[PATCH] Fix zstyle option completion

[PATCH] Fix zstyle option completion

[dana]

I often retrieve zstyle commands from history and then edit them to add -d or
whatever at the beginning. But the completion function won't complete for you
there because of the change made by Daniel in workers/39659.

@Daniel: I think this achieves what you wanted without breaking my use case,
doesn't it? Or did you have something else in mind that i'm not seeing?

dana


diff --git a/Completion/Zsh/Command/_zstyle b/Completion/Zsh/Command/_zstyle
index 75acde5f7..9e82d8ad5 100644
--- a/Completion/Zsh/Command/_zstyle
+++ b/Completion/Zsh/Command/_zstyle
@@ -217,7 +217,7 @@ taglist=(
 )
 
 # Be careful with the context arguments here.  They like to masquerade.
-_arguments -C \
+_arguments -C -A '-*' \
   '(: -)-L[output in form of zstyle commands]:pattern for context patterns:->metapatterns:style:->metastyles' \
   '(: -)-d[delete style definitions]:verbatim context pattern:->patterns:*:styles:->pstyles' \
   '(: -)-e[value is evaluated when style is looked up]:context pattern:->contexts:style:->styles:*:command:_cmdstring' \
@@ -228,7 +228,7 @@ _arguments -C \
   '(: -)-t[test a style, returning false if it'\''s undefined]:context name:->contexts:style:->styles:*:strings to test presence of: ' \
   '(: -)-T[test a style, returning true if it'\''s undefined]:context name:->contexts:style:->styles:*:strings to test presence of: ' \
   '(: -)-m[pattern-match values of a style]:context name:->contexts:style:->styles:pattern: ' \
-  '(-):context pattern:->contexts' '(-):style:->styles' '(-)*:argument:->style-arg'
+  ':context pattern:->contexts' ':style:->styles' '*:argument:->style-arg'
 
 while (( $#state )); do
   case "$state[1]" in

Re: [PATCH] Fix zstyle option completion

[Daniel Shahaf]

dana wrote on Sat, Dec 12, 2020 at 22:15:47 -0600:
> I often retrieve zstyle commands from history and then edit them to add -d or
> whatever at the beginning. But the completion function won't complete for you
> there because of the change made by Daniel in workers/39659.
> 

A regression, then?  Sorry about that.

Could you clarify the precise LBUFFER/RBUFFER values that "won't
complete"?

> @Daniel: I think this achieves what you wanted without breaking my use case,
> doesn't it? Or did you have something else in mind that i'm not seeing?

What I had in mind in that commit was to complete the correct non-option
arguments after the options listed in the log message; e.g., «zstyle -s
<TAB>», «zstyle -s foo <TAB>», «zstyle -s foo bar <TAB>».

The «(-)» exclusions in the line changed were designed to make that line
complete the "set a zstyle" syntax, «zstyle foo bar baz» without any of
the options (not even -e, which is handled separately).

As written, I think the patch will cause «zstyle -g foo bar baz <TAB>»
to offer completions, whereas currently it correctly completes nothing.
(Just observing; not implying any conclusions.)

Cheers,

Daniel
(ENOTIME to test, sorry)

