一对一关系数据库关系的可能优化

Question

What are the strategies for optimizing one-to-one database joins, except indexing and possibly merging the tables?

Assume the tables contain 1 million rows in each.

Answer 1

Indexing your tables like you mentioned is already pretty good, but physically ordering table on disk according to an index order makes it almost ideal.

For this, you can create clustered indexes in SQL Server (clustered indexes are also supported on some other databases like PostgreSQL ). Unfortunately, SQLite does not support clustered indexes. Obviously, only one index can be clustered - basically, this is the one that table pages will try to be physically ordered by on disk.

Also, on all databases (including SQLite), you can simply copy your table while reordering it according to the index, using something like

CREATE TABLE mytable_ordered AS
SELECT * FROM mytable
ORDER BY key_column;

DROP TABLE mytable;

ALTER TABLE mytable_ordered RENAME TO mytable;

CREATE INDEX mytable_key_column_idx ON mytable (key_column);

You should do the same on another big table you mentioned. After this, joining these tables by ordered key 1:1 should be as fast as it could possibly get (short of merging these tables into one).