When reviewing execution statistics, I am frequently interested in the behavior of "classes" of queries, rather than individual queries, for example queries which
- Contain a join
- Touch a specific column
- Use POSIX regular expressions
AFAIK this sort of summary/rollup information can't be computed reliably from pg_stat_statements because of rows being discarded in the face of a large number of (post normalization/jumble) distinct statements.
Is there a way to retrieve these kinds of statistics in Postgres today?
Thanks,
Jacob
