MySQL - Optimising Selects

Archive - Originally posted on "The Horse's Mouth" - 2004-12-21 05:51:52 - Graham Ellis
If you're going to be doing a complex select in MySQL, how should you formulate it to make it run efficiently?

This sounds like a simple question, but is actually very complex!

Firstly, consider creating indexes on any columns that are likely to be used in the selection of joining of records; choose a UNIQUE INDEX if a field is known to be unique, and a FULLTEXT INDEX if you want to be able to quickly search a text field for a particular value. Use just an INDEX for non-unique fields. Here are some example commands that create indeexes:



create unique index piidindex on b_ptab (piid(5));

create index piidindex on b_btab (piid(5));

create fulltext index synindex on b_btab (synopsis);

Your data structure is now optimised to allow for selects searching on the indexed fields.

Secondly, when you specify your tables to be joined, try to specify the tables at the "heart" of the join operation first before the tables that tag data onto the structure; you may want to specify a STRAIGHT JOIN to force these tables to be read first.

Third, if you're selecting based on text word content
- use MATCH in preference to LIKE
- and use LIKE in preference to RLIKE

Using these techniques, a well written query can run thousands of times faster that a poor one. I'm writing this note on a laptop running MySQL and I ran a following query to join:
* A table of books
* A table of Authors
* A pivot table to link authors to books
* A table of book subject
* and a table of book publishers

Out of some 550 books (with a total of around 700 different authors), I selected the 53 records that included the word "focus" in their Synopsis, using the command:
select sql_no_cache title,fullname,b_btab.biid,b_atab.aiid,pvid,subject,pubname from b_btab, b_atab, b_pivot, b_stab, b_ptab where b_btab.biid = b_pivot.biid and b_atab.aiid = b_pivot.aiid and b_stab.biid = b_btab.biid and b_ptab.piid = b_btab.piid and match (synopsis) against ('focus');
and it took 0.02 seconds. The same join, without indexes and any optimisation, took 3 minutes and 18 seconds.

Main Content

MySQL - Optimising Selects