从数百万条记录中选择排名

Question

I will ask SO for help since I could not find similar situation, questions/post with my question

Assuming I have millions of records, columns are
user_id - assuming its recorded from 1 to 1,000,000
name - assume its also recorded up to 20 characters in alphabet
score - 0 to 100 assuming its also recorded
date - date it was recorded (timestamp)

user_id |   name   | score |        date       |
------------------------------------------------
23131   |   name1  |   15  | 2017-01-04 02:01:25
26824   |   name2  |   63  | 2017-01-04 02:41:33
19684   |   name3  |   28  | 2017-01-04 02:56:15
74937   |   name4  |   01  | 2017-01-04 04:07:55
27486   |   name5  |   75  | 2017-01-04 13:07:45
86476   |   name6  |   56  | 2017-01-04 14:21:47
36479   |   name7  |   19  | 2017-01-04 17:11:15
86752   |   name8  |   38  | 2017-01-04 18:22:23
11267   |   name9  |  100  | 2017-01-04 20:34:42
88763   |   name10 |   89  | 2017-01-04 22:45:43

1. I want to know my own rank assuming I know what my user_id is
2. I also want to get other 10 user records above and below my ranking, let's say my rank is 100, I also want to select user with rank of 90 to 99 (above my rank) and 101 to 110 (below my rank).
   if different user has same score order rank by date recorded, earlier record has higher rank.

is it possible?
assuming all records are unique and no index are set.

I know how to sort ranking

SELECT * FROM record order by score

but this would let me select all the records, what are the practical way of selecting specific data without selecting every record?

here is what I would like to achieve

user_id |   name   | score |        date          |     rank     |
------------------------------------------------------------------
12341   |   namep  |   90  | 2017-01-01 04:02:36  |      90      |
45341   |   nameo  |   88  | 2017-01-02 00:05:45  |      91      |
24341   |   namex  |   88  | 2017-01-03 00:11:15  |      92      |
26867   |   namec  |   83  | 2017-01-03 01:41:23  |      93      |
19156   |   nameb  |   81  | 2017-01-03 02:36:45  |      94      |
74973   |   namem  |   79  | 2017-01-03 04:07:55  |      95      |
23134   |   namek  |   78  | 2017-01-04 02:01:25  |      96      |
21424   |   namet  |   77  | 2017-01-04 02:41:33  |      97      |
19534   |   nameg  |   77  | 2017-01-04 02:56:15  |      98      |
74912   |   namez  |   75  | 2017-01-04 04:07:55  |      99      |

my_uid  |  my_name |   75  | 2017-01-04 13:07:45  |     100      |

86766   |   namen  |   75  | 2017-01-04 14:21:47  |     101      |
67976   |   namey  |   74  | 2017-01-04 16:22:23  |     102      |
34676   |   nameu  |   74  | 2017-01-04 17:33:32  |     103      |
86236   |   namei  |   73  | 2017-01-04 18:11:09  |     104      |
98636   |   nameo  |   73  | 2017-01-04 19:21:47  |     105      |
14326   |   namep  |   73  | 2017-01-04 20:33:22  |     106      |
45333   |   namet  |   72  | 2017-01-04 20:44:12  |     107      |
33323   |   namer  |   72  | 2017-01-04 21:34:26  |     108      |
11322   |   namee  |   71  | 2017-01-04 22:51:54  |     109      |
86633   |   namew  |   70  | 2017-01-04 22:55:33  |     110      |

---

ok so here is what I got as of now, sorry that I did not mention anything about not using union or union all, I cannot use that in my project.

but anyway here is my query I used "multi_query()" function

$sql = "SELECT score, date FROM table_name WHERE user_id=your_user_id;" //assume you already know your user_id
$sql .= "SELECT name, score, date FROM table_name WHERE score >= your_score ORDER BY score, date LIMIT 10;"; //to get 10 rows that have greater or same score of your score order by date, earlier date is higher rank if score is the same with other user.
$sql .= "SELECT name, score, date table_name WHERE score <= your_score DESC, date ASC LIMIT 10"; //select score less than or equal to my score order by score and date

and I get something like this

my_uid  |  my_name |   75  | 2017-01-04 13:07:45  |     100      |

---

12341   |   namep  |   90  | 2017-01-01 04:02:36  |      90      |
45341   |   nameo  |   88  | 2017-01-02 00:05:45  |      91      |
24341   |   namex  |   88  | 2017-01-03 00:11:15  |      92      |
26867   |   namec  |   83  | 2017-01-03 01:41:23  |      93      |
19156   |   nameb  |   81  | 2017-01-03 02:36:45  |      94      |
74973   |   namem  |   79  | 2017-01-03 04:07:55  |      95      |
23134   |   namek  |   78  | 2017-01-04 02:01:25  |      96      |
21424   |   namet  |   77  | 2017-01-04 02:41:33  |      97      |
19534   |   nameg  |   77  | 2017-01-04 02:56:15  |      98      |
74912   |   namez  |   75  | 2017-01-04 04:07:55  |      99      |

