Comments on: Hash functions in T-SQL

By: Bernie

Bernie — Wed, 28 Jan 2009 22:20:29 +0000

It is some what ridiculous to have a checksum function that returns collisions to this degree. In my own experience using a population of a 100K rows and 3 fields(char, datetime, int) to compute a checksum on a record. 23 times the Checksum function return the same value. I deem this to be hardly deterministic and therefore this function and sql server by association is making my job more difficult. Temporarily, I am using the hashbyte function and passing that value to the checksum. No collisions so far :). I am very concerned that SQLServer has such a poor implementation of a CRC-32 hash algorith. I might switch to Oracle if I have anymore issues with hashing or third party hash written in C++.

By: Ryan WIlliams

Ryan WIlliams — Thu, 04 Dec 2008 00:10:27 +0000

FFF61235-EC1A-428C-9FC8-DD794D4115A4 also has the same checksum

By: Generating a unique number for each row « Exploring Microsoft.NET & Apple Macintosh

Mon, 04 Aug 2008 18:05:48 +0000

[…] Update: I happen to read that checksum() function is not guaranteed to give unique values. This is the case with SQL Server 2000 and I believe the same case with SQL Server 2005. For more information on checksum() collisions, take a look at this URL: http://blogs.clarience.com/davide/?p=11. […]