On Mon, Jun 2, 2014 at 7:22 AM, Andrew Dunstan <andrew@dunslane.net> wrote:
> Yes, I've just been looking at that. I think this is probably a hangover
> from when these routines were recast to some extent. Given that we're not
> seeing any errors from it, I'd be inclined to remove the the "|| val->type
> == jbvBinary" part. One of the three call sites to convertJsonbValue asserts
> that this can't be true, and doing so doesn't result in a regression
> failure.
Well, I guess the latter condition should be moved, since clearly the
jbvBinary case isn't handled within convertJsonbScalar(). It would be
mildly preferable to have a more accurate error message (the
convertJsonbValue() error) if that "can't happen" condition should
ever happen.
--
Peter Geoghegan