git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [RFC PATCH 0/2] Support enabling bash completion of all options
@ 2020-08-19 17:50 Ryan Zoeller
  2020-08-19 17:51 ` [RFC PATCH 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
                   ` (4 more replies)
  0 siblings, 5 replies; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 17:50 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller

This series adds --git-completion-helper-all, which behaves similarly to
--git-completion-helper but includes options which are typically hidden.
An environment variable GIT_COMPLETION_SHOW_ALL is used to opt-in to
this behavior.

I often prototype automated tooling around git by running commands locally
until I'm satisfied with the behavior, then migrating them into a script.
Having bash completion for some of the more obscure options is useful.

I wasn't sure how to add automated tests around this without making the
tests overly brittle (e.g. depending on a verbatim list of options), so
I welcome ideas and feedback.

-Ryan

Ryan Zoeller (2):
  parse-options: add --git-completion-helper-all
  completion: add GIT_COMPLETION_SHOW_ALL env var

 contrib/completion/git-completion.bash | 14 +++++++++++++-
 parse-options.c                        | 18 ++++++++++++------
 2 files changed, 25 insertions(+), 7 deletions(-)

-- 
2.28.0.260.g5fadab5a9c



^ permalink raw reply	[flat|nested] 9+ messages in thread

* [RFC PATCH 1/2] parse-options: add --git-completion-helper-all
  2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
@ 2020-08-19 17:51 ` Ryan Zoeller
  2020-08-19 19:39   ` Junio C Hamano
  2020-08-19 17:51 ` [RFC PATCH 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller
                   ` (3 subsequent siblings)
  4 siblings, 1 reply; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 17:51 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller

--git-completion-helper excludes hidden options, such as --allow-empty
for git commit. This is typically helpful, but occasionally we want
auto-completion for obscure flags. --git-completion-helper-all returns
all options, even if they are marked as hidden or nocomplete.

Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
---
 parse-options.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/parse-options.c b/parse-options.c
index c57618d537..cc7239e1c6 100644
--- a/parse-options.c
+++ b/parse-options.c
@@ -535,8 +535,9 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
 
 		if (!opts->long_name)
 			continue;
-		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
-			continue;
+		/* Don't check PARSE_OPT_HIDDEN or PARSE_OPT_NOCOMPLETE,
+		 * we expect the caller to handle these appropriately.
+		 */
 		if (opts->flags & PARSE_OPT_NONEG)
 			continue;
 
@@ -572,7 +573,7 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
 	}
 }
 
-static int show_gitcomp(const struct option *opts)
+static int show_gitcomp(const struct option *opts, int show_all)
 {
 	const struct option *original_opts = opts;
 	int nr_noopts = 0;
@@ -582,7 +583,8 @@ static int show_gitcomp(const struct option *opts)
 
 		if (!opts->long_name)
 			continue;
-		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
+		if (!show_all &&
+			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
 			continue;
 
 		switch (opts->type) {
@@ -723,9 +725,13 @@ int parse_options_step(struct parse_opt_ctx_t *ctx,
 		if (internal_help && ctx->total == 1 && !strcmp(arg + 1, "h"))
 			goto show_usage;
 
-		/* lone --git-completion-helper is asked by git-completion.bash */
+		/* lone --git-completion-helper and --git-completion-helper-all
+		 * are asked by git-completion.bash
+		 */
 		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper"))
-			return show_gitcomp(options);
+			return show_gitcomp(options, 0);
+		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper-all"))
+			return show_gitcomp(options, 1);
 
 		if (arg[1] != '-') {
 			ctx->opt = arg + 1;
-- 
2.28.0.260.g5fadab5a9c



^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [RFC PATCH 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var
  2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
  2020-08-19 17:51 ` [RFC PATCH 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
@ 2020-08-19 17:51 ` Ryan Zoeller
  2020-08-19 23:06 ` [PATCH v2 0/2] Support enabling bash completion of all options Ryan Zoeller
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 17:51 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller

When set to 1, GIT_COMPLETION_SHOW_ALL causes --git-completion-helper-all
to be passed instead of --git-completion-helper.

Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
---
 contrib/completion/git-completion.bash | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/contrib/completion/git-completion.bash b/contrib/completion/git-completion.bash
index 700d44af5b..9147fba3d5 100644
--- a/contrib/completion/git-completion.bash
+++ b/contrib/completion/git-completion.bash
@@ -39,6 +39,11 @@
 #     When set to "1", do not include "DWIM" suggestions in git-checkout
 #     and git-switch completion (e.g., completing "foo" when "origin/foo"
 #     exists).
+#
+#   GIT_COMPLETION_SHOW_ALL
+#
+#     When set to "1" suggest all options, including options which are
+#     typically hidden (e.g. '--allow-empty' for 'git commit').
 
 case "$COMP_WORDBREAKS" in
 *:*) : great ;;
@@ -411,10 +416,17 @@ __gitcomp_builtin ()
 	local options
 	eval "options=\${$var-}"
 
+	local completion_helper
+	if [ "$GIT_COMPLETION_SHOW_ALL" = "1" ]; then
+		completion_helper="--git-completion-helper-all"
+	else
+		completion_helper="--git-completion-helper"
+	fi
+
 	if [ -z "$options" ]; then
 		# leading and trailing spaces are significant to make
 		# option removal work correctly.
-		options=" $incl $(__git ${cmd/_/ } --git-completion-helper) " || return
+		options=" $incl $(__git ${cmd/_/ } $completion_helper) " || return
 
 		for i in $excl; do
 			options="${options/ $i / }"
-- 
2.28.0.260.g5fadab5a9c



^ permalink raw reply related	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 1/2] parse-options: add --git-completion-helper-all
  2020-08-19 17:51 ` [RFC PATCH 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
@ 2020-08-19 19:39   ` Junio C Hamano
  2020-08-19 20:51     ` Ryan Zoeller
  0 siblings, 1 reply; 9+ messages in thread
From: Junio C Hamano @ 2020-08-19 19:39 UTC (permalink / raw)
  To: Ryan Zoeller; +Cc: git

Ryan Zoeller <rtzoeller@rtzoeller.com> writes:

> --git-completion-helper excludes hidden options, such as --allow-empty
> for git commit. This is typically helpful, but occasionally we want
> auto-completion for obscure flags.

Hits from "git grep -B2 OPT_NOCOMPLETE" tells me that these are
mostly unsafe options.  Those who accept the risk by saying
"complete all" should be allowed to see them.

The same with OPT_HIDDEN (including OPT_HIDDEN_<TYPE>) gives us a
mixed bag.  Many are unsafe, some are uncommon and the rest are
discouraged, or old synonym to some other option that does get
completed.  I am not sure if letting them be completed is an overall
win or makes the output from "git cmd --<TAB><TAB>" too noisy.

> --git-completion-helper-all returns
> all options, even if they are marked as hidden or nocomplete.

If it is "occasinally", why is the removal of OPT_HIDDEN in
show_negated_gitcomp() unconditional?

> Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
> ---
>  parse-options.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/parse-options.c b/parse-options.c
> index c57618d537..cc7239e1c6 100644
> --- a/parse-options.c
> +++ b/parse-options.c
> @@ -535,8 +535,9 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
>  
>  		if (!opts->long_name)
>  			continue;
> -		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
> -			continue;
> +		/* Don't check PARSE_OPT_HIDDEN or PARSE_OPT_NOCOMPLETE,
> +		 * we expect the caller to handle these appropriately.
> +		 */

	/*
	 * Style: our multi-line comments should begin with
	 * slash-asterisk alone on its own line, and end with
	 * asterisk-slash also on its own line, like this.
	 */

We do not run around and fix existing style violations that would
only raise the patch noise, but we try not to introduce new
violators.

I am not sure what the new comment says is justifiable.  The only
caller of show_negated_gitcomp() is in show_gitcomp() where the
function looped over the options and showed the options normally,
honoring these two flags, but then the original list of options
are passed to this function without filtering any option element
on the list that are marked with these bits, so "the caller"
apparently is not interested in handling the elements with these
bits when making the decision to show "--no-<option>" form itself;
it farms it out to show_negated_gitcomp() and expects the function
to do "the right thing".  Am I misleading the code?


