合并语句临时表 - 如何将数据定向到本地数组？答案

【问题标题】：merge statement temporary table - how to direct data into a local array?合并语句临时表 - 如何将数据定向到本地数组？
【发布时间】：2014-02-20 04:36:18
【问题描述】：

我们在调试过程中遇到了困难，因为我们退休的前任有一个查询语句，该语句在上传到数据库之前将提取的数据放入一系列临时表中。它现在不起作用（我们在 TestPatterns 表中添加了项目，并为 tp7 而不是 tp21 运行，因此数据可能会有所不同）并且我们在调试它时遇到了困难。如果我们能以某种方式打印或访问放在临时表中的数据，也许我们可以更好地调试它。关于如何更好地调试这个的任何想法，也许看到数据？我们可以打印我们的第一个 select/where 语句，但不能打印它在执行 qry 时提取的数据。另外还有我们所做的 Source/Target sql 合并。 在我们执行 $qry1d 之前，我们知道一切似乎都可以正常工作。

我在网上查找了如何查看此信息，并查看了 sql 分析器，但我担心这些语句太复杂而无法破译分析器的输出。当我尝试使用http://youtu.be/mJ8Dyv4Uk6E 进行分析时，对于简单的选择前 1000 行，它说：

 exec sp_executesql N'SELECT
clmns.name AS [Name]
FROM
sys.tables AS tbl
INNER JOIN sys.all_columns AS clmns ON clmns.object_id=tbl.object_id
WHERE
(CAST(clmns.is_sparse AS bit)=@_msparam_0)and((tbl.name=@_msparam_1 and SCHEMA_NAME(tbl.schema_id)=@_msparam_2))
ORDER BY
clmns.column_id ASC',N'@_msparam_0 nvarchar(4000),@_msparam_1 nvarchar(4000),@_msparam_2 nvarchar(4000)',@_msparam_0=N'0',@_msparam_1=N'Measurements',@_msparam_2=N'dbo'

失败了

MERGE 语句尝试更新或删除同一行不止一次。当目标行匹配多个源时会发生这种情况排。 MERGE 语句不能更新/删除目标的同一行表多次。细化 ON 子句以确保目标行最多匹配一个源行，或使用 GROUP BY 子句进行分组源行。

这是几个 php sql 合并语句：

$qry = 'SELECT "PrintSamples"."PrintSampleID", "PrintSamples"."TestPatternName", "PrintSamples"."PrintCopyID", 
"DigitalImages"."CaptureTime", "PrintSampleAnalyses"."psaTicket", "Measurements"."MeasurementID", "Measurements"."MeasurementUuid", 
SUBSTRING("OperatorLastName",1,1) AS "lastInitial", SUBSTRING("OperatorFirstName",1,1) AS "firstInitial",
"ParameterValues"."ParameterID", "ParameterName", "TargetName", "ParameterValues"."ParameterValue"
FROM "ParameterValues"
LEFT JOIN "Measurements" ON "ParameterValues"."MeasurementID"="Measurements"."MeasurementID" 
LEFT JOIN "PrintSampleAnalyses" ON "PrintSampleAnalyses"."psaID"="Measurements"."psaID"
LEFT JOIN "DigitalImages" ON "DigitalImages"."ImageID"="PrintSampleAnalyses"."ImageID" 
LEFT JOIN "PrintSamples" ON "DigitalImages"."PrintSampleID"="PrintSamples"."PrintSampleID"
LEFT JOIN "Sessions" ON "Sessions"."SessionID"="PrintSampleAnalyses"."SessionID" 
LEFT JOIN "Operators" ON "Operators"."OperatorID"="Sessions"."OperatorID"
LEFT JOIN "ParameterNames" ON "ParameterNames"."ParameterID"="ParameterValues"."ParameterID"
LEFT JOIN "Targets" ON "Targets"."TargetID"="Measurements"."TargetID"
WHERE ('; 

