Mysql, get rows before specific row in a multi-column index - mysql

Question

Say I have this table for high-scores:
id : primary key
username : string
score : int
User names and scores themselves can be repeating, only id is unique for each person. I also have an index to get high-scores fast:
UNIQUE scores ( score, username, id )
How can I get rows below the given person? By 'below' I mean they go before the given row in this index.
E.g. for ( 77, 'name7', 70 ) in format ( score, username, id ) I want to retrieve:
77, 'name7', 41
77, 'name5', 77
77, 'name5', 21
50, 'name9', 99
but not
77, 'name8', 88 or
77, 'name7', 82 or
80, 'name2', 34 ...

Answer 1

Here's one way to get the result:
SELECT t.score
, t.username
, t.id
FROM scores t
WHERE ( t.score < 77 )
OR ( t.score = 77 AND t.username < 'name7' )
OR ( t.score = 77 AND t.username = 'name7' AND t.id < 70 )
ORDER
BY t.score DESC
, t.username DESC
, t.id DESC
(NOTE: the ORDER BY clause may help MySQL decide to use the index to avoid a "Using filesort" operation. Your index is a "covering" index for the query, so we'd expect to see "Using index" in the EXPLAIN output.)
I ran a quick test, and in my environment, this does perform a range scan of the index and avoids a sort operation.
EXPLAIN OUTPUT
id select_type table type possible_keys key rows Extra
-- ----------- ----- ----- ------------------ ---------- ---- --------------------------
1 SIMPLE t range PRIMARY,scores_UX1 scores_UX1 3 Using where; Using index
(You may want to add a LIMIT n to that query, if you don't need to return ALL the rows that satisfy the criteria.)
If you have an unique id of a row, you could avoid specifying the values in the table by doing a join. Given the data in your question:
Here we use a second reference to the same table, to get the row id=70, and then a join to get all the rows "lower".
SELECT t.score
, t.username
, t.id
FROM scores k
JOIN scores t
ON ( t.score < k.score )
OR ( t.score = k.score AND t.username < k.username )
OR ( t.score = k.score AND t.username = k.username AND t.id < k.id )
WHERE k.id = 70
ORDER
BY t.score DESC
, t.username DESC
, t.id DESC
LIMIT 1000
The EXPLAIN for this query also shows MySQL using the covering index and avoiding a sort operation:
id select_type table type possible_keys key rows Extra
-- ----------- ----- ----- ------------------ ---------- ---- ------------------------
1 SIMPLE k const PRIMARY,scores_UX1 PRIMARY 1
1 SIMPLE t range PRIMARY,scores_UX1 scores_UX1 3 Using where; Using index

Answer 2

The concept of "below" for repeating scores is quite fuzzy: Think of 11 users having the same score, but you want the "10 below" a special row. That said, you can do something like (assuming you start with id=70)
SELECT score, username, id
FROM scores
WHERE score<=(SELECT score FROM scores WHERE id=77)
ORDER BY if(id=77,0,1), score DESC
-- you might also want e.g. username
LIMIT 5 -- you might want such a thing
;
Which will give you the rows in question inside this fuzzy factor, with the anchor row first.
Edit
Re-reading your question, you don't want the anchor row, so you need WHERE score<=(...) AND id<>77 and forget the first part of the ORDER BY
Edit 2
After your update to the question, I understand you want only those rows, that have one of
score < score in anchor row
score == score in anchor row AND name < name in anchor row
score == score in anchor row AND name == name in anchor row AND id < id in anchor row
We just have to put that into a query (again assuming your anchor row has id=70):
SELECT score, username, id
FROM scores, (
SELECT
#ascore:=score,
#ausername:=username,
#aid:=id
FROM scores
WHERE id=70
) AS seed
WHERE
score<#ascore
OR (score=#ascore AND username<#ausername)
OR (score=#ascore AND username=#ausername AND id<#aid)
ORDER BY
score DESC,
username DESC,
id DESC
-- limit 5 //You might want that
;

Answer 3

I think this is the query you want:
select s.*
from scores s
where s.score <= (select score
from scores
where id = 70
) and
s.id <> 70
order by scores desc
limit 4;