SQL count joins distinct by multiple columns

Question

我有一个问题，我需要在同一个 table 上多次加入并计算加入次数。

这里是数据库设置（SQL Fiddle ):

CREATE TABLE state
(
  [t_id] int,
  [true_id] int,
  [false_id] int,
  [msg] varchar(32) 
);

INSERT INTO state
  (t_id, true_id, false_id, msg)
VALUES
  (5, 6, 7, 'CASE_1'),
  (10, 11, 12, 'CASE_2'),
  (20, 21, 22, 'CASE_N'),
  (30, 31, 32, 'FOOOO');

CREATE TABLE step
(
      [id] int,
      [f_id] int,
      [state_type] int,
      [state_value] int
);

INSERT INTO step
    (id,f_id,state_type, state_value)
VALUES
    (1, 5, 5, 7),
    (2, 5, 5, 7),
    (3, 5, 5, 6),

    (4, 5, 10, 12),
    (5, 5, 10, 12),
    (6, 5, 10, 11),

    (7, 6, 10, 12),
    (8, 6, 10, 12),

    (9, 7, 20, 21),
    (10, 7, 20, 21),

    (11, 7, 30, 32),
    (12,7, 30, 31);

这是我当前的查询：

SELECT state.msg, 
  COUNT(state_true.true_id) AS Trues,
  COUNT(state_false.false_id) AS Falses
FROM state
  INNER JOIN step ON state.t_id = step.state_type
  LEFT OUTER JOIN state AS state_true ON step.state_value = state_true.true_id
  LEFT OUTER JOIN state AS state_false ON step.state_value = state_false.false_id
GROUP BY state.msg, step.f_id

这是我得到的：

msg     Trues   Falses
CASE_1  1       2
CASE_2  1       2
CASE_2  0       2
CASE_N  2       0
FOOOO   1       1

这是我需要的：

msg     Trues   Falses
CASE_1  1       0
CASE_2  1       1
CASE_N  1       0
FOOOO   1       0

解释：

我需要计算每个 state_type 和 f_id 组合有多少正确和失败。

有 6 个条目 f_id = 5 -> (1,2,3,4,5,6)。如果有条目具有相同的 (f_id,state_type) 组合，则只计算最后一个。因此，对于 f_id 5，不应将条目 1,2,4,5 计入计数，因为它们被 3 和 6 覆盖。

所以在处理前 6 个条目后应该有 CASE_1 true => 1 false => 0 和 CASE_2 true => 1 false => 0

__ 编辑 __

 TABLE step:

(1, 5, 5, 7),  -- do not count
(2, 5, 5, 7),  -- do not count
(3, 5, 5, 6),  -- this is the last entry with 
               -- (f_id,state_type) => (5,5) combination. 
               -- it overwrites the 2 previous ones => count CASE_1 true

(4, 5, 10, 12), -- do not count
(5, 5, 10, 12), -- do not count
(6, 5, 10, 11), -- count CASE_2 true

(7, 6, 10, 12), -- do not count
(8, 6, 10, 12), -- count CASE_2 false

(9, 7, 20, 21),  -- do not count
(10, 7, 20, 21), -- count CASE_N false

(11, 7, 30, 32), -- do not count
(12,7, 30, 31);  -- count FOOOO true

Answer 1

我不确定我是否完全理解您的意图，但也许下面的查询就是您想要的？鉴于您的示例数据，它似乎产生了正确的输出。

SELECT state.msg, 
  SUM(CASE WHEN true_id  = state_value THEN 1 ELSE 0 END) AS Trues,
  SUM(CASE WHEN false_id = state_value THEN 1 ELSE 0 END) AS Falses
FROM state
JOIN step ON state.t_id = step.state_type
JOIN (SELECT MAX(id) mid FROM step GROUP BY f_id, state_type) a ON a.mid = step.id
GROUP BY state.msg;

请试一试。如果我误解了我会删除答案。

SQL count joins distinct by multiple columns

SQL count joins distinct by multiple columns

sql

join

count

left-join

sql-server-2008