Re: How to optimize an ORDER BY query using a range to avoid any filesorts

[Faury Rodriguez]

Actually, yeah it is not using a filesort, and it is much faster, but I don't understand the output of the explain.

Weight is 3 bytes and db_id is 1 byte, so according to key_len it's using the full key, however, the # of rows is all the rows in the table (which makes sense because it's getting that from the weight portion) vs. just the # of rows that matches my db_id range, which should have been ~1.9 million.

But doesn't this mean it won't scale well? Let's say I have 10 million rows, but my db_id range only matches 2 million. Will it really do a index scan of the entire table of 10 million or stop when it has looked at all the db_ids in my range.