> diff --git a/Completion/Zsh/Command/_zstyle b/Completion/Zsh/Command/_zstyle
> index 75acde5f7..9e82d8ad5 100644
> --- a/Completion/Zsh/Command/_zstyle
> +++ b/Completion/Zsh/Command/_zstyle
> @@ -217,7 +217,7 @@ taglist=(
>  )
>  
>  # Be careful with the context arguments here.  They like to masquerade.
> -_arguments -C \
> +_arguments -C -A '-*' \
>    '(: -)-L[output in form of zstyle commands]:pattern for context patterns:->metapatterns:style:->metastyles' \
>    '(: -)-d[delete style definitions]:verbatim context pattern:->patterns:*:styles:->pstyles' \
>    '(: -)-e[value is evaluated when style is looked up]:context pattern:->contexts:style:->styles:*:command:_cmdstring' \
> @@ -228,7 +228,7 @@ _arguments -C \
>    '(: -)-t[test a style, returning false if it'\''s undefined]:context name:->contexts:style:->styles:*:strings to test presence of: ' \
>    '(: -)-T[test a style, returning true if it'\''s undefined]:context name:->contexts:style:->styles:*:strings to test presence of: ' \
>    '(: -)-m[pattern-match values of a style]:context name:->contexts:style:->styles:pattern: ' \
> -  '(-):context pattern:->contexts' '(-):style:->styles' '(-)*:argument:->style-arg'
> +  ':context pattern:->contexts' ':style:->styles' '*:argument:->style-arg'
>  
>  while (( $#state )); do
>    case "$state[1]" in
> 
> 

Re: [PATCH] Fix zstyle option completion

[dana]

On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> Could you clarify the precise LBUFFER/RBUFFER values that "won't
> complete"?

For example, when i do:

  # LBUFFER='zstyle -'
  # RBUFFER=' foo bar'
  % zstyle -<TAB> foo bar

... it tries (unsuccessfully) to complete argument-1 instead of showing me
options.

On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> As written, I think the patch will cause «zstyle -g foo bar baz <TAB>»
> to offer completions, whereas currently it correctly completes nothing.

I don't think it should. All the options are exclusive with each other, so
that takes care of that already, and the -A '-*' should handle cases like
`zstyle foo bar <TAB>`. It gives me 'no more arguments' when i test your
example with my change. But i might not fully understand what you mean

dana

Re: [PATCH] Fix zstyle option completion

[Lawrence Velázquez]

On Mon, Dec 14, 2020, at 4:55 AM, dana wrote:
> On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> > Could you clarify the precise LBUFFER/RBUFFER values that "won't
> > complete"?
> 
> For example, when i do:
> 
>   # LBUFFER='zstyle -'
>   # RBUFFER=' foo bar'
>   % zstyle -<TAB> foo bar
> 
> ... it tries (unsuccessfully) to complete argument-1 instead of showing me
> options.
> 
> On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> > As written, I think the patch will cause «zstyle -g foo bar baz <TAB>»
> > to offer completions, whereas currently it correctly completes nothing.
> 
> I don't think it should. All the options are exclusive with each other, so
> that takes care of that already, and the -A '-*' should handle cases like
> `zstyle foo bar <TAB>`. It gives me 'no more arguments' when i test your
> example with my change. But i might not fully understand what you mean

Anything further on this?

vq

Re: [PATCH] Fix zstyle option completion

[Daniel Shahaf]

Lawrence Velázquez wrote on Sat, Mar 27, 2021 at 16:08:12 -0400:
> On Mon, Dec 14, 2020, at 4:55 AM, dana wrote:
> > On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> > > Could you clarify the precise LBUFFER/RBUFFER values that "won't
> > > complete"?
> > 
> > For example, when i do:
> > 
> >   # LBUFFER='zstyle -'
> >   # RBUFFER=' foo bar'
> >   % zstyle -<TAB> foo bar
> > 
> > ... it tries (unsuccessfully) to complete argument-1 instead of showing me
> > options.
> > 
> > On 13 Dec 2020, at 11:31, Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
> > > As written, I think the patch will cause «zstyle -g foo bar baz <TAB>»
> > > to offer completions, whereas currently it correctly completes nothing.
> > 
> > I don't think it should. All the options are exclusive with each other, so
> > that takes care of that already, and the -A '-*' should handle cases like
> > `zstyle foo bar <TAB>`. It gives me 'no more arguments' when i test your
> > example with my change. But i might not fully understand what you mean
> 
> Anything further on this?

Pushed.  Also fixed the X-Seq's in ChangeLog while I was at it.  Thank
you both!