Some day soon (5.6.4), MySQL will provide fractional seconds in TIMESTAMP columns, however, even fractional seconds aren't guaranteed to be unique, though theoretically, they'd most often be unique, especially if you limited MySQL to a single thread.
You can use a UUID if you need a unique number that is ordered temporally.
SELECT UUID()
yields something like:
45f9b8d6-8f00-11e1-8920-842b2b55ce56
And some time later:
004b721a-8f01-11e1-8920-842b2b55ce56
The first three portions of a UUID consist of the time, however, they're in order from highest precision to least, so you'd need to reverse the first three portions using SUBSTR()
and CONCAT()
like this:
SELECT CONCAT(SUBSTR(UUID(), 15, 4), '-', SUBSTR(UUID(), 10, 4),
'-', SUBSTR(UUID(), 1, 8))
Yields:
11e1-8f00-45f9b8d6
You obviously couldn't use a function like this as a default value, so you'd have to set it in code, but it's a guaranteed unique temporally ordered value. UUID() works at a much lower level than seconds (clock cycles), so it's guaranteed unique with each call and has low overhead (no locking like auto_increment).
Using the UUID() on the database server may be preferred to using a similar function, such as PHP's microtime()
function on the application server because your database server is more centralized. You may have more than one application (web) server, which may generate colliding values, and microtime() still doesn't guarantee unique values.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…