maxakuru
commented
6 months ago dumping some stats here to compare the two id methods using 1M iterations to simulate ~24 hours of events for a large site
[A]: Array.from({ length: 75 }, (_, i) => String.fromCharCode(48 + i)).filter((a) => /\d|[A-Z]/i.test(a)).filter(() => Math.random() * 75 > 70).join('')
11825.2205ms duration, 0.0118ms average
unique results: 604597
average frequency: 1.6540
collision rate: 0.3954%
top 10:
=> 13730 (2.27%)
8 => 1065 (0.18%)
V => 1051 (0.17%)
1 => 1044 (0.17%)
h => 1038 (0.17%)
K => 1038 (0.17%)
S => 1033 (0.17%)
i => 1032 (0.17%)
T => 1029 (0.17%)
G => 1023 (0.17%)
bottom 10:
2ACMNUqtx => 1 (0.00%)
01GJQU => 1 (0.00%)
6CTVYcir => 1 (0.00%)
BEFKQj => 1 (0.00%)
016JXg => 1 (0.00%)
anuw => 1 (0.00%)
45NRcm => 1 (0.00%)
8HSrs => 1 (0.00%)
79ahv => 1 (0.00%)
MYek => 1 (0.00%)[B]: Math.random().toString(36).slice(-4)
556.9435ms duration, 0.0006ms average
unique results: 728323
average frequency: 1.3730
collision rate: 0.2717%
top 10:
72fh => 8 (0.00%)
84mb => 8 (0.00%)
kr1l => 7 (0.00%)
b8ta => 7 (0.00%)
8vdc => 7 (0.00%)
4hll => 7 (0.00%)
eszi => 7 (0.00%)
s629 => 7 (0.00%)
oj8n => 7 (0.00%)
hjyb => 7 (0.00%)
bottom 10:
nb13 => 1 (0.00%)
smwi => 1 (0.00%)
4q0f => 1 (0.00%)
prvx => 1 (0.00%)
8bis => 1 (0.00%)
47qb => 1 (0.00%)
q4nw => 1 (0.00%)
ilhg => 1 (0.00%)
m57f => 1 (0.00%)
t8mv => 1 (0.00%)
current implementation has a high rate of empty ids generated (~2%)
we should change it to have similar collision rate (~0.4%), but more evenly spread throughout id values