On Sat, Nov 17, 2018 at 11:07:32PM +0900, Junio C Hamano wrote: > Jeff King <peff@xxxxxxxx> writes: > > > I suspect it would be less confusing if the rewrite were inverted, like: > > > > [url "gh:"] > > rewriteTo = https://github.com > > rewritePrivate > > > > [url "git://github.com"] > > rewriteTo = https://github.com > > > > where the mapping of sections to rewrite rules must be one-to-one, not > > one-to-many (and you can see that the flip side is that we have to > > repeat ourselves). > > > > I hate to introduce two ways of doing the same thing, but maybe it is > > simple enough to explain that url.X.insteadOf=Y is identical to > > url.Y.rewriteTo=X. I do think people have gotten confused by the > > ordering of insteadOf over the years, so this would let them specify it > > in whichever way makes the most sense to them. > > I do admit that the current "insteadOf" was too cute a way to > configure this feature and I often have to read it aloud twice > before convincing myself I got X and Y right. It would have been > cleaner if the definition were in the opposite direction, exactly > because you can rewrite a single thing into just one way, but we do > want to support that many things mapping to the same, and the > approach to use "url.Y.rewriteTo=X" does express it better. In case anybody is interested in picking this up, the code is actually quite simple (the underlying data structure remains the same; we're just inverting the order in the config). It would need documentation and tests, and probably a matching pushRewriteTo. diff --git a/remote.c b/remote.c index 28c7260ed1..26b1a7b119 100644 --- a/remote.c +++ b/remote.c @@ -345,6 +345,11 @@ static int handle_config(const char *key, const char *value, void *cb) return config_error_nonbool(key); rewrite = make_rewrite(&rewrites_push, name, namelen); add_instead_of(rewrite, xstrdup(value)); + } else if (!strcmp(subkey, "rewriteto")) { + if (!value) + return config_error_nonbool(key); + rewrite = make_rewrite(&rewrites, value, strlen(value)); + add_instead_of(rewrite, xmemdupz(name, namelen)); } } However, I did notice the cleanup below as part of writing that. It may be worth applying independently. -- >8 -- Subject: [PATCH] remote: check config validity before creating rewrite struct When parsing url.foo.insteadOf, we call make_rewrite() and only then check to make sure the config value is a string (and return an error if it isn't). This isn't quite a leak, because the struct we allocate is part of a global array, but it does leave a funny half-finished struct. In practice, it doesn't make much difference because we exit soon after due to the config error anyway. But let's flip the order here to avoid leaving a trap for somebody in the future. Signed-off-by: Jeff King <peff@xxxxxxxx> --- remote.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/remote.c b/remote.c index b850f2feb3..28c7260ed1 100644 --- a/remote.c +++ b/remote.c @@ -336,14 +336,14 @@ static int handle_config(const char *key, const char *value, void *cb) if (!name) return 0; if (!strcmp(subkey, "insteadof")) { - rewrite = make_rewrite(&rewrites, name, namelen); if (!value) return config_error_nonbool(key); + rewrite = make_rewrite(&rewrites, name, namelen); add_instead_of(rewrite, xstrdup(value)); } else if (!strcmp(subkey, "pushinsteadof")) { - rewrite = make_rewrite(&rewrites_push, name, namelen); if (!value) return config_error_nonbool(key); + rewrite = make_rewrite(&rewrites_push, name, namelen); add_instead_of(rewrite, xstrdup(value)); } } -- 2.20.0.rc1.703.g93fba25b62
