On Jul 13, 2008, at 5:42 PM, Eric Lemoine wrote:
> > So far, so good; user can do: > > wifi_table = Table('wifi', metadata, > Column('the_geom', Geometry(4326)), > autoload=True) > > But ultimately I'd like that my users can do: > > wifi_table = Table('wifi', metadata, autoload=True) > > I tried this: > > from sqlalchemy.databases import postgres > postgres.ischema_names['geometry'] = Geometry > > This is ok, but during reflection, when SQLA creates Geometry objects, > it obviously passes no "srid" argument to the Geometry constructor, so > the Geometry objects all end up with the "srid" property set to -1. > The proper "srid" value to pass to the Geometry constructor is > actually in a PostGIS table (geometry_columns). So if a geometry > column is discovered, the table's "srid" value could be read from that > table and passed to the Geometry constructor. I thought about doing > something like that: > > from sqlalchemy.databases import postgres > def geometry_factory(): > // go read srid associated with table from geometry_columns > srid = > return Geometry(srid) > postgres.ischema_names['geometry'] = geometry_factory > > but geometry_factory doesn't have any connection object to go read the > srid value. > > My question is simple: do you see solutions to my problem? like before with asdecimal=False, we dont have a standard API for the "ischema_names" dict and again here is a place where you're looking for one. Such an API might look like: def create_postgis_type(table, connection): srid = connection.execute("select whatever you need to figure out SRID value").scalar() return Geometry(srid=srid) engine = create_engine('postgres://...', type_reflectors={ 'numeric':PGFloat, 'PostGIS':create_postgis_type }) where reflecttable() distinguishes between a TypeEngine class and a plain callable, which is assumed to implement a particular API. But thats just a guess. I wouldn't implement such an API casually since while its very easy to add little features like this, its much harder to change them or take them away after you've observed they're a bad idea or were not well thought out (additionally this one's a pretty big job to implement across every dialect). Any opinions from Jason/ Rick/other ? --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "sqlalchemy" group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~----------~----~----~----~------~----~------~--~---