Re: scm_string_hash suboptimal?

[Florian Weimer]

Dirk Herrmann <address@hidden> writes:

> For the case of the symbols hashtable, however, a different
> implementation might be better, namely to use a weak vector and put
> the SCM values directly in there.  If the slot hash(obj) is taken,
> try hash(obj)+1 and so on.

Sorry, I forgot to mention this:

Yes, open addressing is worth a consideration, especially if you can
put rather complete records into the hash table (but you've got to be
careful when you implement deletion).  But please use hash(obj)-1
instead of hash(obj)+1, the wrap-around check will safe a few cycles,
then.