Add --expand to 'git notes get-ref'
cover Add --expand to 'git notes get-ref'
From: "W. Trevor King" <[email protected]>
I was recently confused when
$ git notes merge -v refs/remotes/origin/notes/commits
failed to do (or print) anything. It turns out that note refs must live under 'refs/notes/', so my command line ref was expanding to
refs/notes/refs/remotes/origin/notes/commits
which wasn't matching anything.
The first of my commits here adds the '--expand' option, so you can see how Git is expanding your ref internally:
$ GIT_NOTES_REF=commits git notes get-ref --expand refs/notes/commits
The second commit makes the expansion less strict about the location of note refs. In his initial mail introducing 'git notes', Johan says that note refs should live under 'refs/notes' [1]. This seems like a good place for local notes, but note refs from remote repos should probably live somewhere else (e.g. 'refs/remote-notes/' or 'refs/remotes/<name>/notes/'. In the initial thread there are a few messages talking about looking up reverse mappings under 'refs/notes/', but this seems to all have been before the 'refs/notes/<namespace>/' stage. If I'm missing a good reason to keep everything under 'refs/notes/', feel free to ignore the second patch.
Cheers,
Trevor
[1]: http://permalink.gmane.org/gmane.comp.version-control.git/48540
W. Trevor King (2):
notes get-ref: --expand expands the output notes ref.
notes: don't alter refs starting with 'refs/' in expand_notes_ref
Documentation/git-notes.txt | 6 +++++- builtin/notes.c | 26 +++++++++++++++++++++----- notes.c | 2 +- t/t3301-notes.sh | 12 ++++++++++++ 4 files changed, 39 insertions(+), 7 deletions(-)
Signature · 1 line
1.7.12.176.g3fc0e4c.dirty
PATCH 1/2 notes get-ref: --expand expands the output notes ref.
From: "W. Trevor King" <[email protected]>
Useful for debugging refs that don't seem to be expanding correctly.
--- Documentation/git-notes.txt | 6 +++++- builtin/notes.c | 26 +++++++++++++++++++++----- t/t3301-notes.sh | 8 ++++++++ 3 files changed, 34 insertions(+), 6 deletions(-)
Documentation/git-notes.txt +5 -1
@@ -19,7 +19,7 @@ SYNOPSIS 'git notes' merge --abort [-v | -q] 'git notes' remove [--ignore-missing] [--stdin] [<object>...] 'git notes' prune [-n | -v] -'git notes' get-ref +'git notes' get-ref [-e] DESCRIPTION
@@ -165,6 +165,10 @@ OPTIONS input (there is no reason you cannot combine this with object names from the command line). +-e:: +--expand:: + Expand the notes ref before printing it. + -n:: --dry-run:: Do not remove anything; just report the object names whose notes
builtin/notes.c +21 -5
@@ -31,7 +31,7 @@ static const char * const git_notes_usage[] = {"git notes merge --abort [-v | -q]" ,"git notes [--ref <notes_ref>] remove [<object>...]" ,"git notes [--ref <notes_ref>] prune [-n | -v]" , -"git notes [--ref <notes_ref>] get-ref" , +"git notes [--ref <notes_ref>] get-ref [-e]" , NULL };
@@ -84,7 +84,7 @@ static const char * const git_notes_prune_usage[] = { };static const char *const git_notes_get_ref_usage[] = { -"git notes get-ref" , +"git notes get-ref [<options>]" ,NULL };
@@ -1046,16 +1046,32 @@ static int prune(int argc, const char **argv, const char *prefix)static int get_ref (int argc,const char **argv,const char *prefix) { -struct option options[] = {OPT_END () }; +const char *ref; +struct strbuf buf =STRBUF_INIT ; +int expand =0 ; +struct option options[] = { +OPT_BOOL ('e' ,"expand" , &expand, +"expand the notes ref before printing it" ), +OPT_END () + }; argc =parse_options (argc, argv, prefix, options, git_notes_get_ref_usage,0 ); -if (argc) { +if (argc >1 ) {error ("too many parameters" );usage_with_options (git_notes_get_ref_usage, options); } -puts (default_notes_ref ()); + ref =default_notes_ref (); +if (expand) { +strbuf_insert (&buf,0 , ref,strlen (ref)); +expand_notes_ref (&buf); +puts (buf.buf); +strbuf_release (&buf); + }else { +puts (ref); + } +return 0 ; }
t/t3301-notes.sh +8 -0
@@ -1222,4 +1222,12 @@ test_expect_success 'git notes get-ref (--ref)' ' test "$(GIT_NOTES_REF=refs/notes/bar git notes --ref=baz get-ref)" = "refs/notes/baz" ' +test_expect_success 'git notes get-ref (no expand)' ' + test "$(GIT_NOTES_REF=commits git notes get-ref)" = "commits" +' + +test_expect_success 'git notes get-ref (--expand)' ' + test "$(GIT_NOTES_REF=commits git notes get-ref --expand)" = "refs/notes/commits" +' + test_done
Signature · 1 line
1.7.12.176.g3fc0e4c.dirty
Re: [PATCH 1/2] notes get-ref: --expand expands the output notes ref.
