommands and
bug IDs are both valid. Since there can be lots of bug IDs, they overshadow
the subcommands.
I wonder if instead of

git bug status open 0123acd

we should use

git bug status --open 0123acd

or if there's some other way to avoid ambiguity.

If that doesn't work out, fish has a -k/--keep-order flag, so subcommand
completions can be listed before bug IDs, but that's neither portable nor
a pretty solution.

Part of #493

MichaelMure · 2021-01-24T13:59:52Z

THIS IS AWESOME!

This completes bug IDs in bash and fish. I haven't gotten completions to work in zsh.

I think the ZSH completion code in cobra does support yet this new completion framework. I think it will work eventually, without any changes required on the git-bug side.

In fish, the bug titles are shown as completion label, and can be searched for.
I don't know if/how other shells can show the completion label.
Maybe they don't use "\t" as separator but it seems to do no harm.

I think the \t thing is to tell cobra about the label associated with the completion. If the shell doesn't support that, it will just get ignored.

The interface does not yet seem ideal because sometimes subcommands and
bug IDs are both valid. Since there can be lots of bug IDs, they overshadow
the subcommands.
I wonder if instead of
git bug status open 0123acd
we should use
git bug status --open 0123acd
or if there's some other way to avoid ambiguity.

Hmmm dunno. Even with changing status we'll have the problem with other commands that have subcommands. I think the docker completion has the same "issue" and it still works AFAIK. Now, if the completion order is not maintained between your code and the shell, maybe that's something worth reporting to https://github.com/spf13/cobra?

krobelus · 2021-01-24T18:51:51Z

I'll push a minor fix - completions don't work yet for a command like git bug comment add -> fixed

commands/helper_completion.go

krobelus · 2021-01-24T19:07:53Z

Hmmm dunno. Even with changing status we'll have the problem with other commands that have subcommands. I think the docker completion has the same "issue" and it still works AFAIK.

Yeah, --open is probably no good, that was just a quick thought.
I can't think of a docker command that has the same issue. I think it's always be possible to know if the next argument is a container/image or something else.
Another way to avoid the ambiguity would be to split it into two subcommands, comment show and comment add, though that's ugly. Anyway, it's not a big deal, and I'm quite new to git-bug, so I'll have a better perspective as I use it more.

Now, if the completion order is not maintained between your code and the shell, maybe that's something worth reporting to https://github.com/spf13/cobra?

Ok, I'll report that. Looks like there are some other fish-specific issues over there that need addressing, I'll take care of those.
(Edit: not sorting doesn't seem to be easy to add with the way fish completions are implemented.. maybe I can think of a way)

krobelus · 2021-01-25T16:37:59Z

Ok, I've extended this to all subcommands and flags. This should cover all subcommands.
The label completion is especially clever, because it also takes into account labels which are already on the commandline.

In fish, git-bug ... completions work fine but git bug ... completions need a small wrapper, which will be merged soon (fish-shell/fish-shell#7652).. With that, fish completions can be enabled with
ln -s $PWD/misc/fish_completion/git-bug ~/.config/fish/completions/git-bug.fish

commands/helper_completion.go

MichaelMure · 2021-01-25T18:07:05Z

doc/gen_docs.go

@@ -21,22 +20,15 @@ func main() {
 		"Markdown": genMarkdown,
 	}

-	var wg sync.WaitGroup


why are you removing that? Fast build is nice

See the commit message - without this commit I usually get a data race when running make.
I didn't really investigate what the real fix should be. Do you think we should use global variables for the Cobra commands?

It's the opposite actually. Cobra use a global map here. This global map is shared even if multiple independent root command are created.

It looks like it could be easily solved in cobra though ...

well, not really easy but doable

MichaelMure · 2021-01-26T10:22:43Z

I was going to report the issue to cobra but I see you did it already ;)

I made it into its own issue though, with some extra possible solutions: spf13/cobra#1320

MichaelMure · 2021-01-31T19:19:35Z

Let's wait a bit to see if something is moving on the cobra side. We'll merge either way at the very last before the next version, don't worry.

MichaelMure · 2021-07-02T16:44:16Z

The cobra issue got resolved upstream, there is hope! spf13/cobra#1438

6543 · 2021-08-01T12:33:40Z

cobra bump is here ... need to be tested and then we can add it in ?!

6543 · 2021-08-01T12:47:32Z

ok tested and bump is in @krobelus can you rebase?

MichaelMure · 2022-03-01T20:20:49Z

hey hello,

I finally get more and more time to work on this project, sorry for the crazy long delay.