//----------------------------------------------------------------------------------------------------
// The two statements where added to make AvgGhostValAsSir77 and MaxNegGhostingValAsSir77 work.
// After much testing it was found that the query did not recognize these to ParameterNames in the 
// Postgres database. It was never discovered why when testing the ParameterName to be equal to
// AvgGhostValAsSir77 or MaxNegGhostingValAsSir77 the query fails but using the LIKE statement 
// corrected the problem. It is possible that the names contain a hidden character or space that
// caused the proble. More investigation will need to be done to find a better resolution to this 
// strange problem. On 8/1/13 - It was found that the two Parameters AvgGhostValAsSir77 and MaxNegGhostingValAsSir77
// have a trailing space to thier names in the Postgres database and that's why these two parameters
// were NOT working. It was decided instead of having IQAF people modify the database that the two
// statements using the LIKE will remain in place. The file was modified on this date 8/1/13.
//----------------------------------------------------------------------------------------------------
// only take values that actually get reported on the dashboard
// this list comes from the "MeasurementNames" table
foreach ($measurementIDs as $mid){
    if($mid[0] == "AvgGhostValAsSir77") $qry .= '(("ParameterName" LIKE ' . "'%AvgGhostValAsSir77%'" . ') AND ("TargetName"=' . "'" . $mid[1] . "')) OR "; 
    else if($mid[0] == "MaxNegGhostingValAsSir77") $qry .= '(("ParameterName" LIKE ' . "'%MaxNegGhostingValAsSir77%'" . ') AND ("TargetName"=' . "'" . $mid[1] . "')) OR "; 
    else $qry .= '(("ParameterName"=' . "'" . $mid[0] . "'" . ') AND ("TargetName"=' . "'" . $mid[1] . "')) OR ";
}    
$qry = substr ($qry, 0, -4);
$qry .= ") ";

if (isset($captureTime)){   // used for incremental updates
    $qry .= ' AND ("CaptureTime">' . "'" . $captureTime . "')";
}
// steve invalid reading code is -99999. 
$qry .= ' AND ("ParameterValues"."ParameterValue" != -99999) ORDER BY "PrintSampleID", "MeasurementID"';

$actionString = '$action';

$qryCreate = "CREATE TABLE #tmpMeasurementTable (TestGUID uniqueidentifier, 
            IQAFid uniqueidentifier, pqID int, MeasurementID int,
            EvaluationDate datetime, EvaluatorID int, 
            TestUnitID int, TestUnitCountID int,
            TestPatternID int, ColorID int,
            TargetID int, ParameterID int,
            ParameterValue real)
            CREATE TABLE #MergeOutput (ActionType nvarchar(10))";
//start putting data into measurement tables
$qry1a= "INSERT INTO #tmpMeasurementTable VALUES ";
//put 1a data in MeasurementData
$qry1b = "
    MERGE INTO MeasurementData AS Target
    USING #tmpMeasurementTable AS Source 
    ON Target.pqID=Source.pqID 
        AND Target.MeasurementID=Source.MeasurementID  
        AND Target.TargetID=Source.TargetID  
        AND Target.ParameterID=Source.ParameterID 
        AND Target.TestPatternID=Source.TestPatternID 
        AND Target.IQAFMeasurementID=Source.IQAFid  
    WHEN MATCHED THEN
        UPDATE SET Target.ParameterValue = Source.ParameterValue,
                Target.TestUnitID=Source.TestUnitID, 
                Target.TestUnitCountID=Source.TestUnitCountID,
                Target.EvaluationDate=Source.EvaluationDate,  
                Target.EvaluatorID=Source.EvaluatorID  
    WHEN NOT MATCHED BY Target THEN
        INSERT (TestGUID, IQAFMeasurementID, pqID, 
                MeasurementID, EvaluationDate, EvaluatorID, 
                TestUnitID, TestUnitCountID, TestPatternID, ColorID,
                TargetID, ParameterID, ParameterValue)
        VALUES (Source.TestGUID, Source.IQAFid, Source.pqID, 
                Source.MeasurementID, Source.EvaluationDate, Source.EvaluatorID, 
                Source.TestUnitID, Source.TestUnitCountID,
                Source.TestPatternID, Source.ColorID, 
                Source.TargetID, Source.ParameterID, Source.ParameterValue)
    OUTPUT 
        $actionString INTO #MergeOutput;
        DROP TABLE #tmpMeasurementTable";

$qryOutput = "SELECT ActionType, COUNT(ActionType)AS [count] FROM #MergeOutput GROUP BY ActionType";
$qryCleanup = "DROP TABLE #MergeOutput";

