I am hoping to use plpython to perform various transforms on query results of very large size.
The documentation in the official 8.3 manual makes it appear as if the results of plpy.execute are read in at once (e.g., they appear to have random access and are mutable) rather than in the hidden cursor fashion of looping through a PgSql query result set. If this correct? If so does it mean that I need to avoid plpy.execute for very large queries? If so, a cursor/generator interface would seem to be a substantial improvement for the future.
If I cannot use plpy.execute, is there some way to declare and use a standard cursor from within plpython? I can find nothing on this on the web, and my own experimentation has been fruitless. Any quick example would be hugely appreciated.
If none of the above works, my fallback will be to execute the query in PgSql, then within the fetch loop call a plpython procedure. This works as far as my testing has gone, but seems unfortunate.