Re: problem with timezone parsing

Il giorno mar, 17/02/2009 alle 16.57 +0100, Karsten Hilbert ha scritto:
> On Tue, Feb 17, 2009 at 11:39:30PM +0900, James Henstridge wrote:
> 
> > >> As mentioned before, if you want behaviour different to this I'd
> > >> suggest installing an alternative typecaster.  That would also let you
> > >> fix the busted time zone parsing for released versions of psycopg2.
> > >
> > > Is there a way to install a custom typecaster, then from
> > > inside which letting the default typecaster handle things
> > > but listen for the above exception and handle that
> > > appropriately ?
> > 
> > You can call the other typecaster directly:
> > 
> >     timestamp = psycopg2.DATETIME(some_string, cursor)
> > 
> > Your typecast function will be passed the string and cursor as arguments.
> 
> Ah, OK, thanks. Now, two more questions ;-)
> 
> Is there a way to install a module wide type caster ?
> 
>  Per-connection type casters were introduced not too long
>  ago (and the examples show how to use them) but can we
>  still get the per-psycopg2 instance casters? The
>  register_type() seems to want either a connection or a
>  cursor. The "help(psycopg2.extensions.register_type)"
>  interactive help doesn't say whether the connection/cursor
>  argument is mandatory (in fact, it doesn't mention it at
>  all ...)

Yes, just don't pass that parameter or pass None.

> Is there a way to not need to access the DB (for learning
> the type OID) and still use new_type (which seems a
> prerequisite for register_type).

If you know the oid you can hard-code it into your code.

>  It seems possible as one might assume the OID for
>  "timestamp with time zone" to be fixed - but I remember
>  reading on the PG list that that isn't guarantueed for all
>  times. However, UNICODE and UNICODEARRAY do seem to assume
>  a fixed type OID (because they can be register_type()ed w/o
>  needing to determince the type OID). So I'd try to be
>  clever and deduce the OID from psycopg2.DATETIME.values but
>  unfortunately there is no documentation as to what those
>  values are. Digging in psycopg/pgtypes.h reveals that those
>  are, indeed, the needed OIDs.

Yes, the values are the OIDs.

> Thereby the second question can be answered yes but needs
> reconfirmation that re-using those OIDs is reasonably safe ?

It is completely safe. They are generated from your PostgreSQL headers
and guaranteed to be stable.

federico

-- 
Federico Di Gregorio                         http://people.initd.org/fog
Debian GNU/Linux Developer                                [email protected]
INIT.D Developer                                           fog-NGVKUo/i/[email protected]
 I did appreciate the irony that I was whining about encoding issues on
  a mailing list that was unable to show those chars, too.
                                -- Antti S. Lankila to mono-devel-list@

_______________________________________________
Psycopg mailing list
Psycopg-IAPFreCvJWPBWskQ1e/[email protected]
http://lists.initd.org/mailman/listinfo/psycopg

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEABECAAYFAkma3zUACgkQvcCgrgZGjevlKACgoyiw8MSTyeJQ5sbK/1iuaRlq
GXcAniiSYiZTEwamfjvAJMXTq45+e6JA
=fuFd
-----END PGP SIGNATURE-----