Re: Changing the function used in an index. - Mailing list pgsql-general

From Tom Lane
Subject Re: Changing the function used in an index.
Date
Msg-id 21420.1373644747@sss.pgh.pa.us
Whole thread Raw
In response to Changing the function used in an index.  (Clodoaldo Neto <clodoaldo.pinto.neto@gmail.com>)
List pgsql-general
Clodoaldo Neto <clodoaldo.pinto.neto@gmail.com> writes:
> How does the planner know that the function used in an index has changed?
> If the function's body is changed keeping the argument and return types the
> planner will not use it anymore:

Ordinarily, changing a function definition like that *would* break
things.  The only thing that's saving you from yourself in this example
is that the function is inline-able SQL, and so what the planner is
seeing is not "f(i)" vs "f(i)", but "i * -1" vs just "i".  Even so,
the index is broken/corrupt, or will be as soon as you make more
insertions into it with the changed function.

            regards, tom lane


pgsql-general by date:

Previous
From: Kevin Grittner
Date:
Subject: Re: transactional swap of tables
Next
From: Bradley McCune
Date:
Subject: Re: V8.4 TOAST table problem