On Aug 19, 2010, at 8:08 AM, Robert Haas wrote:
>> Another possibility is for EXECUTE USING to coerce any unknowns to TEXT
>> before it calls the parser at all. This would square with the typical
>> default assumption for unknown literals, and it would avoid having to
>> have any semantics changes below the SPI call.
>
> That seems more intuitive than just chucking an error.
It'd be nice if SPI itself could work this way for UNKNOWNs, too.
Best,
David