>  		if (opts->flags & PARSE_OPT_NONEG)
>  			continue;
>  
> @@ -572,7 +573,7 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
>  	}
>  }
>  
> -static int show_gitcomp(const struct option *opts)
> +static int show_gitcomp(const struct option *opts, int show_all)
>  {
>  	const struct option *original_opts = opts;
>  	int nr_noopts = 0;
> @@ -582,7 +583,8 @@ static int show_gitcomp(const struct option *opts)
>  
>  		if (!opts->long_name)
>  			continue;
> -		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
> +		if (!show_all &&
> +			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
>  			continue;
>  
>  		switch (opts->type) {
> @@ -723,9 +725,13 @@ int parse_options_step(struct parse_opt_ctx_t *ctx,
>  		if (internal_help && ctx->total == 1 && !strcmp(arg + 1, "h"))
>  			goto show_usage;
>  
> -		/* lone --git-completion-helper is asked by git-completion.bash */
> +		/* lone --git-completion-helper and --git-completion-helper-all
> +		 * are asked by git-completion.bash
> +		 */

Ditto.

>  		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper"))
> -			return show_gitcomp(options);
> +			return show_gitcomp(options, 0);
> +		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper-all"))
> +			return show_gitcomp(options, 1);

This is not your fault, but the micro-optimization to avoid
comparison between *arg (which already is known to be "-") and "-"
is not worth the reduced readability.  

		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper"))
			return show_gitcomp(options, 0);
		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper-all"))
			return show_gitcomp(options, 1);

would be much clearer for readers to know what is going on.

With an extra "const char *rest" variable in the same scope as
"const char *arg",

		if (ctx->total == 1 && 
		    !skip_prefix(arg, "--git-completion-helper", &rest) && 
		    (!*rest || !strcmp(rest, "-all")))
			return show_gitcomp(options, *rest);

would further avoid repetitions, but some folks may find it a bit
too dense.  I dunno.

>  
>  		if (arg[1] != '-') {
>  			ctx->opt = arg + 1;

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 1/2] parse-options: add --git-completion-helper-all
  2020-08-19 19:39   ` Junio C Hamano
@ 2020-08-19 20:51     ` Ryan Zoeller
  2020-08-19 21:05       ` Junio C Hamano
  0 siblings, 1 reply; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 20:51 UTC (permalink / raw)
  To: Junio C Hamano; +Cc: git

On 8/19/20 2:39 PM, Junio C Hamano wrote:
> 
> Ryan Zoeller <rtzoeller@rtzoeller.com> writes:
> 
>> --git-completion-helper excludes hidden options, such as --allow-empty
>> for git commit. This is typically helpful, but occasionally we want
>> auto-completion for obscure flags.
> 
> Hits from "git grep -B2 OPT_NOCOMPLETE" tells me that these are
> mostly unsafe options.  Those who accept the risk by saying
> "complete all" should be allowed to see them.
> 
> The same with OPT_HIDDEN (including OPT_HIDDEN_<TYPE>) gives us a
> mixed bag.  Many are unsafe, some are uncommon and the rest are
> discouraged, or old synonym to some other option that does get
> completed.  I am not sure if letting them be completed is an overall
> win or makes the output from "git cmd --<TAB><TAB>" too noisy.

If options marked OPT_HIDDEN are considered too internal to
meaningfully expose, I'm happy to hide them. I defaulted to
"show everything", and backing off from that is easy enough.

> 
>> --git-completion-helper-all returns
>> all options, even if they are marked as hidden or nocomplete.
> 
> If it is "occasinally", why is the removal of OPT_HIDDEN in
> show_negated_gitcomp() unconditional?

It shouldn't have been. I visually clumped the calls to it
as being inside the for loop, and assumed they were being skipped
over as part of the continue.

> 
>> Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
>> ---
>>   parse-options.c | 18 ++++++++++++------
>>   1 file changed, 12 insertions(+), 6 deletions(-)
>>
>> diff --git a/parse-options.c b/parse-options.c
>> index c57618d537..cc7239e1c6 100644
>> --- a/parse-options.c
>> +++ b/parse-options.c
>> @@ -535,8 +535,9 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
>>
>>   		if (!opts->long_name)
>>   			continue;
>> -		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
>> -			continue;
>> +		/* Don't check PARSE_OPT_HIDDEN or PARSE_OPT_NOCOMPLETE,
>> +		 * we expect the caller to handle these appropriately.
>> +		 */
> 
> 	/*
> 	 * Style: our multi-line comments should begin with
> 	 * slash-asterisk alone on its own line, and end with
> 	 * asterisk-slash also on its own line, like this.
> 	 */
> 
> We do not run around and fix existing style violations that would
> only raise the patch noise, but we try not to introduce new
> violators.