Quoted 5 lines · W. Trevor King
From: "W. Trevor King" <[email protected]>
Useful for debugging refs that don't seem to be expanding correctly.
Signed-off-by W. Trevor King <[email protected]>
Signature · 2 lines
Johan Herland, <[email protected]> www.herland.net
Re: [PATCH 1/2] notes get-ref: --expand expands the output notes ref.
Quoted 8 lines · Johan Herland
Quoted 5 lines · W. Trevor King
From: "W. Trevor King" <[email protected]>
Useful for debugging refs that don't seem to be expanding correctly.
Signed-off-by W. Trevor King <[email protected]>Acked-by Johan Herland <[email protected]>
Really? Would "git log --expand master" be useful?
I'd have to say that a DWIM that the user needs to diagnose with such a "feature" is a total failure. Can't we fix that root cause instead?
That is, for a name, e.g. "refs/remotes/origin/notes/commit" (the original example in [PATCH 0/2]) or "commit" (the usual):
- If it is already a ref, just use that (e.g. use that stuff in the refs/remotes/ hierarchy); or
- If it is not, prefix refs/notes/ to it.
or something?
Re: [PATCH 1/2] notes get-ref: --expand expands the output notes ref.
Quoted 1 line · Junio C Hamano
Really? Would "git log --expand master" be useful?
I'm clearly not an expert on this, but isn't that what
git show-ref master
is for? Or is the fact that show-ref returns hashes the more important feature?
There was a lot of "ambiguous refs" discussion in the follow-up for
http://thread.gmane.org/gmane.comp.version-control.git/165885
Subject: [1.8.0] Provide proper remote ref namespaces
For example:
Quoted 6 lines · Jeff King
Will the same apply to refs/heads/foo versus refs/remotes/*/foo? Will it also apply to refs/heads/foo versus refs/remotes/*/tags/foo? In the final case, that does matter to "git push" (should the destination be in the head or tag namespace?). So the actual names of the ref can matter, and should probably be taken into account when deciding what is ambiguous.
So it seems like having a way to figure out how Git is interpreting a shorthand ref would be useful, especially while those of us with less experience are learning to think like Git.
The expansion command doesn't have to be notes-specific, but since the current ref expansion is role specific (e.g. branch/tag/note/…), it seemed like the best place to put it. With more consistent ref interpretation, I'd be less interested in this expansion command ;).
Trevor
Signature · 2 lines
This email may be signed or encrypted with GnuPG (http://www.gnupg.org). For more information, see http://en.wikipedia.org/wiki/Pretty_Good_Privacy
Re: [PATCH 1/2] notes get-ref: --expand expands the output notes ref.
Quoted 8 lines · W. Trevor King
Quoted 1 line · Junio C Hamano
Really? Would "git log --expand master" be useful?
I'm clearly not an expert on this, but isn't that what
git show-ref master
is for?
But then can't you say the same against "git notes get-ref --expand" with "git show-ref refs/remotes/origin/notes/foobla"?
My primary point is that I think it is a UI mistake if the DWIM ignores the user input that directly names that can be interpreted without DWIMming. When the user wants to avoid ambiguity, it should always be possible to spell it out without having to worry about the DWIM code to get in the way.
The problem with the current notes.c:expand_notes_ref() is that it does not allow you to do that, and if you fixed that problem, the user never has to ask "what does this expand to", no?
Perhaps something like this. Note that this only deals with an existing ref, and that is semi-deliberate (because I am not yet convinced that it is a good idea to let any ref outside refs/notes/ to be created to hold a notes tree).
notes.c | 2 ++ 1 file changed, 2 insertions(+)
notes.c +2 -0
@@ -1289,6 +1289,8 @@ int copy_note(struct notes_tree *t,void expand_notes_ref (struct strbuf *sb) { +if (!prefixcmp (sb->buf,"refs/" ) &&ref_exists (sb->buf)) +return ;if (!prefixcmp (sb->buf,"refs/notes/" ))return ;/* we're happy */ else if (!prefixcmp (sb->buf,"notes/" ))
Re: [PATCH 1/2] notes get-ref: --expand expands the output notes ref.
Quoted 19 lines · Junio C Hamano
Quoted 8 lines · W. Trevor King
Quoted 1 line · Junio C Hamano
Really? Would "git log --expand master" be useful?
I'm clearly not an expert on this, but isn't that what
git show-ref master
is for?
But then can't you say the same against "git notes get-ref --expand" with "git show-ref refs/remotes/origin/notes/foobla"?
My primary point is that I think it is a UI mistake if the DWIM ignores the user input that directly names that can be interpreted without DWIMming. When the user wants to avoid ambiguity, it should always be possible to spell it out without having to worry about the DWIM code to get in the way.
Quoted 8 lines
The problem with the current notes.c:expand_notes_ref() is that it does not allow you to do that, and if you fixed that problem, the user never has to ask "what does this expand to", no?
