Closed
Description
As reported by @ge_o the randomness of db.collection.any() should be improved.
Example:
/* set up a collection with n = 100 values */
db._drop("rand");
var c = db._create("rand");
var i, n = 100, dist = { };
for (i = 0; i < n; ++i) {
c.save({ value: i });
dist[i] = 0;
}
/* get some random values and increase counters */
for (i = 0; i < n * 100; ++i) {
dist[c.any().value]++;
}
/* print results */
dist;
On my machine this returns some very unbalanced distribution, e.g.:
...
"80" : 82,
"81" : 36,
"82" : 52,
"83" : 75,
"84" : 28,
"85" : 102,
"86" : 54,
"87" : 120,
"88" : 40,
"89" : 61,
"90" : 34,
"91" : 24,
"92" : 172,
"93" : 32,
"94" : 99,
"95" : 268,
"96" : 24,
"97" : 25,
"98" : 36,
"99" : 59
Contrary, generating pseudo-random values with JavaScript Math.random() generates a much more balanced distribution:
...
"80" : 101,
"81" : 95,
"82" : 103,
"83" : 111,
"84" : 86,
"85" : 90,
"86" : 98,
"87" : 104,
"88" : 102,
"89" : 88,
"90" : 103,
"91" : 83,
"92" : 90,
"93" : 86,
"94" : 95,
"95" : 104,
"96" : 99,
"97" : 96,
"98" : 102,
"99" : 88