> Michael Stassen wrote:
>> For example, if the selected random id is missing, we take the next id we
>> find, like this:
>>
>> SELECT @rand_id:= CAST( 1 + MAX(id)*RAND() AS UNSIGNED) FROM history;
>> SELECT * FROM history WHERE id >= @rand_id LIMIT 1;
>
> That will have a possibly undesired effect. Records that have gaps in the
> IDs before them will be twice, three times, etc. (depending on the size of
> the gap), as likely to be selected as records with no preceding gaps.
Replace MAX with COUNT and the WHERE clause with an OFFSET and the gap
problem should go away...
-philip
