On Wed, Apr 22, 2015 at 6:12 PM, Jim Nasby <Jim.Nasby@bluetreble.com> wrote: > On 4/22/15 2:12 PM, Merlin Moncure wrote: >> That being said, I think json types with their associated API, given >> that they are core types, will ultimately handle these types of >> problems. That way, at least, we can avoid adding syntax and >> functionality that will basically do the same thing. This reminds me >> a little bit of the json_build() vs enhanced row() syntax we discussed >> some time back. I didn't say so at the time, but for posterity, I >> think you were right...json_build() is working fine for building >> arbitrary record types and moving a record to json and deconstructing >> it should work just as well. > > The one part I don't care for in that is it seems rather inefficient to cast > something to JSON just so we can do things we really should be able to do > with a record. But perhaps it's not all that costly. > > As for allowing SQL and plpgsql functions to accept a record, I think our > JSON functionality just provided plenty of reason we should allow accepting > them, even if you can't do much with it: you *can* hand it to row_to_json(), > which does allow you to do something useful with it. So it seems reasonable > to me that we should at least accept it as a function argument.
I agree that that would be useful. I think the problem with an expression like rowvar.something is that PL/pgsql cannot infer the type of the result, and nothing else works without that. I doubt that it's practical to lift that restriction. PL/pgsql could introduce dedicated syntax for this operation, like DYNAMIC_EXTRACT(rowvar, colname, resulttype) or something, but that's going to be clunky at best. Whether we eventually do that or not, if we can allow rows to be passed in and then let people use json or hstore operators on them, that would be a helpful step forward, IMHO. I'm not sure if that's practical either, but maybe...
this need significant changes in plpgsql - it can enforce work with different types in expressions in cycle - so we should to leave expressions based on plans or we have to introduce alternated plans for different input types.