On Wednesday 23 January 2019 20:19:27 Tim Bunce wrote:
> A key goal of the DBI is to provide a database independant interface to
> databases to enable application portability. What you're suggesting
> seems inherently database specific.
As wrote DBI already has API for providing last error/warning/note and I
just tried to extend this API to provide list, not just one value. So
tried to describe some API which returns list.
Do you think that providing list of errors/warnings/notes is database
specific?
For example MySQL and MariaDB databases may produce more then one
warning for last executed statement. Same apply for PostgreSQL and
in databases with some kind of PL/SQL you can in most cases raise
warnings / errors too.
For example, here is simple PostgreSQL statement which cause generation
of more warnings and info messages:
\set VERBOSITY verbose
SET client_min_messages TO DEBUG;
DO $$ BEGIN RAISE NOTICE 'my notice' USING ERRCODE = '00001'; RAISE WARNING
'my warning' USING ERRCODE = '00002'; RAISE INFO 'my info' USING ERRCODE =
'00003'; RAISE DEBUG 'my debug' USING ERRCODE = '00004'; RAISE WARNING 'my
warning' USING ERRCODE = '00005'; END $$;
NOTICE: 00001: my notice
WARNING: 00002: my warning
INFO: 00003: my info
DEBUG: 00004: my debug
WARNING: 00005: my warning
And SQL statement for MariaDB which cause 2 warnings:
\W
SELECT CAST("1a" AS INT), CAST("2b" AS INT);
Warning (Code 1292): Truncated incorrect INTEGER value: '1a'
Warning (Code 1292): Truncated incorrect INTEGER value: '2b'
Both examples does not produce any error, just warnings or info
messages.
I guess that similar thing -- more then one warning for one statement --
can happen also in other databases.
> The DBI defines a way for drivers to offer database specific functionality.
>
> A good way to move a suggestion like this move forward is to work with
> multiple driver authors to add similar functionality to their drivers.
Therefore I'm writing to this list :-) What other developers think...
> Exposing the full features that a specific database makes available, via a
> database specific interface, is immediately helpful for users of that driver.
>
> Once multiple drivers provide similar functionality then it would be
> reasonable to work on a design for extending the DBI itself.
>
> Tim.
