Re: [Sqlalchemy-tickets] [sqlalchemy] #1511: [PATCH] PGEnum PostgreSQL enum type

[sqlalchemy <mi...@zz...>]

#1511: [PATCH] PGEnum PostgreSQL enum type
-----------------------------+----------------------------------------------
    Reporter:  guest         |       Owner:  zzzeek           
        Type:  enhancement   |      Status:  new              
    Priority:  medium        |   Milestone:  0.6.0            
   Component:  postgres      |    Severity:  major - 1-3 hours
  Resolution:                |    Keywords:  patch            
Status_field:  needs review  |  
-----------------------------+----------------------------------------------
Comment (by zzzeek):

 im still a little antsy on the `__metadatas` collection.  its almost
 always harmless but there are some cases where we've had "memory leak"
 issues, for applications which make many engines or metadatas.  Even
 though this is not the typical programming paradigm, it would be nice if
 there weren't a "surprise" memory leak if someone happened to use many
 metadatas and this particular type object.   It would be better if we at
 some point added a "has_event()" type of hook to `MetaData` and `Table`
 for this kind of thing.


 another thing missing here is that there's no hook for "does this type
 already exist ?" (i.e. in the actual PG database).  otherwise its not
 possible here to call `metadata.create_all()` multiple times.  If the PG
 dialect also implemented `has_type()` the way it implements `has_table`
 and `has_sequence`, the hook can be placed as the "on" callable within the
 `CreateType`/`DropType` objects which would only create/drop the type if
 it didn't/did exist.

 also when would `create_statement()`, `drop_statement()`,
 `__ddl_listener()` on `SchemaDefinedType` be called ?  I tried to follow
 each patch visually and it seems like these are still there (I may have
 missed something though).

-- 
Ticket URL: <http://www.sqlalchemy.org/trac/ticket/1511#comment:3>
sqlalchemy <http://www.sqlalchemy.org/>
The Database Toolkit for Python