Perhaps something like this. Note that this only deals with an existing ref, and that is semi-deliberate (because I am not yet convinced that it is a good idea to let any ref outside refs/notes/ to be created to hold a notes tree).
(sorry for the late reply; I was away this weekend)
This patch looks like an acceptable solution to me. Especially since it prevents the notes code from "accidentally" creating new notes trees outside refs/notes/*.
That said, we should check for more cases of hardcoded "refs/notes/" that potentially needs changing as well.
...Johan
Signature · 2 lines
Johan Herland, <[email protected]> www.herland.net
PATCH 2/2 notes: don't alter refs starting with 'refs/' in expand_notes_ref
In reply to W. Trevor KingFrom: "W. Trevor King" <[email protected]>
This avoids surprising cases like:
$ GIT_NOTES_REF=refs/remotes/origin/notes/commits git notes get-ref --expand refs/notes/refs/remotes/origin/notes/commits
With the old implementation, all note refs had to live under 'refs/notes/' which was not mentioned in the git-notes.txt.
--- notes.c | 2 +- t/t3301-notes.sh | 4 ++++ 2 files changed, 5 insertions(+), 1 deletion(-)
notes.c +1 -1
@@ -1289,7 +1289,7 @@ int copy_note(struct notes_tree *t,void expand_notes_ref (struct strbuf *sb) { -if (!prefixcmp (sb->buf,"refs/notes/" )) +if (!prefixcmp (sb->buf,"refs/" ))return ;/* we're happy */ else if (!prefixcmp (sb->buf,"notes/" ))strbuf_insert (sb,0 ,"refs/" ,5 );
t/t3301-notes.sh +4 -0
@@ -1230,4 +1230,8 @@ test_expect_success 'git notes get-ref (--expand)' ' test "$(GIT_NOTES_REF=commits git notes get-ref --expand)" = "refs/notes/commits" ' +test_expect_success 'git notes get-ref (--expand)' ' + test "$(GIT_NOTES_REF=refs/remotes/origin/notes/commits git notes get-ref --expand)" = "refs/remotes/origin/notes/commits" +' + test_done
Signature · 1 line
1.7.12.176.g3fc0e4c.dirty
Quoted 10 lines · W. Trevor King
The second commit makes the expansion less strict about the location of note refs. In his initial mail introducing 'git notes', Johan says that note refs should live under 'refs/notes' [1]. This seems like a good place for local notes, but note refs from remote repos should probably live somewhere else (e.g. 'refs/remote-notes/' or 'refs/remotes/<name>/notes/'. In the initial thread there are a few messages talking about looking up reverse mappings under 'refs/notes/', but this seems to all have been before the 'refs/notes/<namespace>/' stage. If I'm missing a good reason to keep everything under 'refs/notes/', feel free to ignore the second patch.
This has been discussed a couple of times on this list, but it never resulted in any actual changes. Read up on this thread to get some context:
http://thread.gmane.org/gmane.comp.version-control.git/160503
...Johan
Signature · 2 lines
Johan Herland, <[email protected]> www.herland.net
Quoted 9 lines · Johan Herland
Quoted 2 lines · W. Trevor King
If I'm missing a good reason to keep everything under 'refs/notes/', feel free to ignore the second patch.
This has been discussed a couple of times on this list, but it never resulted in any actual changes. Read up on this thread to get some context:
http://thread.gmane.org/gmane.comp.version-control.git/160503
Thanks for the pointer, it looks like there are a bunch of good ideas. I assume the lack of changes was due to nobody having the time/inclination to implement
http://article.gmane.org/gmane.comp.version-control.git/160655
Trevor
Signature · 11 lines
As a side note, does anyone know of a thread-specific version of http://download.gmane.org/<xyz>/<start>/<stop> It would be easier for me (and less load for them) if I could run http://download.gmane.org/gmane.comp.version-control.git/thread/160503 -- This email may be signed or encrypted with GnuPG (http://www.gnupg.org). For more information, see http://en.wikipedia.org/wiki/Pretty_Good_Privacy
Quoted 9 lines · W. Trevor King
Quoted 8 lines · Johan Herland
Quoted 2 lines · W. Trevor King
If I'm missing a good reason to keep everything under 'refs/notes/', feel free to ignore the second patch.
This has been discussed a couple of times on this list, but it never resulted in any actual changes. Read up on this thread to get some context:
Quoted 7 lines
Thanks for the pointer, it looks like there are a bunch of good ideas. I assume the lack of changes was due to nobody having the time/inclination to implement
http://article.gmane.org/gmane.comp.version-control.git/160655
Yes, I think time/inclination probably has a lot to do with it. Also, I forgot to link to the largest mailing list thread that discusses these changes in _much_ more detail:
http://thread.gmane.org/gmane.comp.version-control.git/165799/focus=165885
As you can see, the discussion quickly balloons into something much larger than a simple discussion of where remote notes should be located...
That said, your patch 2/2 might be an acceptable stopgap measure for the time being, and then we can reevaluate it if/when we implement a larger ref restructuring.
...Johan
Signature · 2 lines
Johan Herland, <[email protected]> www.herland.net