Let’s say my Primary Key is: PRIMARY KEY(A,B)

I also need a standard BTREE index on B: KEY(B)

Now, would it be in my interest to change PRIMARY KEY(A,B) to PRIMARY KEY(B,A) and thus get the covering index on B and have no need for the additional KEY(B) ?

Thanks for your help and all the great information on this site.

This will catch indexes (A,B) and (A,C) as duplicate would not it ?
It would be simple if indexes would be single column only

SELECT * FROM information_schema.statistics WHERE table_schema = ‘?’ GROUP BY table_schema, table_name, column_name HAVING count(column_name) > 1;

It would be better to include ‘index_type’ in GROUP BY clause to more accurate answer.

In the above query bind the ‘database name’ parameter.

-Bog

This is pretty strange idea. Innodb PK is usually most efficient key to use for row data retrieval.

And what about Innodb PK ? I heard that to speed UP queries, I should declare an additionnal key on the PK, and force the use of this key instead of the primary (because innodb PK are stored in fragments). So, true or false ?