Cache plan invalidation

Bruce Momjian wrote:

The current TODO list has:

Dependency Checking
===================

* Flush cached query plans when the dependent objects change,
when the cardinality of parameters changes dramatically, or
when new ANALYZE statistics are available

A more complex solution would be to save multiple plans for different
cardinality and use the appropriate plan based on the EXECUTE values.

* Track dependencies in function bodies and recompile/invalidate

This is particularly important for references to temporary tables
in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
in PL/PgSQL is to use EXECUTE. One complexity is that a function
might itself drop and recreate dependent tables, causing it to
invalidate its own query plan.

Which of these are done or not done?

Also, is this done:

* Invalidate prepared queries, like INSERT, when the table definition
is altered

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce@momjian.us> writes:

Bruce Momjian wrote:

The current TODO list has:

Dependency Checking
===================

* Flush cached query plans when the dependent objects change,
when the cardinality of parameters changes dramatically, or
when new ANALYZE statistics are available

A more complex solution would be to save multiple plans for different
cardinality and use the appropriate plan based on the EXECUTE values.

This is partially done --- you'll have to split it into multiple items
if you want to preserve the bit about keeping different plans for
different parameter values. Note that in the current code, any VACUUM
or ANALYZE on a table will force relcache inval and hence replan; see
vac_update_relstats. So the only case not covered as far as
non-parameterized queries go is large growth of a table without any
vacuuming or analyzing ... and you're going to have problems anyway
if you don't analyze after loading a table. We may in fact find that
our problem is now too many replans rather than too few.

* Track dependencies in function bodies and recompile/invalidate

This is particularly important for references to temporary tables
in PL/PgSQL because PL/PgSQL caches query plans.

This is done.

Also, is this done:

* Invalidate prepared queries, like INSERT, when the table definition
is altered

This too.

regards, tom lane

I removed the cardinality item and marked the others as done:

* -Flush cached query plans when the dependent objects change or
when new ANALYZE statistics are available
* -Track dependencies in function bodies and recompile/invalidate
* -Invalidate prepared queries, like INSERT, when the table definition
is altered

Let's see if the cardinality issue is still needed after this release.

---------------------------------------------------------------------------

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

Bruce Momjian wrote:

The current TODO list has:

Dependency Checking
===================

* Flush cached query plans when the dependent objects change,
when the cardinality of parameters changes dramatically, or
when new ANALYZE statistics are available

A more complex solution would be to save multiple plans for different
cardinality and use the appropriate plan based on the EXECUTE values.

This is partially done --- you'll have to split it into multiple items
if you want to preserve the bit about keeping different plans for
different parameter values. Note that in the current code, any VACUUM
or ANALYZE on a table will force relcache inval and hence replan; see
vac_update_relstats. So the only case not covered as far as
non-parameterized queries go is large growth of a table without any
vacuuming or analyzing ... and you're going to have problems anyway
if you don't analyze after loading a table. We may in fact find that
our problem is now too many replans rather than too few.

* Track dependencies in function bodies and recompile/invalidate

This is particularly important for references to temporary tables
in PL/PgSQL because PL/PgSQL caches query plans.

This is done.

Also, is this done:

* Invalidate prepared queries, like INSERT, when the table definition
is altered

This too.

regards, tom lane

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +