CODEWITHSUNDEEP
mysqloptimizationdate-format
plamen
plamen

Reputation: 63

How to optimize mysql date_format() for speed in where clause?

I'm trying to optimize relatively big mysql (myisam) table with 220,000 rows. The table itself is not so big - about 23.5MB in size. So, what's the real problem? - i got query like this:

SELECT * FROM table WHERE DATE_FORMAT(date_field, '%m%d') = '1128' LIMIT 10

I tried to set an index on date_field but EXPLAIN show that the index was not used at all ... i guess this is not so strange because of the DATE_FORMAT() . So, i'm planing to add another column that will hold the dates as '%m%d' and put an index on it. The only reason i don't want to do this is because of the data duplication.
Btw I use date_field is a birthdate field and I'm sure i always need the date_field as %Y-%m-%d or just %m%d

Do you have better suggestion about how to optimize the query above ? Thanks in advance !!!

Some info:

MySQL version: 5.0.51b-log
OS: slackware 12.1
CPU: Pentium III (Coppermine) at 996.783Mhz
RAM: 512MB DDR
HDD: 80GB SATA

P.S I tried to add another column that hold the dates as %m%d . The results are very good but i still don't like this approach. I'm waiting for more suggestions!

Upvotes: 5

Views: 3318

Answers (1)

nos
nos

Reputation: 229088

If you always need a wildcard on the year, like your query there, I'm not sure mysql will be able to use an index on a date/datetime field

If these are only dates, you can create a time_dimension table though, and prepopulate that with a calendar for the next handful of years. I've a stored procedure to do that if you should need one.

create table time_dimension (
 dbdate date primary key,
 year int NOT NULL,
 month int  NOT NULL , 
 day int NOT NULL,
 KEY(year),
 KEY(month);
 KEY(day);
);

You'd join your big data table to this relativly small table and filter on its field. e.g. 

SELECT * FROM data_table d 
  inner join time_dimension t 
    on d.date_field=t.dbdate 
where t.day=28 and t.month=11 LIMIT 10

This leverages filtering on the little time_dimension, and the joins on date_field = dbdate will normally use indexes.

Upvotes: 2

Related Questions