Rich: On Tue, 10 Sept 2024 at 17:32, Rich Shepard <rshepard@xxxxxxxxxxxxxxx> wrote: > Only a duplicate problem because when I created this table I didn't make the > proj_nbr column a PK. Always report this is if you have future problems, so people know if it is a case of pilot error or corruption, solutions differ and the good for ones may harm others. > > Also, if you do not have any uniqueness criteria consider adding an > > "id identity" column, it is useful when shit hits the fan. > Yep. that's what I will do. Bear in mind the relational model on which relational databases are modeled needs unique rows ( i.e., no two full rows should be equal on all fields ). It is not enforced in SQL, but now having it normally leads to problems. When it cannot be done in any other way, normally adding an identity column is a cheap way to make them unique. I had that problem with CDR tables ( it means call detail record, and until I managed to add circuit identification I had the problem that you can have two calls from and two the same two numbers with equal start and end times ) and solved it using an identity column ( added just for this purpose, after doing it a couple times using ctid in testing, it is slightly more expensive, but a lot more civilized ). Francisco Olarte.