---

74912   |   namez  |   75  | 2017-01-04 04:07:55  |      99      |
my_uid  |  my_name |   75  | 2017-01-04 13:07:45  |     100      |
86766   |   namen  |   75  | 2017-01-04 14:21:47  |     101      |
67976   |   namey  |   74  | 2017-01-04 16:22:23  |     102      |
34676   |   nameu  |   74  | 2017-01-04 17:33:32  |     103      |
86236   |   namei  |   73  | 2017-01-04 18:11:09  |     104      |
98636   |   nameo  |   73  | 2017-01-04 19:21:47  |     105      |
14326   |   namep  |   73  | 2017-01-04 20:33:22  |     106      |
45333   |   namet  |   72  | 2017-01-04 20:44:12  |     107      |
33323   |   namer  |   72  | 2017-01-04 21:34:26  |     108      |

my problem is when using multiple query, it is still the same as doing 3 different query since I have 3 queries, how can I combine it as one? without using union or union all?
and on the 3rd query how can I set my starting point from my data?

Answer 1

In MySQL, view is like a function in other languages.

CREATE VIEW rankall AS SELECT * FROM record ORDER BY score;

SELECT * FROM rankall
    WHERE rank > (SELECT rank FROM rankall WHERE user_id = ID) - 11
    AND rank < (SELECT rank FROM rankall WHERE user_id = ID) + 11;

That's the basic idea, but no guarantee that above code will work :P

Answer 2

Try this

SELECT user_id, name, score FROM record
WHERE score BETWEEN 
CONVERT((SELECT score FROM record WHERE user_id = (SELECT user_id FROM record WHERE score = '100') ), INTEGER) - 10
AND
CONVERT((SELECT score FROM record WHERE user_id = (SELECT user_id FROM record WHERE score = '100') ), INTEGER) + 10
ORDER BY score DESC, date DESC

Answer 3

Possible way to do it, but there are issues which will need work. This gets the records either side (score wise) of the chosen user, along with the rank of that user. Then orders that by score and calculates the ranking. This does work normally, but will mess up if the chosen user is the top / bottom scorer. Would be possible to sort this, but not sure with real data it will be usable anyway. But will give you some ideas.

This is using user id 86474 as the one you are interested in, and just getting 1 either side - just to suit the test data you have given:-

SELECT user_id,
        name, 
        score, 
        date,
        def_rank - (@ranking := @ranking -1) AS rank
FROM
(
    SELECT *
    FROM
    (
        (SELECT r1.user_id,
                r1.name, 
                r1.score, 
                r1.date,
                sub0.def_rank
        FROM record r1
        INNER JOIN record r2 ON r2.user_id = 86476
        CROSS JOIN 
        (
            SELECT COUNT(*) def_rank
            FROM record r1
            INNER JOIN record r2 ON r2.user_id = 86476
            WHERE r1.score >= r2.score
        ) sub0
        WHERE r1.score >= r2.score
        ORDER BY score ASC
        LIMIT 2) 
        UNION
        (SELECT r1.user_id,
                r1.name, 
                r1.score, 
                r1.date,
                sub0.def_rank
        FROM record r1
        INNER JOIN record r2 ON r2.user_id = 86476
        CROSS JOIN 
        (
            SELECT COUNT(*) def_rank
            FROM record r1
            INNER JOIN record r2 ON r2.user_id = 86476
            WHERE r1.score >= r2.score
        ) sub0
        WHERE r1.score <= r2.score
        ORDER BY score DESC
        LIMIT 2)
    ) sub97
    ORDER BY score
) sub1
CROSS JOIN 
(
    SELECT @ranking := 2
) sub2

Answer 4

I finally got what I wanted to get, so I will just answer my own question

SELECT score, date FROM rank WHERE uid=your_user_id; //your score and date recorded
SELECT (count(*) + 1) AS rank FROM rank WHERE score > your_score OR (score = your_score AND date < date of your score recorded); //your rank
SELECT * FROM rank WHERE score > your_score OR (score = your_score AND date < date of your score recorded) ORDER BY score ASC, date DESC LIMIT 10; //10 users above my rank, in your output you have to reverse the order
SELECT * FROM rank WHERE score < your_score OR (score = your_score AND date > date of your score recorded) ORDER BY score DESC, date ASC LIMIT 10; //10 users below my rank

thank you for other users who replied :)