Kevin,
> This would require capture of information beyond what I was thinking
> about in terms of schema. Do you think we need to capture just index
> type, or something more? Do you propose that we capture the pg_*
> metadata related to every object referenced in the plan, every object
> related to every table in the query, or everything in the entire
> database?
Not right away. However, having collected the info in the first place
would give us the option of duping a copy of the system tables so that we
could mine them for more information.
Also, for our test harness, we will have pre-programmed object information
for the objects we use.
--
--Josh
Josh Berkus
Aglio Database Solutions
San Francisco