Benjamin Peterson added the comment: 2013/1/2 Giovanni Bajo <rep...@bugs.python.org>: > > Giovanni Bajo added the comment: > > Il giorno 02/gen/2013, alle ore 06:52, Christian Heimes > <rep...@bugs.python.org> ha scritto: > >> >> Christian Heimes added the comment: >> >> Thanks for the information! I'm working on a PEP for the issue at hand. > > Since you're collecting ideas on this, I would like to stress that, in the > Python 3 transition, it was deemed acceptable to switch all objects to use > unicode strings for attribute names, making the hash computation of such > attributes (in the context of the instance dictionary) at least twice as slow > than it used to be (the 'at least' refers to the fact that longer strings > might have even worse effects because of a higher number of cache misses). > SipHash isn't twice as slow as the current hash function, not even for short > strings. > > So there is a precedent in slowing down the hash computation time in a very > important use case, and it doesn't look like hell froze over.
It's probably not to bad for attribute names because a) they're short b) they're interned c) the hash is cached. ---------- _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue14621> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
