Changes between Initial Version and Version 1 of Ticket #19259, comment 12
- Timestamp:
- Nov 9, 2012, 5:15:19 PM (12 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #19259, comment 12
initial v1 6 6 > I am leaning towards just generating inefficient SQL if the query happens to be ran first thing in a new ConnectionWrapper (not just as first thing in a connection - first thing for the object). The reason is that otherwise we have to close the connection used for version checking and that is somewhat expensive. I am not feeling strongly at all about this... 7 7 8 Since it's a chicken and egg problem, is it really necessary to "autodetect" the feature then? I couldn't find any other example where Django enables/disables features based on the database version. What about just introducing a 'django.db.backends.postgresql9_psycopg2'backend? It's a trivial patch.8 Since it's a chicken and egg problem, is it really necessary to "autodetect" the feature then? I couldn't find any other example where Django enables/disables features based on the database version. What about just introducing a `django.db.backends.postgresql9_psycopg2` backend? It's a trivial patch. 9 9 10 10 > The harder problem is that you can't group by the main PK alone. You will need the primary keys of joined tables too, and that information isn't directly available in group by clause generation. (Actually, it might be in select and related_select_cols as "second" part of the tuple, not sure of this). In any case the MySQL code will not work directly.