| re: randomly select records
"Erwin Moller"
<since_humans_read_this_I_am_spammed_too_much@spam yourself.com> a écrit dans
le message de news: 4242dc41$0$154$e4fe514c@news.xs4all.nl...[color=blue]
> Bob Bedford wrote:
>
> Hi Bob,
>[color=green]
>> I've a site where companies add their article.
>>
>> I'de like to provide a "lasts articles" table. By this, I'll show last
>> articles inserted. But I won't always the same articles at any refresh.
>> Question 1: how to get a "random" selection from the database, giving
>> more
>> priority to the last inserted (the ones with higher articleID)[/color]
>
> Do that in PHP and not in SQL.
> I don't know if it is possible in SQL, but it sounds like a monsterquery
> to
> me.
>
> How to approach?
> If you have all your articleid present in the database CONTINIOUSLY (no
> gaps
> caused by deleting records), I think you can relatively conjure up some
> algoritm that will produce one of the articleid's.
>
> The excact formula depends on the distribution you want, of course.
> How much bigger must the chance be for recent articles?
>
> If you have gaps between the articleid's, I suggest you first get them in
> ORDER BY articleid, and add indexnumbers to them in some array, like
> 1 articleid = 2
> 2 articleid = 3
> 3 articleid = 5
> 4 articleid = 9
> etc
>
> Then use the indexnumbers for the distribution.
>
> (If you have trouble conjuring up some formula, ask here again)
>[color=green]
>> Question 2: I'd like to provide one article by client. I won't show 3
>> articles from the same client only because he has been the last one to
>> insert his articles.
>>
>> So I must randomly select by choosing between last articleID, but only
>> one
>> per clientID.[/color]
>
> I would also do this in PHP and not in SQL.
> use ORDER BY clientid, articleid in your query, and loop over the result,
> picking a random articleid PER clientid
>[color=green]
>>
>> simplified table is:
>>
>> articleID
>> clientID
>> details
>> price.
>>
>> It is possible with MySQL ? or any idea how to achieve this ?
>>
>> PS:I've no access to mysql NG.[/color]
>
> EVERYBODY has access to mySQL.
> You can download it right now.
> :-)
>[/color]
Hi Erwin,
Thanks for suggestions.
I have a huge database, quering like you say will have a huge load on the
server.
I've seen MySQL allow to use RANDOM and subquery.
I think I must mix those things for doing that. The purpose of doing that is
to avoid to return 1000 results then thread only 5 of them. If the server
may returns directly the 5 items, it would just be great, and I'm sure it's
possible.
select distinct ClientID from client order by reverse(RAND()) LIMIT 0,5)
//get 5 different clients
Then for each client:
select distinct articleID from article where article.ClientID =
client.ClientID order by reverse(RAND()) LIMIT 0,1
This should work even if I haven't tested yet. The problem is that it
doesn't really select the last articles, it just select any random article
from any random client. I've a field called lastdatetime in the article
table. It does log the last modification datetime. I'd like to use this
value to get the best result, but actually don't know how.
Bob
| 
