On Jul 11, 2013, at 10:41 AM, Junio C Hamano <gits...@pobox.com> wrote:

> Doug Bell <madcity...@gmail.com> writes:
>> The docs seem to say that doing
>>      git show-ref --head --tags
>> would show both the HEAD ref and all the tag refs. However, doing
>> both --head and either of --tags or --heads would filter out the HEAD
>> ref.
>> Signed-off-by: Doug Bell <madcity...@gmail.com>
>> ---
> I think this patch fell through the cracks, and looking at it, I am
> somewhat torn.
> The command help for "--head" says "show the HEAD reference", which
> may mean:
> (1) in addition to everything else the command would do if there
>     weren't this option specified, show HEAD;
> (2) show the HEAD and nothing else; or
> (3) add HEAD to the candidates to be shown, but apply the usual
>     filtering rule based on --heads, --tags and/or pattern
>     arguments.
> While the last interpretation is what we have used since the
> beginning of the command at 358ddb62 (Add "git show-ref" builtin
> command, 2006-09-15), I tend to agree with you that the first
> interpretation may make more sense, at least from the end user's
> point of view.

> But at a deeper level, it makes the command somewhat inconsistent.
> What happens in the command is
> - We iterate over "candidates to be shown", which is usually
>   "everything under refs/", but with "--head", HEAD is added to
>   this set.  For each of these candidates:
>   - if one or more <pattern> parameters are given, reject the
>     candidate ref if it does not tail-match with any of the
>     patterns;
>   - if either "--heads" or "--tags" is given, among the ones that
>     pass <pattern> filter, check if they:
>     - begin with "refs/heads/" (if "--heads" is given); or
>     - begin with "refs/tags/" (if "--tags" is given).
>     and reject those that don't.
>   - show it if it is still surviving after these two tests.
> And taht is why "git show-ref --tags master v1.3.0" shows only the
> v1.3.0 tag without showing the master branch, and giving "--heads"
> instead of "--tags" yields only the master branch without the tag.
> The semantics your patch wants, by changing the definition of
> "--head" from (3) to (1), is:
> - If "--head" is given, show HEAD no matter what.
> - Iterate over everything under refs/, and for each of them, do the
>   same filter-and-show as we currently do (see above).
> While I think the new semantics is also understandable as the
> current one, and personally I think it is a better behaviour than
> the current one, it will require an update to the document to
> highlight that "--head" is special-cased in a big way, to bypass all
> the filtering that is applied to normal refs.
> A few additional observations (these are not complaints to this
> patch and please do not take them as such):
> - The command help says "(can be combined with heads)" for "--tags"
>   and vice versa, but does not mention their interaction with
>   "--head".  This is because we take interpretation (3) above and
>   do not treat "--head" as a mechanism to add to <pattern>
>   parameter like these two.
> - The command help for "--heads" and "--tags" says "only show
>   heads/tags", which technically does not contradict with "can be
>   combined with" above, but a logical consequence of combining
>   ought to be showing nothing, as a ref cannot be a head (an old
>   nomenclature for a "branch") and a tag at the same time.  
> I think we should find a word better than "only" to use here, but I
> am not sure what would be a good phrase to use.

The reason I had initially wanted both --tags and --head was I wanted to 
compare HEAD against all the tags to see which one(s) I was on (if any). I was 
eventually pointed to `git describe`, but I ended up just using show-ref 
without any options and filtering the result using Perl (the entire application 
is in Perl, so this wasn't a big deal). Then, yeah, I figured it was confusing 
enough to either patch the code or the docs.

For the doc changes, I think if it's explained that by default it show-ref 
shows refs/{tags,heads,remotes}, it becomes easier to explain what the options 
will end up doing. I'll put together a second patch.--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to