Re: some random() clarification needed

Please don't top-post. Inline (with trim) is better but at minimum
bottom-post.

On Tue, Jul 14, 2020 at 9:01 AM Marc Millas <marc(dot)millas(at)mokadb(dot)com> wrote:

> Hi,
> your answer helps me understand my first problem.
> so, I rewrote a simple loop so as to avoid the "volatile" behaviour.
> (at least I was thinking I did... looks like I was wrong !)
> step by step loop:
> DO $$
> BEGIN
> FOR counter IN 1..1000 LOOP
> begin
> declare
> id1 integer =ceiling(random()*2582);
> id3 date= '2000-01-01';
> id2 date;
> pren varchar;
> begin
> id2=id3 + (random()*7200)::integer;
> SELECT prenom FROM prenoms WHERE id=id1 into pren;
> INSERT INTO testparttransac (datenaissance, prenoms) values(id2,pren);
> end;
> end;
> END LOOP;
> END; $$;
>
> I truncated the table, executed the loop with no errors, and expected that
> a select count(*)
> may answer 1000 !
> no.
> it varies, from less than 1000 (much less, something like 900)
> and more than 1000 (up to 1094)
>
> so... what s "volatile" in the loop ?
>

Everything...

You are setting id1 to the result of an expression inside the loop.
Everytime that statement gets executed within the loop a new random number
is produced.

I mean, even "id 3 date = '2000-01-01'" is repeatedly casting (I think) the
string to a date and assigning it to the variable even though that
statement overall is effectively immutable.

David J.