Question

So I was writing some queries today and was using top 10 on a query just to get some sample data without an orderby and noticed it was actually returning different data on each call. It did not seem random however as it is a table with millions of rows and some entries always showed up, while others changed every call. What logic goes on in the back end to decide which rows to return when no orderby is specified? I am using SQL Server 2008r2, but I’d be interested if it is handled differently in other DBs.

Answer 1

Of course the order is entirely undefined. But I’ll try to mention some actual causes in real implementations because this was explicitly asked for. Here is a small sample:

• parallel query plan
• read-ahead results returning in random order
• the execution plan changed for some reason
• read of a column store index returns rows in internal order
• SQL Server is choosing an unordered index scan which is reading the pages in allocation order and not in logical order (the allocation order can change by reorganizing the table for example)

These causes are specific for SQL Server.