Moved from: bobsql.com
I ran across a check in the XEvent string comparison logic which can be used to improve performance of the ‘like’ comparison.
At first glance the string comparison seems straightforward. However, the logic may require complex comparison (SQL Server collations and/or surrogate characters involved in UNICODE, UTF_16, etc.)
When performing comparison, simple checks are attempted first. For example:
If the compare value is ‘Bob%’ the logic may compare the first 3 characters and can avoid a more involved check.
If the compare value is ‘%Bob’ the logic may compare the last 3 characters and avoid a more involved check
However, a compare value of ‘%Bob%’ may require more work than