Dev/Sqlite/Tips

Avoid LIKE

• Like is not performant, as it will search both lowercase and uppercase. You can get around this with PRAGMA case_sensitive_like=OFF;
• Like does not use an index.

Consider the following table:

CREATE TABLE tbl1
( id    integer      NOT NULL    PRIMARY KEY
, name  varchar(255) NOT NULL
);
CREATE INDEX tbl1__name ON tbl1 (name);

Now, consider the following query:

SELECT  id
FROM    tbl1
WHERE   name LIKE 'a%';

The clause name LIKE 'a%' does not use the index tbl1__name.

= Alternative: Use comparison

If the like has a wildcard at the end, and the data is standardized to one case (all lowercase or all uppercase), use an equality instead:

SELECT  id
FROM    tbl1
WHERE   name >= 'a' AND name < 'b';

This uses the index tbl1__name and returns the same data.

Avoid ORDER BY

ORDER BY will do an expensive sort operation which may not use an index. This will come into play with multiple-table joins

SELECT  t1.id
FROM    tbl1 t1
        JOIN tbl2 t2 ON t1.id = t2.id
ORDER BY t2.age
;

In most cases, SQLite will only apply one index. In this example, it will be the index for the JOIN. No index gets applied for the score column, even if one is available In some cases, SQLite will actually sort the result-set by the ORDER BY first, before it applies the WHERE.

Alternative: denormalize the table

Putting all the columns in one table and building an index across them.

Alternative: multiple where clauses

SELECT  t1.id
FROM    tbl1 t1
        JOIN tbl2 t2 ON t1.id = t2.id
WHERE   t2.age > 100 AND t2.age <= 130
;
SELECT  t1.id
FROM    tbl1 t1
        JOIN tbl2 t2 ON t1.id = t2.id
WHERE   t2.age >  70 AND t2.age <= 100
;

Beware bulk INSERTs into a table with a PRIMARY KEY

INSERT INTO 

INSERT INTO tbl2 (id, key)
SELECT  t1.id
FROM    tbl1 t1
        JOIN tbl2 t2 ON t1.id = t2.id
WHERE   t2.age > 100 AND t2.age <= 130
;

Instead, insert into the table with an ORDER BY

ATTACH

ATTACH database does not affect performance. There is no difference between the following

SELECT  t1.id
FROM    tbl1 t1
        JOIN tbl2 t2 ON t1.id = t2.id
;

ATTACH  'db2.sqlite3' AS 'db2';
SELECT  t1.id
FROM    tbl1 t1
        JOIN db2.tbl2 t2 ON t1.id = t2.id
;

Get a Solid State Drive

SQLite does a lot of disk reading and writing

Use an INDEX for your SELECTs, especially JOINs

Use TRANSACTIONs for multiple INSERT / UPDATE / DELETE

Use PRAGMA page_size 4096