pgsql ext: pg_query_params/pg_send_query_params false constant handli…

[devnexen]

…ng proposal.

making a new subset of zend_get_string calls and concealed for the pgsql module for now.

[Girgias]

I really don't think adding a Zend API for this case is the way to go, when you can just to a check against zend_empty_string in PGSQL to branch of.

[devnexen]

I thought of something like this was more in case later it can be used beyond pgsql ext.

[devnexen]

ping :)

[cmb69]

Thanks for the PR! I'm not quite sure what to do here; the issue had been reported as https://bugs.php.net/68156, and re-classified as feature request back then. And at least the fix in this PR appears to introduce a BC break; consider using str instead of num in the when-clause; passing false would no longer match an empty string. I think we cannot change that behavior for any stable release; it might be okay to change it for master, but even that might be debatable.

Also, someone mentioned in that bug report that PDO_PGSQL would behave the same (I haven't checked whether this is still the case), so it might be reasonable to change its behavior as well.

BTW, I just noticed that the pgsql test suite is pretty volatile regarding the setup of the test tables; unless 01createdb.phpt is run, most other tests would fail since there is no explicit table creation.

[Girgias]

I think with our new policy we can ValueError on invalid types.