Comments on: SQL: Ranking without self join

By: links for 2010-06-25 | Digitalistic - Mashup or die trying

links for 2010-06-25 | Digitalistic - Mashup or die trying — Fri, 25 Jun 2010 14:03:51 +0000

[...] SQL: Ranking without self join | code.openark.org (tags: sql join ranking optimization mysql) [...]

By: shlomi

shlomi — Tue, 02 Feb 2010 19:53:03 +0000

George,

Reset as in @c := 0, nothing fancy…

By: George

George — Tue, 02 Feb 2010 19:38:16 +0000

Hi Shlomi,

This piece of gold has already helped me immensely, thanks.

As per #18, how would I reset the other variable? I cannot seem to find anything on mysql.com

By: MySQL: Another Ranking trick « JZ Talk Blogger

MySQL: Another Ranking trick « JZ Talk Blogger — Thu, 21 Jan 2010 07:28:32 +0000

[...] Another Ranking trick I just read SQL: Ranking without self join, in which Shlomi Noach shares a nice MySQL-specific trick based on user-defined variables to [...]

By: shlomi

shlomi — Thu, 14 Jan 2010 18:59:41 +0000

@Pau

Very nice! Thanks for sharing

By: Pau Sanchez

Pau Sanchez — Wed, 13 Jan 2010 16:27:08 +0000

Fantastic solution!

I was looking for a solution and it come up after looking to this post

My problem was having a sort_index field used to sort some elements in a particular order.

The thing is that the indexes were not consecutive, and that caused some trouble for some tasks I’m not going to talk about.

Thanks to your solution I came up with:

SET @pos := 0;
UPDATE sort_example
SET sort_index = (SELECT @pos := @pos+1)
ORDER BY sort_index ASC

So if before running this code the sort_indexes of several rows were:
1, 32, 34, 58, 99, 1028, 9982

After running the update they turned into:
1, 2, 3, 4, 5, 6, 7

Just wanted to share, maybe it can be useful to someone else.

Thanks!

By: Roland Bouman

Roland Bouman — Wed, 04 Nov 2009 06:59:03 +0000

Hi Sergio,

post your table definition, some 20 rows of sample data that illustrate the desired result.

By: Toby

Toby — Wed, 04 Nov 2009 02:24:31 +0000

Sergio,
You could always just use the self join approach, which makes your scenarios simple.

By: sergio

sergio — Wed, 04 Nov 2009 01:56:32 +0000

Hello again, I have tried what you said but it doesnt seem to work, can you provide a little more insigth really getting fustrated. Also as an added question to add to the previous question on top of the test_ids for multiple tests stored in the same table, what if another column was added with bonus_score which rank is determined first by the regular score and if equal, then the bonus score is used to determine the ranks? I really appricate any insight you can provide as this is very frustrating to me.

By: shlomi

shlomi — Tue, 03 Nov 2009 05:36:54 +0000

@sergio
Yes, it is possible; you would need to maintain two sets of variables ; the primary being the test_id, the secondary being for the score.
Whenever the test id changes, you reset the score variables.
you first ORDER BY test_id DESC, SCORE DESC.