parse-options: don't complete option aliases by default

From: Philippe Blain <[email protected]>

Since 'OPT_ALIAS' was created in 5c387428f1 (parse-options: don't emit "ambiguous option" for aliases, 2019-04-29), 'git clone --git-completion-helper', which is used by the Bash completion script to list options accepted by clone (via '__gitcomp_builtin'), lists both '--recurse-submodules' and its alias '--recursive', which was not the case before since '--recursive' had the PARSE_OPT_HIDDEN flag set, and options with this flag are skipped by 'parse-options.c::show_gitcomp', which implements 'git <cmd> --git-completion-helper'.

At the point where 'show_gitcomp' is called in 'parse_options_step', 'preprocess_options' was already called in 'parse_options', so any aliases are now copies of the original options with a modified help text indicating they are aliases.

Helpfully, since 64cc539fd2 (parse-options: don't leak alias help messages, 2021-03-21) these copies have the PARSE_OPT_FROM_ALIAS flag set, so check that flag early in 'show_gitcomp' and do not print them, unless the user explicitely requested that *all* completion be shown (by setting 'GIT_COMPLETION_SHOW_ALL'). After all, if we want to encourage the use of '--recurse-submodules' over '--recursive', we'd better just suggest the former.

The only other options alias is 'log' and friends' '--mailmap', which is an alias for '--use-mailmap', but the Bash completion helpers for these commands do not use '__gitcomp_builtin', and thus are unnaffected by this change.

Test the new behaviour in t9902-completion.sh. As a side effect, this also tests the correct behaviour of GIT_COMPLETION_SHOW_ALL, which was not tested before. Note that since '__gitcomp_builtin' caches the options it shows, we need to re-source the completion script to clear that cache for the second test.

Signed-off-by Philippe Blain <[email protected]>

---
parse-options: don't complete option aliases by default

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-996%2Fphil-blain%2Fclone-recurse-completion-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-996/phil-blain/clone-recurse-completion-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/996

 parse-options.c       |  2 +-
 t/t9902-completion.sh | 13 +++++++++++++
 2 files changed, 14 insertions(+), 1 deletion(-)

parse-options.c +1 -1

index e6f56768ca5..2abff136a17 100644
@@ -585,7 +585,7 @@ static int show_gitcomp(const struct option *opts, int show_all)
 		if (!opts->long_name)
 			continue;
 		if (!show_all &&
-			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE)))
+			(opts->flags & (PARSE_OPT_HIDDEN | PARSE_OPT_NOCOMPLETE | PARSE_OPT_FROM_ALIAS)))
 			continue;
 
 		switch (opts->type) {

t/t9902-completion.sh +0 -0

index cb057ef1613..11573936d58 100755

On Thu, Jul 15 2021, Philippe Blain via GitGitGadget wrote:

I'm a bit biased here since I like --recursive better, but let's not drag that whole argument up again. I don't find the argument in bb62e0a99fc (clone: teach --recurse-submodules to optionally take a pathspec, 2017-03-17) convincing enough to have moved such a prominent use-case to a longer option name.

But, water under the bridge. Aside from that:

For me a Google search for "git clone --recursive" is ~40k results, but "git clone --recurse-submodules". The former links to various discussions/docs/stackoverflow answers, often --recurse-submodules isn't mentioned at all or as an aside.

I think it's unfortunate that we:

1. Conflate whether something shows up in completion v.s. the help. Given its prominence and that "git clone -h" is ~50 lines why not note --recursive there.

2. Don't have the completion aware of these aliases, i.e. "git clone --rec<TAB>" before your chance offers a completion of both, that sucks, we should fully complete the non-alias instead.

3. Making it PARSE_OPT_HIDDEN "solves" #2 at the cost of hiding it in "git help -h", and now this won't work, but did before:

git clone --recursi<TAB>

I.e. even if we didn't want to do #2 *and* wanted to hide it in the usage output surely completing an unmbigous prefix is better, even if it's a hidden option?

E.g. the user has used --recursive for years, but can't even find it in -h (I also think it's a mistake to have entirely removed it from the docs, even if one agrees with its "deprecation" I'd say we should keep some "used to be called --recursive" note there).

I'm a bit biased here since I like --recursive better, but let's not drag that whole argument up again. I don't find the argument in bb62e0a99fc (clone: teach --recurse-submodules to optionally take a pathspec, 2017-03-17) convincing enough to have moved such a prominent use-case to a longer option name.

