[CLJ-1036] Util/hasheq should be hashing a BigInteger to the same values as Long, and BigInt

Fri, 12 Apr 2013 08:48:23 -0500

The doc string for hash states that it defines a hash code function that is consistent with =, but for java.math.BigInteger hash is not consistent with =.

user=> (apply = [(Long. -1) -1N (biginteger -1)])
true
user=> (map hash [(Long. -1) -1N (biginteger -1)])
(0 0 -1)
user=>

It is possible to have a PHM with two key/value pairs where the keys are equal, and the hash codes are different:

user=> (assoc clojure.lang.PersistentHashMap/EMPTY (biginteger -1) :oops! -1N :one)
{-1N :one, -1 :oops!}

The expected behavior is the same as PersistentArrayMap, which does not have this issue, because it does not hash its keys:

user=> (assoc clojure.lang.PersistentArrayMap/EMPTY (biginteger -1) :oops! -1N :one)
{-1 :one}

This same misbehavior also occurs for Doubles and Floats:

thalia.core=> (apply = [(Float. 1e9) (Double. 1e9)])
true
thalia.core=> (map hash [(Float. 1e9) (Double. 1e9)])
(1315859240 1104006501)

That leads to the same difference in array-map and hash-map behavior as above for BigInteger and BigInt.

Clojure JIRA

[CLJ-1036] Util/hasheq should be hashing a BigInteger to the same values as Long, and BigInt