Hi,
Here is the next version of the patch where, I think, all of Roberts's
claims related to the code have been fixed.
For example, expression 'x < 1 OR x < 2' is transformed to
'x < ANY (1,2)'.
Here, we still need to deal with the architectural issues. I like the
approach mentioned by Peter: try to transform the expression tree to
some 'normal' form, which is more laconic and simple; delay the search
for any optimization ways to the following stages.
Also, it doesn't pass pg_dump test. At first glance, it is a problem of
regex expression, which should be corrected further.
--
regards,
Andrei Lepikhov
Postgres Professional