> CREATE OR REPLACE FUNCTION myfunction(myrow mytable)
> RETURNS INTEGER AS $$
> SELECT myrow.c + myrow.b FROM myrow;
> $$ LANGUAGE sql;
> where "myrow" is a table with a different set of column names from
> "mytable". The existing behavior for that is to seek the column name
> in "myrow" (the table), failing that to seek it in the parameter,
> and only to throw an error if both fail.
Wow I never thought this would be possible. why didn't the designers of the language use myrow mytable%ROWTYPE for rows of a table as a parameter, given that it's a valid type in PL/PGSQL? I figure that way it would have been way easier to disambiguate the definitions.