[Python-checkins] cpython: Add comments to frozenset_hash().
raymond.hettinger
python-checkins at python.org
Sun Jan 5 21:00:37 CET 2014
http://hg.python.org/cpython/rev/247f12fecf2b
changeset: 88314:247f12fecf2b
user: Raymond Hettinger <python at rcn.com>
date: Sun Jan 05 12:00:31 2014 -0800
summary:
Add comments to frozenset_hash().
Also, provide a minor hint to the compiler on how to group the xors.
files:
Objects/setobject.c | 15 ++++++++++++++-
1 files changed, 14 insertions(+), 1 deletions(-)
diff --git a/Objects/setobject.c b/Objects/setobject.c
--- a/Objects/setobject.c
+++ b/Objects/setobject.c
@@ -738,6 +738,17 @@
static Py_hash_t
frozenset_hash(PyObject *self)
{
+ /* Most of the constants in this hash algorithm are randomly choosen
+ large primes with "interesting bit patterns" and that passed
+ tests for good collision statistics on a variety of problematic
+ datasets such as:
+
+ ps = []
+ for r in range(21):
+ ps += itertools.combinations(range(20), r)
+ num_distinct_hashes = len({hash(frozenset(s)) for s in ps})
+
+ */
PySetObject *so = (PySetObject *)self;
Py_uhash_t h, hash = 1927868237UL;
setentry *entry;
@@ -754,8 +765,10 @@
hashes so that many distinct combinations collapse to only
a handful of distinct hash values. */
h = entry->hash;
- hash ^= (h ^ (h << 16) ^ 89869747UL) * 3644798167UL;
+ hash ^= ((h ^ 89869747UL) ^ (h << 16)) * 3644798167UL;
}
+ /* Make the final result spread-out in a different pattern
+ than the algorithem for tuples or other python objects. */
hash = hash * 69069U + 907133923UL;
if (hash == -1)
hash = 590923713UL;
--
Repository URL: http://hg.python.org/cpython
More information about the Python-checkins
mailing list