Re: [PATCH] general style: replaces memcmp() with proper starts_with()

[Jeff King <peff@peff.net>]

On Wed, Mar 12, 2014 at 12:39:01PM -0700, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> >>  static inline int standard_header_field(const char *field, size_t len)
> >>  {
> >> -	return ((len == 4 && !memcmp(field, "tree ", 5)) ||
> >> -		(len == 6 && !memcmp(field, "parent ", 7)) ||
> >> -		(len == 6 && !memcmp(field, "author ", 7)) ||
> >> -		(len == 9 && !memcmp(field, "committer ", 10)) ||
> >> -		(len == 8 && !memcmp(field, "encoding ", 9)));
> >> +	return ((len == 4 && starts_with(field, "tree ")) ||
> >> +		(len == 6 && starts_with(field, "parent ")) ||
> >> +		(len == 6 && starts_with(field, "author ")) ||
> >> +		(len == 9 && starts_with(field, "committer ")) ||
> >> +		(len == 8 && starts_with(field, "encoding ")));
> >
> > These extra "len" checks are interesting.  They look like an attempt to
> > optimize lookup, since the caller will already have scanned forward to
> > the space.
> 
> If one really wants to remove the magic constants from this, then
> one must take advantage of the pattern
> 
> 	len == strlen(S) - 1 && !memcmp(field, S, strlen(S))
> 
> that appears here, and come up with a simple abstraction to express
> that we are only using the string S (e.g. "tree "), length len and
> location field of the counted string.
> 
> Blindly replacing starts_with() with !memcmp() in the above part is
> a readability regression otherwise.

I actually think the right solution is:

  static inline int standard_header_field(const char *field, size_t len)
  {
          return mem_equals(field, len, "tree ") ||
                 mem_equals(field, len, "parent ") ||
                 ...;
  }

and the caller should tell us it's OK to look at field[len]:

  standard_header_field(line, eof - line + 1)

We could also omit the space from the standard_header_field. The caller
just ran strchr() looking for the space, so we know that either it is
there, or we are at the end of the line/buffer. Arguably a string like
"parent\n" should be "a parent header with no data" (but right now it is
not matched by this function). I'm not aware of an implementation that
writes such a thing, but it seems to fall in the "be liberal in what you
accept" category.

-Peff