←back to thread

The probability of a hash collision (2022)

(kevingal.com)
140 points subset | 1 comments | 22 Jun 25 03:34 UTC | HN request time: 0.22s | source
Show context
Simon_O_Rourke ◴[25 Jun 25 11:30 UTC] No.44376039[source]▶
I've had the privilege (?) of experiencing one in the wild only once. Was hashing a few columns to create a hash key for a dimensional table in snowflake and two wholy different rows clashed.
replies(1): >>44379193 #
1. mjuarez ◴[25 Jun 25 16:36 UTC] No.44379193[source]▶
By default, Snowflake doesn't seem to use any crytographic hash algorithms, so that's not completely unexpected. Based on (1) this link, it uses a proprietary 64-bit hash, and has this tidbit: "Do not use HASH to create unique keys."

(1) https://docs.snowflake.com/en/sql-reference/functions/hash