Phil Hord <phil.hord@xxxxxxxxx> writes:
> With many thousands of references, a simple `git rev-parse HEAD` may take
> more than a second to return because it first loads all the refs into
> memory even though it will never use them.
>
> Defer loading any references until we actually need them.
>
> Signed-off-by: Phil Hord <phil.hord@xxxxxxxxx>
> ---
> log-tree.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/log-tree.c b/log-tree.c
> index 3b904f037..c1509f8b9 100644
> --- a/log-tree.c
> +++ b/log-tree.c
> @@ -84,8 +84,10 @@ void add_name_decoration(enum decoration_type type, const char *name, struct obj
> res->next = add_decoration(&name_decoration, obj, res);
> }
>
> +static void maybe_load_ref_decorations();
I'll tweak that "()" and the other one we see below to "(void)"
while queuing.
I am not sure if "maybe_" is a good name here, though. If anything,
you are making the semantics of "load_ref_decorations()" to "maybe"
(but I do not suggest renaming that one).
How about calling it to load_ref_decorations_lazily() or something?
I also wonder if decoration_loaded should now become function-scope
static in this new helper, but that can be left outside of the
topic.
Other than that, I like what this patch attempts to do. A nicely
identified low-hanging fruit ;-).
Thanks.
> const struct name_decoration *get_name_decoration(const struct object *obj)
> {
> + maybe_load_ref_decorations();
> return lookup_decoration(&name_decoration, obj);
> }
>
> @@ -150,10 +152,13 @@ static int add_graft_decoration(const struct commit_graft *graft, void *cb_data)
>
> void load_ref_decorations(int flags)
> {
> - if (!decoration_loaded) {
> + decoration_flags = flags;
> +}
>
> +static void maybe_load_ref_decorations()
> +{
> + if (!decoration_loaded) {
> decoration_loaded = 1;
> - decoration_flags = flags;
> for_each_ref(add_ref_decoration, NULL);
> head_ref(add_ref_decoration, NULL);
> for_each_commit_graft(add_graft_decoration, NULL);
