查找 DISTINCT 缺失的 SQL 依赖项答案

【问题标题】：Find DISTINCT Missing SQL Dependencies查找 DISTINCT 缺失的 SQL 依赖项
【发布时间】：2016-12-02 15:15:24
【问题描述】：

我有返回数据库中所有无效依赖项的脚本。但是此脚本返回许多重复项。所以我只想看到DISTINCT 结果。

    /*
modified version of script from http://michaeljswart.com/2009/12/find-missing-sql-dependencies/
Added columns for object types &amp; generated refresh module command...
filter out user-define types: http://stackoverflow.com/questions/2330521/find-broken-objects-in-sql-server
*/

SELECT TOP (100) PERCENT
    QuoteName(OBJECT_SCHEMA_NAME(referencing_id)) + '.' + QuoteName(OBJECT_NAME(referencing_id)) AS [this Object...],
        o.type_desc,
    ISNULL(QuoteName(referenced_server_name) + '.', '')
    + ISNULL(QuoteName(referenced_database_name) + '.', '')
    + ISNULL(QuoteName(referenced_schema_name) + '.', '')
    + QuoteName(referenced_entity_name) AS [... depends ON this missing entity name]
    ,sed.referenced_class_desc
    ,case when o.type_desc in( 'SQL_STORED_PROCEDURE' ,'SQL_SCALAR_FUNCTION' ,'SQL_TRIGGER' ,'VIEW')
          then 'EXEC sys.sp_refreshsqlmodule ''' + QuoteName(OBJECT_SCHEMA_NAME(referencing_id)) + '.' + QuoteName(OBJECT_NAME(referencing_id)) + ''';'
          else null
       end as [Refresh SQL Module command]
FROM sys.sql_expression_dependencies as sed
LEFT JOIN sys.objects o
            ON sed.referencing_id=o.object_id
WHERE (is_ambiguous = 0)
    AND (OBJECT_ID(ISNULL(QuoteName(referenced_server_name) + '.', '')
    + ISNULL(QuoteName(referenced_database_name) + '.', '')
    + ISNULL(QuoteName(referenced_schema_name) + '.', '')
    + QuoteName(referenced_entity_name)) IS NULL)
    AND NOT EXISTS
       (SELECT * 
        FROM sys.types 
        WHERE types.name = referenced_entity_name 
        AND types.schema_id = ISNULL(SCHEMA_ID(referenced_schema_name), SCHEMA_ID('dbo'))
       )
ORDER BY [this Object...],
[... depends ON this missing entity name]
go

这是脚本。

我尝试添加GROUP BY [... depends ON this missing entity name]，但出现以下错误：

无效的列名'...取决于这个缺失的实体名'。

我尝试将聚合函数添加到结果中的每一列仅用于测试，但仍然出现相同的错误。

【问题讨论】：

SELECT DISTINCT 或 GROUP BY ISNULL(QuoteName(referenced_server_name) + '.', '') + ISNULL(QuoteName(referenced_database_name) + '.', '') + ISNULL(QuoteName(referenced_schema_name) + '.', '') + QuoteName(referenced_entity_name)
@Matt，是的，GROUP BY 有帮助。总是认为 GROUP BY by alias 应该可以工作......
GROUP BY 工作了吗？
@Matt，好像是的。不再显示错误
抱歉，您不能在 group by 子句中使用别名。这是The Logical Processing Order 的结果。

标签： sql sql-server sql-server-2012 group-by

【解决方案1】：

GROUP BY 实际字段名不是别名。

SELECT TOP (100) PERCENT
    QuoteName(OBJECT_SCHEMA_NAME(referencing_id)) + '.' + QuoteName(OBJECT_NAME(referencing_id)) AS [this Object...],
        o.type_desc,
    ISNULL(QuoteName(referenced_server_name) + '.', '')
    + ISNULL(QuoteName(referenced_database_name) + '.', '')
    + ISNULL(QuoteName(referenced_schema_name) + '.', '')
    + QuoteName(referenced_entity_name) AS [... depends ON this missing entity name]
    ,sed.referenced_class_desc
    ,case when o.type_desc in( 'SQL_STORED_PROCEDURE' ,'SQL_SCALAR_FUNCTION' ,'SQL_TRIGGER' ,'VIEW')
          then 'EXEC sys.sp_refreshsqlmodule ''' + QuoteName(OBJECT_SCHEMA_NAME(referencing_id)) + '.' + QuoteName(OBJECT_NAME(referencing_id)) + ''';'
          else null
       end as [Refresh SQL Module command]
FROM sys.sql_expression_dependencies as sed
LEFT JOIN sys.objects o
            ON sed.referencing_id=o.object_id
WHERE (is_ambiguous = 0)
    AND (OBJECT_ID(ISNULL(QuoteName(referenced_server_name) + '.', '')
    + ISNULL(QuoteName(referenced_database_name) + '.', '')
    + ISNULL(QuoteName(referenced_schema_name) + '.', '')
    + QuoteName(referenced_entity_name)) IS NULL)
    AND NOT EXISTS
       (SELECT * 
        FROM sys.types 
        WHERE types.name = referenced_entity_name 
        AND types.schema_id = ISNULL(SCHEMA_ID(referenced_schema_name), SCHEMA_ID('dbo'))
       )
GROUP BY ISNULL(QuoteName(referenced_server_name) + '.', '') + ISNULL(QuoteName(referenced_database_name) + '.', '') + ISNULL(QuoteName(referenced_schema_name) + '.', '') + QuoteName(referenced_entity_name)
ORDER BY [this Object...],
[... depends ON this missing entity name]
go

【讨论】：

【解决方案2】：

您可以在查询的前面使用 APPLY 运算符来访问别名，例如（寻找oa.x oa.y oa.z）

SELECT
        o.type_desc, sed.referenced_class_desc
      , oa.x , oa.y , oa.z
FROM sys.sql_expression_dependencies AS sed
  LEFT JOIN sys.objects o ON sed.referencing_id = o.object_id
  OUTER APPLY (
      SELECT
              ISNULL(QUOTENAME(sed.referenced_server_name) + '.', '')
            + ISNULL(QUOTENAME(sed.referenced_database_name) + '.', '')
            + ISNULL(QUOTENAME(sed.referenced_schema_name) + '.', '')
            + QUOTENAME(sed.referenced_entity_name)                    AS x
          ,  QUOTENAME(OBJECT_SCHEMA_NAME(referencing_id)) 
            + '.' 
            + QUOTENAME(OBJECT_NAME(referencing_id))                   AS y
          , CASE
                  WHEN o.type_desc IN ('SQL_STORED_PROCEDURE', 'SQL_SCALAR_FUNCTION', 'SQL_TRIGGER', 'VIEW') 
                    THEN 'EXEC sys.sp_refreshsqlmodule ''' 
                      + QUOTENAME(OBJECT_SCHEMA_NAME(referencing_id)) 
                      + '.' + QUOTENAME(OBJECT_NAME(referencing_id)) 
                      + ''';'
                  ELSE NULL
            END                                                        AS z
      ) AS OA
WHERE (is_ambiguous = 0)
      AND (oa.x IS NULL)
      AND NOT EXISTS (
            SELECT
                  *
            FROM sys.types
            WHERE types.name = referenced_entity_name
                  AND types.schema_id = ISNULL(SCHEMA_ID(referenced_schema_name), SCHEMA_ID('dbo'))
            )
GROUP BY
        o.type_desc, sed.referenced_class_desc
      , oa.x , oa.y , oa.z
;

【讨论】：