Will fix.

> 
> I am not sure what the new comment says is justifiable.  The only
> caller of show_negated_gitcomp() is in show_gitcomp() where the
> function looped over the options and showed the options normally,
> honoring these two flags, but then the original list of options
> are passed to this function without filtering any option element
> on the list that are marked with these bits, so "the caller"
> apparently is not interested in handling the elements with these
> bits when making the decision to show "--no-<option>" form itself;
> it farms it out to show_negated_gitcomp() and expects the function
> to do "the right thing".  Am I misleading the code?
> 
> 

You're not misreading it; I apparently neglected to test the completion
for '--no-' options with '--git-completion-helper', only
'--git-completion-helper-all'. I'll apply the same show_all logic
to this function.

>>   		if (opts->flags & PARSE_OPT_NONEG)
>>   			continue;
>>
>> @@ -572,7 +573,7 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
>>   	}
>>   }
>>
>> -static int show_gitcomp(const struct option *opts)
>> +static int show_gitcomp(const struct option *opts, int show_all)
>>   {
>>   	const struct option *original_opts = opts;
>>   	int nr_noopts = 0;
>> @@ -582,7 +583,8 @@ static int show_gitcomp(const struct option *opts)
>>
>>   		if (!opts->long_name)
>>   			continue;
>> -		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
>> +		if (!show_all &&
>> +			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
>>   			continue;
>>
>>   		switch (opts->type) {
>> @@ -723,9 +725,13 @@ int parse_options_step(struct parse_opt_ctx_t *ctx,
>>   		if (internal_help && ctx->total == 1 && !strcmp(arg + 1, "h"))
>>   			goto show_usage;
>>
>> -		/* lone --git-completion-helper is asked by git-completion.bash */
>> +		/* lone --git-completion-helper and --git-completion-helper-all
>> +		 * are asked by git-completion.bash
>> +		 */
> 
> Ditto.
> 

Will fix.

>>   		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper"))
>> -			return show_gitcomp(options);
>> +			return show_gitcomp(options, 0);
>> +		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper-all"))
>> +			return show_gitcomp(options, 1);
> 
> This is not your fault, but the micro-optimization to avoid
> comparison between *arg (which already is known to be "-") and "-"
> is not worth the reduced readability.
> 
> 		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper"))
> 			return show_gitcomp(options, 0);
> 		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper-all"))
> 			return show_gitcomp(options, 1);
> 
> would be much clearer for readers to know what is going on.
> 

I completely agree, and will clean these up.

> With an extra "const char *rest" variable in the same scope as
> "const char *arg",
> 
> 		if (ctx->total == 1 &&
> 		    !skip_prefix(arg, "--git-completion-helper", &rest) &&
> 		    (!*rest || !strcmp(rest, "-all")))
> 			return show_gitcomp(options, *rest);
> 
> would further avoid repetitions, but some folks may find it a bit
> too dense.  I dunno.

I'm inclined to be repetitive in order to keep
'--git-completion-helper-all' intact, e.g. for grepping.

> 
>>
>>   		if (arg[1] != '-') {
>>   			ctx->opt = arg + 1;


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 1/2] parse-options: add --git-completion-helper-all
  2020-08-19 20:51     ` Ryan Zoeller
@ 2020-08-19 21:05       ` Junio C Hamano
  0 siblings, 0 replies; 9+ messages in thread
From: Junio C Hamano @ 2020-08-19 21:05 UTC (permalink / raw)
  To: Ryan Zoeller; +Cc: git

Ryan Zoeller <rtzoeller@rtzoeller.com> writes:

> On 8/19/20 2:39 PM, Junio C Hamano wrote:
>> 
>> Ryan Zoeller <rtzoeller@rtzoeller.com> writes:
>> 
>>> --git-completion-helper excludes hidden options, such as --allow-empty
>>> for git commit. This is typically helpful, but occasionally we want
>>> auto-completion for obscure flags.
>> 
>> Hits from "git grep -B2 OPT_NOCOMPLETE" tells me that these are
>> mostly unsafe options.  Those who accept the risk by saying
>> "complete all" should be allowed to see them.
>> 
>> The same with OPT_HIDDEN (including OPT_HIDDEN_<TYPE>) gives us a
>> mixed bag.  Many are unsafe, some are uncommon and the rest are
>> discouraged, or old synonym to some other option that does get
>> completed.  I am not sure if letting them be completed is an overall
>> win or makes the output from "git cmd --<TAB><TAB>" too noisy.
>
> If options marked OPT_HIDDEN are considered too internal to
> meaningfully expose, I'm happy to hide them. I defaulted to
> "show everything", and backing off from that is easy enough.