Quoted 5 lines · Ævar Arnfjörð Bjarmason

I'm a bit biased here since I like --recursive better, but let's not drag that whole argument up again. I don't find the argument in bb62e0a99fc (clone: teach --recurse-submodules to optionally take a pathspec, 2017-03-17) convincing enough to have moved such a prominent use-case to a longer option name.

I do agree that it is a separate topic to pick one or the other as the primary. I think this change means that the users no longer need to stop after hitting a tab:

     clone --recurs<TAB>

and instead get to "clone --recurse-submodules" right away, which is a positigve change.

the canonical and which is the alias, the same user would get the new canonical one right away from the same state.

So it does look like a good change to me.

I'm a bit biased here since I like --recursive better, but let's not drag that whole argument up again. I don't find the argument in bb62e0a99fc (clone: teach --recurse-submodules to optionally take a pathspec, 2017-03-17) convincing enough to have moved such a prominent use-case to a longer option name.

But, water under the bridge. Aside from that:

For me a Google search for "git clone --recursive" is ~40k results, but "git clone --recurse-submodules". The former links to various discussions/docs/stackoverflow answers, often --recurse-submodules isn't mentioned at all or as an aside.

I think it's unfortunate that we:

1. Conflate whether something shows up in completion v.s. the help. Given its prominence and that "git clone -h" is ~50 lines why not note --recursive there.

2. Don't have the completion aware of these aliases, i.e. "git clone --rec<TAB>" before your chance offers a completion of both, that sucks, we should fully complete the non-alias instead.

3. Making it PARSE_OPT_HIDDEN "solves" #2 at the cost of hiding it in "git help -h", and now this won't work, but did before:

git clone --recursi<TAB>

I.e. even if we didn't want to do #2 *and* wanted to hide it in the usage output surely completing an unmbigous prefix is better, even if it's a hidden option?

Per-se none of this is a blocker or "we must fix this first" for this particular change, we have this in many existing cases.

I daresay there's no other alias that's in as wide a use in the wild, so we should think about this one particularly carefully though.

It's not fully clear from your commit message which of 1-3 you're aiming for, I think it's more of the "discourage its use".

Sure, fair enough, but PARSE_OPT_HIDDEN is not a 1=1 mapping to that, and can often lead to more user confusion.

E.g. the user has used --recursive for years, but can't even find it in -h (I also think it's a mistake to have entirely removed it from the docs, even if one agrees with its "deprecation" I'd say we should keep some "used to be called --recursive" note there).

From: Philippe Blain <[email protected]>

Since 'OPT_ALIAS' was created in 5c387428f1 (parse-options: don't emit "ambiguous option" for aliases, 2019-04-29), 'git clone --git-completion-helper', which is used by the Bash completion script to list options accepted by clone (via '__gitcomp_builtin'), lists both '--recurse-submodules' and its alias '--recursive', which was not the case before since '--recursive' had the PARSE_OPT_HIDDEN flag set, and options with this flag are skipped by 'parse-options.c::show_gitcomp', which implements 'git <cmd> --git-completion-helper'.

At the point where 'show_gitcomp' is called in 'parse_options_step', 'preprocess_options' was already called in 'parse_options', so any aliases are now copies of the original options with a modified help text indicating they are aliases.

Helpfully, since 64cc539fd2 (parse-options: don't leak alias help messages, 2021-03-21) these copies have the PARSE_OPT_FROM_ALIAS flag set, so check that flag early in 'show_gitcomp' and do not print them,

unless the user explicitely requested that *all* completion be shown (by setting 'GIT_COMPLETION_SHOW_ALL'). After all, if we want to encourage the use of '--recurse-submodules' over '--recursive', we'd better just suggest the former.

The only other options alias is 'log' and friends' '--mailmap', which is an alias for '--use-mailmap', but the Bash completion helpers for these commands do not use '__gitcomp_builtin', and thus are unnaffected by this change.

Test the new behaviour in t9902-completion.sh. As a side effect, this also tests the correct behaviour of GIT_COMPLETION_SHOW_ALL, which was not tested before. Note that since '__gitcomp_builtin' caches the options it shows, we need to re-source the completion script to clear that cache for the second test.

Quoted 6 lines · Philippe Blain via GitGitGadget

From: Philippe Blain <[email protected]>

Helpfully, since 64cc539fd2 (parse-options: don't leak alias help messages, 2021-03-21) these copies have the PARSE_OPT_FROM_ALIAS flag set, so check that flag early in 'show_gitcomp' and do not print them,

Makes sense but I don't think what the patch is doing should be buried here. I think a separate paragraph explaining what you are trying to do will be better.

I agree this is better, and the patch itself looks obviously correct.

Reviewed-by Felipe Contreras <[email protected]>