Waiting for your response
Ilan.

SELECT FIRST 1000 i.”id”, i.”slug”, i.”what_they_do”, i.”does_well”, i.”last_improves” FROM “inspections” i LEFT JOIN “services” s on i.”service_id” = s.”id” WHERE s.”active” = 1 AND s.”suspended” = 0 AND i.”report_date” > ’2005-04-01′ ORDER BY RAND() ASC;

and it executes very slow , Please help…

Then I initialize the values to random numbers in batches of, say, 10000:
update myTable set static_rand = 1000000 * rand() where static_rand is null limit 10000;

Repeat until there are no NULL values.

I can then do:
select id from myTable order by static_rand limit 10;
//get the id values in $VALUES
update myTable set static_rand = 1000000 * rand() where id in ($VALUES)

So basically, it “shuffles” the records to start with, and then pops the required set off the top of the deck, and moves the “used” part of the deck back to random locations within the deck.

This works well for quite large datasets, as the static_rand index fits into RAM and the record set being altered is generally quite small and fast to update.

Odds are quite good, actually, that you’ll be doing something else with the randomly-selected records anyway, so they’ll be in RAM as a working set usually.

Invariably, this is all wrapped up in application logic with a single entry point anyway, and the random constraint “feels” more like an application business requirement than a DB layer, so I have no problem doing this in the application layer…

But I’m sure a mysql guru could re-work it easily enough into an SPROC though.

I personally find this to fit KISS (Keep it simple, stupid) best of all the solutions I’ve tried over the years.

ymmv

Also you assume that id starts at 0. For example, if you have 1000 rows but the least id value is >1000, no random value between 0 and COUNT(*) will ever match any id in your table.

Also I’m not sure why you are setting a User Variable at all. You make no use of the variable in your query. I tried a test with your query and it still functions if I omit the expressions that initialize and increment @num.

SELECT (SELECT myTable2.id from myTable myTable2 where myTable2.id>=b.num LIMIT 1) c FROM (SELECT FLOOR(RAND() * (SELECTcount(*) FROM myTable)) num ,@num:=@num+1 from (SELECT @num:=0) a , myTable LIMIT X) b ;

To improve performance you can combine this solution with a left join:

SELECT IF(c.id is null, (SELECT c2.id FROM myTable c2 WHERE c2.id > b.num LIMIT 1), c.id) d FROM (SELECT FLOOR(RAND() * (SELECTcount(*) FROM myTable)) num ,@num:=@num+1 FROM (SELECT @num:=0) a , myTable LIMIT X) b LEFT JOIN myTable c ON( c.id=b.num);