I think with the current state of HIDDEN which is applied unevenly
and for different purposes, the only sensible thing to do for the
"complete-all" operation is to show them.  If we audit all uses of
hidden and allocate different bit for the reason why they are marked
to be hidden, it may become possible for "complete-all" to be more
intelligent about them (e.g. not showing old synonyms that exist
only for helping muscle memory, while showing others), but before
that, I do not think it makes sense to hide them all.

> You're not misreading it; I apparently neglected to test the completion
> for '--no-' options with '--git-completion-helper', only
> '--git-completion-helper-all'. I'll apply the same show_all logic
> to this function.

That makes sense.

> I'm inclined to be repetitive in order to keep
> '--git-completion-helper-all' intact, e.g. for grepping.

Good reasoning.

Thanks.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v2 0/2] Support enabling bash completion of all options
  2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
  2020-08-19 17:51 ` [RFC PATCH 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
  2020-08-19 17:51 ` [RFC PATCH 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller
@ 2020-08-19 23:06 ` Ryan Zoeller
  2020-08-19 23:06 ` [PATCH v2 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
  2020-08-19 23:06 ` [PATCH v2 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller
  4 siblings, 0 replies; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 23:06 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller, Junio C Hamano

Fix style guide violations and issue with '--no-' options.

Ryan Zoeller (2):
  parse-options: add --git-completion-helper-all
  completion: add GIT_COMPLETION_SHOW_ALL env var

 contrib/completion/git-completion.bash | 14 +++++++++++++-
 parse-options.c                        | 26 +++++++++++++++++---------
 2 files changed, 30 insertions(+), 10 deletions(-)

-- 
2.28.0.260.g5934a15c94



^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v2 1/2] parse-options: add --git-completion-helper-all
  2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
                   ` (2 preceding siblings ...)
  2020-08-19 23:06 ` [PATCH v2 0/2] Support enabling bash completion of all options Ryan Zoeller
@ 2020-08-19 23:06 ` Ryan Zoeller
  2020-08-19 23:06 ` [PATCH v2 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller
  4 siblings, 0 replies; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 23:06 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller, Junio C Hamano

--git-completion-helper excludes hidden options, such as --allow-empty
for git commit. This is typically helpful, but occasionally we want
auto-completion for obscure flags. --git-completion-helper-all returns
all options, even if they are marked as hidden or nocomplete.

Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
---
 parse-options.c | 26 +++++++++++++++++---------
 1 file changed, 17 insertions(+), 9 deletions(-)

diff --git a/parse-options.c b/parse-options.c
index c57618d537..f0507432ee 100644
--- a/parse-options.c
+++ b/parse-options.c
@@ -525,7 +525,8 @@ void parse_options_start(struct parse_opt_ctx_t *ctx,
 	parse_options_start_1(ctx, argc, argv, prefix, options, flags);
 }
 
-static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
+static void show_negated_gitcomp(const struct option *opts, int show_all,
+				 int nr_noopts)
 {
 	int printed_dashdash = 0;
 
@@ -535,7 +536,8 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
 
 		if (!opts->long_name)
 			continue;
-		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
+		if (!show_all &&
+			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
 			continue;
 		if (opts->flags & PARSE_OPT_NONEG)
 			continue;
@@ -572,7 +574,7 @@ static void show_negated_gitcomp(const struct option *opts, int nr_noopts)
 	}
 }
 
