Question

我有以下简化架构：

CREATE TABLE [file]
( 
    id UNIQUEIDENTIFIER NOT NULL, 
    uri NVARCHAR(MAX) NOT NULL, 
    CONSTRAINT PK_file PRIMARY KEY (id ASC) ON [PRIMARY] 
)

CREATE TABLE [property]
( 
    id UNIQUEIDENTIFIER NOT NULL, 
    name NVARCHAR(MAX) NOT NULL, 
    CONSTRAINT PK_property PRIMARY KEY (id ASC) ON [PRIMARY] 
)

CREATE TABLE [metadata]
( 
    fileid UNIQUEIDENTIFIER NOT NULL,
    propertyid UNIQUEIDENTIFIER NOT NULL,
    value NVARCHAR(MAX) NOT NULL, 
    CONSTRAINT PK_metadata PRIMARY KEY (fileid, propertyid ASC) ON [PRIMARY] 
)

[fileid] FK到[file].[id]和[propertyid] FK到[property].[id]的位置。假设[properyid]为CLUSTERED且[value]为NON-CLUSTERED。

我想选择与某组元数据匹配的所有文件;例如，每个文件的属性值对都为size = 1 Kb和extension = 'txt'。

我提出的查询，例如指定了三个属性，是：

SELECT [uri] FROM [file] WHERE [id] IN (
    SELECT a.[fileid] FROM (
        SELECT COUNT(*) [count], [fileid] FROM [metadata]
        WHERE ([propertyid] = '597ddddf-afd2-414f-9774-36f067038064' AND 
                [value] = N'moo') OR
              ([propertyid] = 'd83d12de-e4bc-4d18-be12-743504df3318' AND 
                [value] = N'foo') OR
              ([propertyid] = 'c00c3966-5034-4818-8567-abd660f37f15' AND 
                [value] = N'boo')
        GROUP BY [fileid]
    ) a
    WHERE a.[count] = 3
)

我可以做得更好吗？

Answer 1

;WITH propertylist AS (
  SELECT propertyid = '597ddddf-afd2-414f-9774-36f067038064', value = N'moo' UNION ALL
  SELECT propertyid = 'd83d12de-e4bc-4d18-be12-743504df3318', value = N'foo' UNION ALL
  SELECT propertyid = 'c00c3966-5034-4818-8567-abd660f37f15', value = N'boo'
)
SELECT uri
FROM file
WHERE id IN (
  SELECT m.fileid
  FROM metadata m
    INNER JOIN propertylist p ON m.propertyid = p.propertyid AND m.value = p.value
  GROUP BY m.fileid
  HAVING COUNT(*) = (SELECT COUNT(*) FROM propertylist)
)

Answer 2

也许是这样的？

SELECT 
   [uri] 
FROM 
   [file] 
WHERE 
   EXISTS(
        SELECT 
           NULL
        FROM 
           [metadata]
        WHERE 
              ([propertyid] = '597ddddf-afd2-414f-9774-36f067038064' AND 
                [value] = N'moo') OR
              ([propertyid] = 'd83d12de-e4bc-4d18-be12-743504df3318' AND 
                [value] = N'foo') OR
              ([propertyid] = 'c00c3966-5034-4818-8567-abd660f37f15' AND 
                [value] = N'boo') AND 
              [File].[id] = [metadata].[fileid]
        GROUP BY 
           [fileid]
        HAVING 
           COUNT(*) = 3
       )

Answer 3

我可能会遗漏一些东西，但是如果你想要满足所有3个条件，为什么不每次只使用“AND”而不是计算属性集的数量？您将避免不必要的GROUPING。

我会写：

    SELECT [uri] FROM [file] 
    WHERE EXISTS ( SELECT [fileid] FROM [metadata]
           WHERE [propertyid] = '597ddddf-afd2-414f-9774-36f067038064'
             AND [value] = N'moo'
             AND [propertyid] = 'd83d12de-e4bc-4d18-be12-743504df3318' 
             AND [value] = N'foo'
             AND [propertyid] = 'c00c3966-5034-4818-8567-abd660f37f15'
             AND [value] = N'boo'
             AND [file].[id] = [metadata].[fileid])

用于查找与一组属性匹配的行的SQL查询

3 个答案: