Question

I’m creating a MySQL MyISAM (Full textual searches are needed) table.

1. column_1 – contains a TEXT primary key (Data will be a 64 bit encoded string)
2. column_2 – references another table and is used for joins
3. column_3 – another TEXT column indexed for searches using MATCH
4. …

The table is likely to hold billions of records over time.

column_1 main search would be performed on the primary key column as follows. e.g.

SELECT * FROM table WHERE column_1 = 123;

column_2 main search would be performed as follows:

SELECT * FROM table_1 
JOIN table_2 ON ( table_1.column_2 == table_2.id );

column_3 main search would be performed as follows:

SELECT column_3, MATCH ( column_3 )
AGAINST ( 'TOKEN' ) AS score
FROM table_1;

I would like to take advice on the sort of indexes I would need to setup and any other advice that sounds relevant.

Thanks in advance.

P.S

Am I right in thinking that if you do a search e.g.

SELECT * FROM table WHERE id = 1; (where id column is not indexed)

The search on a substantial db would be slower than if the column was indexed?

Answer 1

1. You don’t need any more indices for the first query since the PRIMARY KEY is indexed already.
2. table_2.id should be indexed (if a text field, make the indexed on the first few bytes of the field). table_1.column_2 does not need to be indexed since you do no selection on that field.
3. column_3 needs a FULL TEXT index.

You are right in your final assumption. The index is made up by a data structure specifically tailored for searching in, with the column as key and a pointer to the correct row as the value. A search on a non-indexed field will require a full table scan (making the db examine every row of the table).