-static int show_gitcomp(const struct option *opts)
+static int show_gitcomp(const struct option *opts, int show_all)
 {
 	const struct option *original_opts = opts;
 	int nr_noopts = 0;
@@ -582,7 +584,8 @@ static int show_gitcomp(const struct option *opts)
 
 		if (!opts->long_name)
 			continue;
-		if (opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE))
+		if (!show_all &&
+			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
 			continue;
 
 		switch (opts->type) {
@@ -610,8 +613,8 @@ static int show_gitcomp(const struct option *opts)
 			nr_noopts++;
 		printf(" --%s%s", opts->long_name, suffix);
 	}
-	show_negated_gitcomp(original_opts, -1);
-	show_negated_gitcomp(original_opts, nr_noopts);
+	show_negated_gitcomp(original_opts, show_all, -1);
+	show_negated_gitcomp(original_opts, show_all, nr_noopts);
 	fputc('\n', stdout);
 	return PARSE_OPT_COMPLETE;
 }
@@ -723,9 +726,14 @@ int parse_options_step(struct parse_opt_ctx_t *ctx,
 		if (internal_help && ctx->total == 1 && !strcmp(arg + 1, "h"))
 			goto show_usage;
 
-		/* lone --git-completion-helper is asked by git-completion.bash */
-		if (ctx->total == 1 && !strcmp(arg + 1, "-git-completion-helper"))
-			return show_gitcomp(options);
+		/*
+		 * lone --git-completion-helper and --git-completion-helper-all
+		 * are asked by git-completion.bash
+		 */
+		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper"))
+			return show_gitcomp(options, 0);
+		if (ctx->total == 1 && !strcmp(arg, "--git-completion-helper-all"))
+			return show_gitcomp(options, 1);
 
 		if (arg[1] != '-') {
 			ctx->opt = arg + 1;
-- 
2.28.0.260.g5934a15c94



^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v2 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var
  2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
                   ` (3 preceding siblings ...)
  2020-08-19 23:06 ` [PATCH v2 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
@ 2020-08-19 23:06 ` Ryan Zoeller
  4 siblings, 0 replies; 9+ messages in thread
From: Ryan Zoeller @ 2020-08-19 23:06 UTC (permalink / raw)
  To: git; +Cc: Ryan Zoeller, Junio C Hamano

When set to 1, GIT_COMPLETION_SHOW_ALL causes --git-completion-helper-all
to be passed instead of --git-completion-helper.

Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
---
 contrib/completion/git-completion.bash | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/contrib/completion/git-completion.bash b/contrib/completion/git-completion.bash
index 700d44af5b..9147fba3d5 100644
--- a/contrib/completion/git-completion.bash
+++ b/contrib/completion/git-completion.bash
@@ -39,6 +39,11 @@
 #     When set to "1", do not include "DWIM" suggestions in git-checkout
 #     and git-switch completion (e.g., completing "foo" when "origin/foo"
 #     exists).
+#
+#   GIT_COMPLETION_SHOW_ALL
+#
+#     When set to "1" suggest all options, including options which are
+#     typically hidden (e.g. '--allow-empty' for 'git commit').
 
 case "$COMP_WORDBREAKS" in
 *:*) : great ;;
@@ -411,10 +416,17 @@ __gitcomp_builtin ()
 	local options
 	eval "options=\${$var-}"
 
+	local completion_helper
+	if [ "$GIT_COMPLETION_SHOW_ALL" = "1" ]; then
+		completion_helper="--git-completion-helper-all"
+	else
+		completion_helper="--git-completion-helper"
+	fi
+
 	if [ -z "$options" ]; then
 		# leading and trailing spaces are significant to make
 		# option removal work correctly.
-		options=" $incl $(__git ${cmd/_/ } --git-completion-helper) " || return
+		options=" $incl $(__git ${cmd/_/ } $completion_helper) " || return
 
 		for i in $excl; do
 			options="${options/ $i / }"
-- 
2.28.0.260.g5934a15c94



^ permalink raw reply related	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2020-08-19 23:06 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-19 17:50 [RFC PATCH 0/2] Support enabling bash completion of all options Ryan Zoeller
2020-08-19 17:51 ` [RFC PATCH 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
2020-08-19 19:39   ` Junio C Hamano
2020-08-19 20:51     ` Ryan Zoeller
2020-08-19 21:05       ` Junio C Hamano
2020-08-19 17:51 ` [RFC PATCH 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller
2020-08-19 23:06 ` [PATCH v2 0/2] Support enabling bash completion of all options Ryan Zoeller
2020-08-19 23:06 ` [PATCH v2 1/2] parse-options: add --git-completion-helper-all Ryan Zoeller
2020-08-19 23:06 ` [PATCH v2 2/2] completion: add GIT_COMPLETION_SHOW_ALL env var Ryan Zoeller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).