Thread: Pl/Python runtime overhead

On Wed, Aug 7, 2013 at 7:43 AM, Seref Arikan
<serefarikan@kurumsalteknoloji.com> wrote:
> When a pl/python based function is invoked, does it keep a python runtime
> running across calls to same function? That is, if I use connection pooling,
> can I save on the python runtime initialization and loading costs?

You can use the following wrapping technique to cache function's body,
that will save you some resources and time. It stores the main() in SD
(session data) built-in object and retrieves it when stored, so
plpython does not need to process it every time stored function is
called.

CREATE OR REPLACE FUNCTION some_plpython_function()
 RETURNS integer
 LANGUAGE plpythonu
AS $function$
""" An example of a function's body caching and error handling """

sdNamespace = 'some_plpython_function'

if sdNamespace not in SD:

    def main():
        """ The function is assumed to be cached in SD and reused """

        result = None

        # Do whatever you need here

        return result

    # Cache body in SD
    SD[sdNamespace] = main

try:
    return SD[sdNamespace]()
except Exception, e:
    import traceback
    plpy.info(traceback.format_exc())

$function$;

I can also recommend you to cache query plans, as plpython does not do
it itself. The code below also works with SD to store prepared plans
and retrieve them. This allows you to avoid preparing every time you
are executing the same query. Just like plpgsql does, but manually.

if SD.has_key('%s_somePlan' % sdNamespace):
    somePlan = SD['%s_planName' % sdNamespace]
else:
    somePlan = plpy.prepare(...)


> Are there any documents/books etc you'd recommend to get a good
> understanding of extending postgres with languages like python? I'd really
> like to get a good grip of the architecture of this type of extension, and
> possibly attempt to introduce a language of my own choosing. The docs I've
> seen so far are mostly too specific, making it a bit for hard for me to see
> the forest from the trees.

AFAIK, this one is the best one
http://www.postgresql.org/docs/9.2/interactive/plpython.html.

--
Kind regards,
Sergey Konoplev
PostgreSQL Consultant and DBA

http://www.linkedin.com/in/grayhemp
+1 (415) 867-9984, +7 (901) 903-0499, +7 (988) 888-1979
gray.ru@gmail.com

On 8/7/13 10:43 AM, Seref Arikan wrote:
> When a pl/python based function is invoked, does it keep a python
> runtime running across calls to same function? That is, if I use
> connection pooling, can I save on the python runtime initialization and
> loading costs?

The Python interpreter is initialized once during a session, normally
when the first PL/Python function is called.  So yes, connection pooling
can be helpful here.

> Are there any documents/books etc you'd recommend to get a good
> understanding of extending postgres with languages like python? I'd
> really like to get a good grip of the architecture of this type of
> extension, and possibly attempt to introduce a language of my own
> choosing. The docs I've seen so far are mostly too specific, making it a
> bit for hard for me to see the forest from the trees.

The basic documentation is here:
http://www.postgresql.org/docs/devel/static/plhandler.html.  The rest is
mainly experience and copying from existing language handler
implementations.