I've been trying out this PR to see if everything was fine. The build process got fixed indeed but ... the completion doesn't work for me. It might be because of the cobra upgrade as they made some related change, but they claim to have kept backward compatibility. I've looked around and everything seems fine.

Any idea? It's probably something silly.

MichaelMure · 2022-03-01T20:29:56Z

It's also possible that I'm just using it wrong.

krobelus · 2022-03-02T20:46:32Z

fish and bash completions seem to work (although bash only works with git-bug, not git bug); added steps to the commit message. zsh didn't work on my brief attempt.

MichaelMure · 2022-03-03T15:42:07Z

@krobelus there is this little bash thing that should map git bug to git-bug for the completion, isn't that working? https://github.com/MichaelMure/git-bug/blob/master/commands/root.go#L54-L60

Maybe it's not being exported as part of the completion script anymore?

MichaelMure · 2022-03-12T10:54:55Z

I tried a few things, I can't get it to work with bash (I didn't try the others):

a simple source misc/bash_completion/git-bug bring the completion of the sub-command names, but not the options (--help and so on, which was working at some point), and git-bug select ^I gives filename completion.
your command bash --rcfile <(echo source misc/bash_completion/git-bug) brings a new shell but any attempt at completion yield bash: _get_comp_words_by_ref : command not found

MichaelMure

I'm trying to figure out what's happening. The following actually returns something that looks ok:

: git-bug __complete select 68
681b740	sqdsqdqsdqsdsq
e4c9831	sqdjlksqdsdiqolz
:0
Completion ended with directive: ShellCompDirectiveDefault

And yet, when I try git-bug select 68^I nothing happens.

MichaelMure · 2022-03-12T11:19:44Z

commands/helper_completion.go

+			completions[i] = bridge + "\t" + "Bridge"
+		}
+
+		return completions, cobra.ShellCompDirectiveDefault


Out of curiosity, why not ShellCompDirectiveNoFileComp here?

MichaelMure · 2022-03-12T11:20:30Z

commands/helper_completion.go

+
+func completeBug(env *Env) validArgsFunction {
+	return func(cmd *cobra.Command, args []string, toComplete string) (completions []string, directives cobra.ShellCompDirective) {
+		if len(args) > 0 {


It'd be nice to complete even if len(args) == 0. That's one of the problem I had, I tried to complete git-bug select which didn't work as your code returns an empty array.
If you did that to avoid returning thousands of completion proposition, maybe we should have an arbitrary limit?

MichaelMure · 2022-03-12T11:21:55Z

commands/helper_completion.go

+		allIds := env.backend.AllBugsIds()
+		bugExcerpt := make([]*cache.BugExcerpt, len(allIds))
+		for i, id := range allIds {
+			var err error


what about filtering here to only output the ones with toCompleteas prefix?

MichaelMure · 2022-03-12T11:52:20Z

I found the root cause, GenBashCompletionV2 need to be called instead of the legacy one.

krobelus · 2022-03-19T13:00:49Z

+func completeBug(env *Env) validArgsFunction { + return func(cmd *cobra.Command, args []string, toComplete string) (completions []string, directives cobra.ShellCompDirective) { + if len(args) > 0 { It'd be nice to complete even if `len(args) == 0`. That's one of the problem I had, I tried to complete `git-bug select` which didn't work as your code returns an empty array.

Right, you found another bug! I'm afraid I don't have much time these days, so if someone wants to take over go ahead.

If you did that to avoid returning thousands of completion proposition, maybe we should have an arbitrary limit?

I think shells will already filter completions, so I don't think we need to. In particular, fish sometimes shows all completions for which the commandline's token is a subsequence of the completion. So if we filter, we might actually lose some valid completions, and I don't think there's a need filter unless we are returning many megabytes of data.

Complete bug IDs, bridges, users, labels where appropriate. This works in bash and fish. ZSH is not yet supported by Cobra. In fish, descriptions (the part of a completion after the "\t") are shown as completion label, and can be searched with Ctrl+S. Reproduce with fish -C 'source misc/fish_completion/git-bug' git bug select ^I (tested with fish version 3.3.1) Also works with bash, but only for "git-bug" (with the dash) bash --rcfile <(echo source misc/bash_completion/git-bug) git-bug select ^I Closes MichaelMure#493

… function

MichaelMure · 2022-05-01T10:54:32Z

With @tst2005 fix for bash and a bunch of small adjustments I think we have something really cool!

Thanks @krobelus, this is awesome.

