[TransWarp] pgdb.py from PostgreSQL 7.3.2

[Phillip J. Eby]

At 05:05 PM 2/14/03 +0100, Radek Kanovsky wrote:

>There was no method pgdbCursor.nextset() in 7.2.4 version of PostgreSQL.
>If I understand DBAPI 2.0 correctly, pgdb.py behaviour follows
>specification.

Yes, but it goes against the recommendation in footnote 3, that says the 
preferred approach is not to have the attribute in the first place.  They 
sure went to a lot of trouble to add a function that doesn't do anything 
and guarantees that any generic SQL layer will waste time calling the 
function and throwing the exception!  I would suggest encouraging them to 
remove the method, if it never does anything.  The spec says quite clearly:

"The preferred approach is to not implement the method and thus have Python 
generate an AttributeError in case the method is requested. This allows the 
programmer to check for database capabilities using the standard hasattr() 
function."


>So I have made small patch:
>
>--- /home/radekk/work/PEAK/src/peak/storage/SQL.py
>+++ /usr/local/python-2.2.2/lib/python2.2/site-packages/peak/storage/SQL.py
>@@ -77,7 +77,10 @@
>
>
>      def nextset(self):
>-        return getattr(self._cursor, 'nextset', _nothing)()
>+        try :
>+            return getattr(self._cursor, 'nextset', _nothing)()
>+        except self.conn.API.NotSupportedError :
>+            return None
>
>
>      def execute(self, *args):
>
>
>Is NotSupportedError dereferencing through self.conn.API correct?

Yes, that's fine.  I will probably make a similar change to PEAK, but I may 
add NotSupportedError to the connection's bindings so that it can be done 
as 'self.conn.NotSupportedError'.  Anyway, thanks for bringing it to my 
attention, although I would again encourage any DB driver developers to 
follow the DBAPI recommendation in this, as it allows for less calling 
overhead in this situation.