> Past a "many-small" certain point on numbers of hash-tables, if that's > the right word, in a program, and intepreter process on a machine, is > it be more time-efficient to allocate a 2**32-byte table? Are > 'modulo' and 'doublesize' the only steps of the lookup process that it > would eliminate, and are they expensive ones? If so, what point? If > not, what's a tighter bottleneck?- Hide quoted text - > > - Show quoted text -
There's something to it. As stated, your idea cannot be implemented with current assumptions that Python makes, specifically that integers are their own hash values, possible others. Relax this, and you may be on to something, have a path, such as keying the object id into the hash value. -- http://mail.python.org/mailman/listinfo/python-list
