Robert Haas wrote:
> experience, most bad plans are caused by bad selectivity estimates,
> and the #1 source of bad selectivity estimates is selectivity
> estimates for unknown expressions.
ISTM unknown expressions should be modeled as a range of
values rather than one single arbitrary value.
For example, rather than just guessing 1000 rows, if an
unknown expression picked a wide range (say, 100 - 10000
rows; or maybe even 1 - table_size), the planner could
choose a plan which wouldn't be pathologically slow
regardless of if the guess was too low or too high.
For that matter, it seems if all estimates used a range
rather than a single value, ISTM less in general we would
product less fragile plans.
