Forum OpenACS Development: Response to concatenated indices in ACS4Tcl (in PG?)

Collapse
Posted by Don Baccus on
I haven't seen core dump, but I have seen referential integrity errors  that were "false negatives", i.e. the foreign key reference didn't "see" the row with the key even though it clearly existed and "select"  would find it.  Dumping the two indices in the stock file storage package and replacing with two indices on the individual keys made that go away (this was Oracle).  I remember another thread about this somewhere on web/db.

PG doesn't have CONNECT BY and we do our tree-ordering sorts using generated sort keys that are stored in the table.  The sort key has a simple b-tree index on it.  This code's undoubtably more portable than  the CONNECT BY code.  It would be interesting to benchmark CONNECT BY and the sort key approach in Oracle to see which is faster ...