2015-03-17 2:51 GMT+01:00 Peter Eisentraut <peter_e@gmx.net>:
On 3/12/15 8:12 AM, Pavel Stehule wrote: > 1. fix missing semicolon pg_proc.h > > Oid protrftypes[1]; /* types for which to apply > transforms */
Darn, I thought I had fixed that. > 2. strange load lib by in sql scripts: > > DO '' LANGUAGE plperl; > SELECT NULL::hstore; > > use load plperl; load hstore; instead
OK > 3. missing documentation for new contrib modules,
OK > 4. pg_dump - wrong comment > > +<-----><------>/* > +<-----><------> * protrftypes was added at v9.4 > +<-----><------> */
OK > 4. Why guc-use-transforms? Is there some possible negative side effect > of transformations, so we have to disable it? If somebody don't would to > use some transformations, then he should not to install some specific > transformation.
Well, there was extensive discussion last time around where people disagreed with that assertion.
I don't like it, but I can accept it - it should not to impact a functionality.
> 5. I don't understand to motivation for introduction of protrftypes in > pg_proc and TRANSFORM clause for CREATE FUNCTION - it is not clean from > documentation, and examples in contribs works without it. Is it this > functionality really necessary? Missing tests, missing examples.
Again, this came out from the last round of discussion that people wanted to select which transforms to use and that the function needs to remember that choice, so it doesn't depend on whether a transform happens to be installed or not. Also, it's in the SQL standard that way (by analogy).
I am sorry, I didn't discuss this topic and I don't agree so it is good idea. I looked to standard, and I found CREATE TRANSFORM part there. But nothing else.
Personally I am thinking, so it is terrible wrong idea, unclean, redundant. If we define TRANSFORM, then we should to use it. Not prepare bypass in same moment.