On 2013-01-25 14:51:39 -0500, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > Its slightly more complex than just making it one hash table with an
> > extended key. When validating a trigger function we don't have a
> > relation to do the cache lookup. I chose to handle that case by not
> > doing a cache lookup at all in that case which imo is a sensible
> > choice.
>
> Seems fair. However ... why is it safe for PLy_procedure_create to be
> using the same name for multiple instances of a trigger function?
> Should we not be including the rel OID when building the procName
> string?
I don't think its a problem, given the way python works I am pretty sure
it will result in independent functions.
Each PLy_procedure_compile will run the source code in a copy of
PLy_interp_globals, therefore the independent comilitions shouldn't
affect each other.
I am not sure why it builds the call to the function via eval'ing a
"$funcname()" instead of using the result of PyRun_String which will
return a reference to the function, but thats an independent issue.
Now I think an argument can be made that it would be nicer for debugging
purposes to have clearly distinguishable function names, but I
personally never needed it and it probably wouldn't be something to
backpatch. People might rely on those function names.
Greetings,
Andres Freund
--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services