MichaelMure approved these changes Jan 24, 2021

View reviewed changes

krobelus force-pushed the complete-bug-ids branch from 7fa0ce0 to df1db38 Compare January 24, 2021 18:53

krobelus commented Jan 24, 2021

View reviewed changes

commands/helper_completion.go Show resolved Hide resolved

krobelus force-pushed the complete-bug-ids branch 2 times, most recently from 15cd8de to 8d23da6 Compare January 25, 2021 16:29

MichaelMure reviewed Jan 25, 2021

View reviewed changes

commands/helper_completion.go Outdated Show resolved Hide resolved

MichaelMure reviewed Jan 25, 2021

View reviewed changes

commands/helper_completion.go Show resolved Hide resolved

MichaelMure reviewed Jan 25, 2021

View reviewed changes

krobelus mentioned this pull request Jan 26, 2021

Default app design encourages data race spf13/cobra#1290

Closed

krobelus force-pushed the complete-bug-ids branch from 8d23da6 to a1db6ce Compare February 10, 2021 17:27

krobelus force-pushed the complete-bug-ids branch from a1db6ce to b2effa9 Compare August 1, 2021 17:48

krobelus force-pushed the complete-bug-ids branch from b2effa9 to 715a8eb Compare March 2, 2022 20:41

MichaelMure reviewed Mar 12, 2022

View reviewed changes

MichaelMure and others added 3 commits May 1, 2022 10:54

completion: use the correct GenBashCompletionV2 instead of the legacy…

f25690d

… function

misc: fix bash completion with "git bug"

edc8b75

ls: fix incorrect query parsing with quotes escaped by the shell

b9991d8

MichaelMure force-pushed the complete-bug-ids branch from dcb3965 to beffeaa Compare May 1, 2022 10:32

completion: lots of small ironing

3d534a7

MichaelMure force-pushed the complete-bug-ids branch from beffeaa to 3d534a7 Compare May 1, 2022 10:47

MichaelMure merged commit 547d626 into MichaelMure:master May 1, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Complete bug IDs where appropriate #531

Complete bug IDs where appropriate #531

krobelus commented Jan 24, 2021 •

edited

MichaelMure commented Jan 24, 2021

krobelus commented Jan 24, 2021 •

edited

krobelus commented Jan 24, 2021 •

edited

krobelus commented Jan 25, 2021

MichaelMure Jan 25, 2021

krobelus Jan 25, 2021

MichaelMure Jan 25, 2021

MichaelMure Jan 25, 2021

MichaelMure commented Jan 26, 2021

MichaelMure commented Jan 31, 2021

MichaelMure commented Jul 2, 2021

6543 commented Aug 1, 2021

6543 commented Aug 1, 2021

MichaelMure commented Mar 1, 2022

MichaelMure commented Mar 1, 2022

krobelus commented Mar 2, 2022

MichaelMure commented Mar 3, 2022

MichaelMure commented Mar 12, 2022

MichaelMure left a comment

MichaelMure Mar 12, 2022

MichaelMure Mar 12, 2022

MichaelMure Mar 12, 2022

MichaelMure commented Mar 12, 2022

krobelus commented Mar 19, 2022 via email

MichaelMure commented May 1, 2022

Complete bug IDs where appropriate #531

Complete bug IDs where appropriate #531

Conversation

krobelus commented Jan 24, 2021 • edited

MichaelMure commented Jan 24, 2021

krobelus commented Jan 24, 2021 • edited

krobelus commented Jan 24, 2021 • edited

krobelus commented Jan 25, 2021

MichaelMure Jan 25, 2021

Choose a reason for hiding this comment

krobelus Jan 25, 2021

Choose a reason for hiding this comment

MichaelMure Jan 25, 2021

Choose a reason for hiding this comment

MichaelMure Jan 25, 2021

Choose a reason for hiding this comment

MichaelMure commented Jan 26, 2021

MichaelMure commented Jan 31, 2021

MichaelMure commented Jul 2, 2021

6543 commented Aug 1, 2021

6543 commented Aug 1, 2021

MichaelMure commented Mar 1, 2022

MichaelMure commented Mar 1, 2022

krobelus commented Mar 2, 2022

MichaelMure commented Mar 3, 2022

MichaelMure commented Mar 12, 2022

MichaelMure left a comment

Choose a reason for hiding this comment

MichaelMure Mar 12, 2022

Choose a reason for hiding this comment

MichaelMure Mar 12, 2022

Choose a reason for hiding this comment

MichaelMure Mar 12, 2022

Choose a reason for hiding this comment

MichaelMure commented Mar 12, 2022

krobelus commented Mar 19, 2022 via email

MichaelMure commented May 1, 2022

krobelus commented Jan 24, 2021 •

edited

krobelus commented Jan 24, 2021 •

edited

krobelus commented Jan 24, 2021 •

edited