The normal ROWID selection algorithm described above will generate
monotonically increasing unique ROWIDs as long as you never use the
maximum ROWID value and you never delete the entry in the table with
the largest ROWID. If you ever delete rows, then ROWIDs from
previously deleted rows might be reused when creating new rows.
The above is only true if you don't have a INTEGER PRIMARY KEY AUTOINCREMENT column (it will still work fine with INTEGER PRIMARY KEY columns). Anyway, this should be more portable / reliable:
SELECT foo FROM bar
WHERE _ROWID_ >= (abs(random()) % (SELECT max(_ROWID_) FROM bar))
LIMIT 1;
ROWID, _ROWID_ and OID are all aliases for the SQLite internal row id.
SELECT * FROM table WHERE id IN (SELECT id FROM table ORDER BY RANDOM() LIMIT x)
This is also applicable to MySQL. This runs faster because SQL engines first load projected fields of rows to memory then sort them, here we just load and random sort the id field of rows, then we get X of them, and find the whole rows of these X ids which is by default indexed.