多个连接和聚合函数产生不正确的总数

Question

我有 4 张桌子。

LAYERS 

id(pk)        location
-----------------------
  1            Bristol
  2            London

FEATURES

fid(pk)      layer_id(fk)     category
-----------------------------------------
  1.001        1                  Rank1
  1.002        1                  Rank2
  1.003        1                  Rank1
  2.001        2                  Rank3
  2.002        2                  Rank1

TOWN_STATUS 

id(pk)     town_id(fk)      fid(fk)      status
--------------------------------------------------
  1            5             1.001       AT_RISK
  2            6             1.001       DANGER
  3            19            2.002       DANGER
  4            23            2.002       DANGER
  5            24            2.002       AT_RISK
  6            30            2.002       DANGER

LANDMARK_STATUS 

id(pk)     landmark_id(fk)      fid(fk)      status
--------------------------------------------------
  1            1             1.002       DANGER
  2            10            1.002       DANGER
  3            11            2.002       DANGER
  4            13            2.002       AT_RISK

FEATURES 表中的每条记录都链接到LAYERS 表中的一条记录。

空间触发器对要素表进行操作，以将记录插入到 TOWN_STATUS 和 LANDMARK_STATUS 表中。

TOWN_STATUS 和 LANDMARK_STATUS 表中的状态列只能具有 2 个值之一，例如'AT RISK' 或 'DANGER'

我想创建一个表格来计算一个类别中的特征数量以及有多少'AT RISK'或'DANGER'statuses 该类别导致并由特定层过滤。例如，表格如下所示：

    category   feature_num    at_risk_num   danger_num
--------------------------------------------------------
    Rank1         3                 1         5
    Rank2         1                 0         3
    Rank3         1                 0         0 

目前我有一个看起来像这样的 (Postgre)SQL 查询

SELECT feature.category,
    count(feature.category) AS count,
    sum(
        CASE
            WHEN town_status.status = 'AT_RISK' OR landmark_status.status 'AT_RISK' THEN 1
            ELSE NULL
        END) AS at_risk_num,
    sum(
        CASE
            WHEN town_status.status::text = 'DANGER' OR landmark_status.status = 'DANGER' THEN 1
            ELSE NULL
        END) AS danger_num
   FROM features
     LEFT JOIN layer ON layer.id = feature.layer_id
     LEFT JOIN town_status ON town_status.fid = feature.fid
     LEFT JOIN landmark_status ON landmark_status.fid = feature.fid
  WHERE layer.location = 'Bristol'
  GROUP BY features.category
  ORDER BY features.category;

但是，当我这样做时，feature_num 列的总数不正确。由于this answer，我知道这与连接有关，但我不知道如何根据特征类别形成最后两列。

Answer 1

聚合之前加入：

SELECT f.category,
       COALESCE(ts.cnt, 0) + (COALESCE(ls.cnt, 0) as count,
       COALESCE(ts.at_risk, 0) + (COALESCE(ls.at_risk, 0) as at_risk,
       COALESCE(ts.danger, 0) + (COALESCE(ls.danger, 0) as danger
FROM features f LEFT JOIN
     layer l
     ON l.id = f.layer_id LEFT JOIN
     (SELECT ts.fid, COUNT(*) as cnt,
             COUNT(*) FILTER (WHERE ts.status = 'AT_RISK') as at_risk,
             COUNT(*) FILTER (WHERE ts.status = 'DANGER') as danger
      FROM town_status ts
      GROUP BY ts.fid
     ) ts
     ON town_status.fid = f.fid
     (SELECT ls.fid, COUNT(*) as cnt,
             COUNT(*) FILTER (WHERE ls.status = 'AT_RISK') as at_risk,
             COUNT(*) FILTER (WHERE ls.status = 'DANGER') as danger
      FROM landmark_status ls
      GROUP BY ls.fid
     ) ls
     ON ls.fid = f.fid;
WHERE l.location = 'Bristol'
GROUP BY f.category
ORDER BY f.category;