[PATCH 3/3] Use GPerf for command names.
sgrimm at facebook.com
Wed Nov 7 19:07:35 UTC 2007
You could use it for commands like "stats" that have subcommands. Hard
to see how it'd be useful for much beyond that.
This was a good idea, BTW. I like it. We should make sure that the
release tarballs don't require gperf to be installed; the hash values
should be the same everywhere for a given release of the code (since
the command set is fixed) so there shouldn't be any need to make
people install gperf unless they're adding/removing commands in the
text protocol. We can just run it once before tarring and let end
users reuse the output.
On Nov 7, 2007, at 10:56 AM, Aaron Stone wrote:
> The resulting code is much prettier. The existing code does satisfy
> O(1) requirement, but as a big fan of gperf, I like this a lot.
> How do you use gperf for further parsing?
> On Wed, Nov 7, 2007, Tomash Brechko <tomash.brechko at gmail.com> said:
>> Using one lookup per command instead of a sequence of strcmp() is
>> scalable. This patch doesn't change command syntax or semantics in
>> any way. The next step would be to use GPerf for the rest of the
> [snip patch]
More information about the memcached