Re: [PATCH 1/6] Add docstrings for lookup_replace_object() and do_lookup_replace_object()

On 02/21/2014 07:21 PM, Junio C Hamano wrote:

Michael Haggerty [off-list ref] writes:

quoted

Signed-off-by: Michael Haggerty <redacted>
---
 cache.h | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/cache.h b/cache.h
index dc040fb..0ecd1c8 100644
--- a/cache.h
+++ b/cache.h

@@ -788,13 +788,29 @@ static inline void *read_sha1_file(const unsigned char *sha1, enum object_type *
 {
 	return read_sha1_file_extended(sha1, type, size, LOOKUP_REPLACE_OBJECT);
 }
+
+/*
+ * If a replacement for object sha1 has been set up, return the
+ * replacement object's name (replaced recursively, if necessary).
+ * The return value is either sha1 or a pointer to a
+ * permanently-allocated value.  This function always respects replace
+ * references, regardless of the value of check_replace_refs.
+ */
 extern const unsigned char *do_lookup_replace_object(const unsigned char *sha1);
+
+/*
+ * If object sha1 should be replaced, return the replacement object's
+ * name.  This function is similar to do_lookup_replace_object(),
+ * except that it when object replacement is suppressed, it always
+ * returns its argument unchanged.
+ */
 static inline const unsigned char *lookup_replace_object(const unsigned char *sha1)
 {
 	if (!read_replace_refs)
 		return sha1;
 	return do_lookup_replace_object(sha1);
 }
+
 static inline const unsigned char *lookup_replace_object_extended(const unsigned char *sha1, unsigned flag)
 {
 	if (!(flag & LOOKUP_REPLACE_OBJECT))

The above description is good, but after reading e1111cef (inline
lookup_replace_object() calls, 2011-05-15) that introduced this
ugliness, I have to wonder if do_lookup_replace(), which nobody
except lookup_replace_object() ever calls, is better removed from
the public API, making lookup_replace_object() an extern definition.

We do name functions that are purely helpers that are internal
implementation detals of the API as "do_blah", but exporting that
kind of name as if that is part of the API people are expected to
call feels very wrong.

I assume that the current design was to avoid the overhead of a function
call in the case that no replace references exist.  If we're willing to
eat that cost, then sure, we should bury do_lookup_replace_object() in
the implementation file.

Unless you say otherwise, I will work that change into my patch series.

Michael

-- 
Michael Haggerty
mhagger@alum.mit.edu
http://softwareswirl.blogspot.com/