Question

Today I came across a quirky behavior from what looked like a simple SQL query. A SELECT, a few JOINs, ORDER BY a single field. What I am unable to explain right now is that the sort changes, for equal values, based on the LIMIT statement. Let me give you an example :

SELECT facture.ID, [ bunch a fields ... ] FROM facture 
CROSS JOIN contact [ a few JOINs ... ] 
WHERE facture.USER_ID=usager.ID AND usager.ID=contact.USER_ID 
ORDER BY contact.NAME DESC LIMIT 50;

The first results will be have the exact same name, and are organized like this : 166, 172, 167, 169 (the IDs). If I get rid of the LIMIT statement, I should have the same order, for those results … right? Nope : 172, 166, 167, 169. If I change the value of the limit, I also do get different order. With a LIMIT 60, I get : 167, 166, 172, 169.

So I’m trying to understand… why a different sorting order for same values? After sorting with the field specified in the ORDER BY, should the DB then sort based on IDs or something else? This just sounds random. BTW the database is PostgreSQL.

Answer 1

The order of the returned rows is not deterministic. as you are only ordering by contact.Name and all rows returned have the same name, the order of the rows can change from query to query. You should order by the ID, too.