$qry1c = "IF EXISTS (SELECT * FROM tempdb.dbo.sysobjects WHERE ID = OBJECT_ID(N'tempdb..#MergeOutput') AND xtype='U')
            DROP TABLE #MergeOutput
    CREATE TABLE #MergeOutput (ActionType nvarchar(10))
    SELECT TestIndex AS TestID, TestID AS TestNumber, MeasurementData.TestGUID, pqID, TestUnitID, TestUnitCountID, TestPatternID,
  ColorID, MeasurementData.MeasurementID, TargetID, ParameterID,
  CAST(ROUND(AVG(ParameterValue*Multiplier), 2)AS DECIMAL(18,2)) AS Value
  INTO #tmpTable
  FROM MeasurementData
  LEFT JOIN Measurements ON Measurements.MeasurementID=MeasurementData.MeasurementID 
  LEFT JOIN Tests ON Tests.TestGUID=MeasurementData.TestGUID AND Tests.PiggybackID IS NULL 
  WHERE MeasurementData.TestGUID='" . $TestGUID . "' AND pqID>=" . $startPQid . " AND pqID<=" .$endPQid;


if (isset($captureTime))    // used for incremental updates
{
    $qry1c .= " AND EvaluationDate>'" . makeDateTime($captureTime) . "'";
}

$qry1c .= " GROUP BY TestIndex, TestID, MeasurementData.TestGUID, pqID, TestUnitID, TestUnitCountID, TestPatternID,
  ColorID, MeasurementData.MeasurementID, TargetID, ParameterID
  ORDER BY ColorID, TestPatternID, TestUnitCountID, MeasurementData.MeasurementID, TargetID, ParameterID;
  SELECT @@ROWCOUNT AS rows INTO #DashboardRows;";

//put temporary table in dashboard table
$qry1d = ";
    MERGE INTO DashboardData AS Target
    USING #tmpTable AS Source
    ON Target.pqID=Source.pqID 
        AND Target.MeasurementID=Source.MeasurementID  
        AND Target.TargetID=Source.TargetID  
        AND Target.ParameterID=Source.ParameterID 
        AND Target.TestPatternID=Source.TestPatternID 
    WHEN MATCHED THEN
        UPDATE SET Target.ParameterValue=Source.Value,
                Target.TestUnitID=Source.TestUnitID, 
                Target.TestUnitCountID=Source.TestUnitCountID 
    WHEN NOT MATCHED BY Target THEN
        INSERT (TestGUID, pqID, 
                MeasurementID, 
                TestUnitID, TestUnitCountID, TestPatternID, ColorID,
                TargetID, ParameterID, ParameterValue,
                TestNumber, TestIndex)
        VALUES (Source.TestGUID, Source.pqID,
                Source.MeasurementID,
                Source.TestUnitID, Source.TestUnitCountID,
                Source.TestPatternID, Source.ColorID, 
                Source.TargetID, Source.ParameterID, Source.Value, 
                Source.TestNumber, Source.TestID)
     OUTPUT 
        $actionString INTO #MergeOutput;
    DROP TABLE #tmpTable";

查询是这样执行的：

 $result = $ms_conn->query($qry1c);     $recordCount = $ms_conn->fetchOne
 ("SELECT rows FROM #DashboardRows");   //dataLog ("OLAP database " .
 $recordCount . " records to process", true);   $result =
 $ms_conn->query($qry1d); dataLog ("end dashboard query");

【问题讨论】：

标签： php sql merge olap

【解决方案1】：

要查看中间表，请从表名中删除第一个字符 #。这将使表成为非临时表，因为 SQL Server 的约定是，如果表的名称以此字符开头，则表是临时的。您可能还想从代码中删除 drop table 语句。然后，您可以像查看任何其他表一样查看这些中间表。

当然，为了能够重新运行代码，您必须手动删除表。

【讨论】：

感谢@FrankPl！我们会试一试。如果我们想手动删除表，是否需要进入一个新的 sql 查询并输入，例如“DROP TABLE #MergeOutput”？
@Michele 是的，可能省略了“#”。但是您也可以通过 Management Studio 中的 GUI 执行此操作：在对象资源管理器中（通常在左侧），打开数据库，打开“表”节点，选择表（您可能必须选择“表”节点点击F5 刷新表格列表以查看它），然后按 Del 键。然后确认删除。
我很难决定要删除哪个#，因为有这么多并且它们在下一个语句中使用，所以如果我删除#，它将使其余代码不工作。
@Michele 删除所有的 '#'s。
@Michele 好吧，如果你在CREATE TABLE 中省略了#，你也必须在MERGE ... USING 中省略它，因为表名不同。它必须是一致的。