Sqlserver
 sql >> Base de Dados >  >> RDS >> Sqlserver

Como converter a consulta baseada em CURSOR para baseada em SET

INSERT INTO T_PROJECTGROUPSDATA (uploadID, fieldA,fieldB,......fieldN )
SELECT t.ID,p.fieldA,mg.fieldB,......mgc.fieldN
FROM T_Projects mp
INNER JOIN T_UPLOADS t mp.ID = t.project_savix_ID
INNER JOIN  SG_Dynamic_Forms..v_projects p ON p.LegacyProjectId = mp.ID 
INNER JOIN Savix_Service_Group..Groups mg ON mg.Grp_Project = p.ProjectID 
INNER JOIN Savix_Service_Group..Group_Cycle mgc ON mgc.Gc_GroupID = mg.Grp_ID
INNER JOIN Savix_Service_Group..Group_Data mgd ON mgd.Gd_CycleID = mgc.Gc_ID
LEFT JOIN Savix_Service_Trainers..Trainers me ON me.Tr_ID = mgc.Gc_MonitoredBy
LEFT JOIN Savix_Service_Dictionaries..Dictionary mgt ON mgt.Dny_ID = me.Tr_Type 
WHERE mp.SyncMode = 1
AND t.[STATUS] = 3 AND t.UPLOADFILENAME = 'Automatic upload from web MIS'
AND mg.Grp_IsDeleted = 0 AND mgd.Gd_IsDeleted != 1


UPDATE t SET 
TOTALEXPENDITURES = CASE WHEN DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,mp.EndDate) != 0 
    THEN p.Cost_Until_Today*DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,dbo.fn_GetEndOfPeriod(t.Period_ID))/DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,mp.EndDate)/dbo.fn_RateAtDate(p.PROJECTCURRENCY_ID,dbo.fn_GetEndOfPeriod(t.Period_ID))
    ELSE 0 
    END,
TotalExpendituresNative = CASE WHEN DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,mp.EndDate) != 0 
    THEN p.Cost_Until_Today*DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,dbo.fn_GetEndOfPeriod(t.Period_ID))/DATEDIFF(d,mp.PROJECTESTABLISHEDDATE,mp.EndDate)
    ELSE 0 
    END
FROM T_UPLOADS t
JOIN T_Projects mp ON mp.ID = t.project_savix_ID AND mp.SyncMode = 1
WHERE t.[STATUS] = 3 AND t.UPLOADFILENAME = 'Automatic upload from web MIS'

Para explicar como cheguei a isso, aqui está uma versão modificada do código abaixo. Não posso garantir que consegui tudo, tive que fazer algumas suposições. Faltou dados de amostra e todas as tabelas savix estão faltando. Esta é a melhor resposta que posso encontrar, dadas as informações e limitando meu tempo a menos de 4 horas investidas. Eu poderia fazer muito mais, mas você teria que me dar mais do que fama na internet.
--ALTER PROCEDURE [dbo].[proc_Upload]  WITH RECOMPILE 
--as
set NoCount on

DECLARE
    @StartTime      datetime,
    @EndTime        datetime,
    @DataID         uniqueidentifier,
    @CollectionDate datetime,
    @Status         int,
    @PeriodID       int,
    @EndDate        datetime,
    @GroupID        uniqueidentifier,
    @ProjectID      INT,
    @FAID           uniqueidentifier,
    @UploadID       int ,
    @Createdate datetime, 
    @MINIDprojects INT,
    @MAXIDprojects INT,
    @MINIDdatasets INT,
    @MAXIDdatasets INT,
    @MINIDperiods INT,
    @MAXIDperiods INT

É difícil trabalhar e testar cursores, então eu os substituí (projects, projects1, datasets e period) por tabelas temporárias contendo uma identidade e passei por elas.
IF OBJECT_ID('tempdb..#projects') IS NOT NULL DROP TABLE #projects
IF OBJECT_ID('tempdb..#projects1') IS NOT NULL DROP TABLE #projects1
IF OBJECT_ID('tempdb..#datasets') IS NOT NULL DROP TABLE #datasets
IF OBJECT_ID('tempdb..#period') IS NOT NULL DROP TABLE #period

CREATE TABLE #projects
(
    [ProjectID] [INT],
    [Title] [varchar](255) ,
    [currency] [int] ,
    [Cost_Until_Today] [float] ,
    [StartDate] [datetime] ,
    [EndDate] [datetime] ,
    [MisID] [uniqueidentifier] ,
    [SystemStatus] [int] ,
    [FacilitatingAgency] [uniqueidentifier] ,
    [SyncMode] [int] 
)

CREATE TABLE #projects1
(
    ID INT IDENTITY(1,1),
    [ProjectID] [INT],
    [FacilitatingAgency] [uniqueidentifier]
)

CREATE TABLE #datasets
(
    ID INT IDENTITY(1,1),
    Gd_ID [uniqueidentifier], 
    Grp_ID [uniqueidentifier], 
    Gd_CollectionDate DATETIME, 
    Gd_IsDeleted BIT, 
    Gd_CreateDate DATETIME
)

CREATE TABLE #period
(
    ID INT IDENTITY(1,1),
    IDPeriod INT,
    EndDate DATETIME
)

 INSERT #projects ( [ProjectID], [Title], {currency], [Cost_Until_Today], [StartDate], [EndDate], [MisID], [SystemStatus], [FacilitatingAgency], [SyncMode] )
 SELECT ID, PROJECTNAME, PROJECTCURRENCY_ID, Cost_Until_Today, PROJECTESTABLISHEDDATE, EndDate, MisID, 4, FacilitatingAgency, SyncMode 
 FROM [dbo].[T_PROJECTS] /*thsi is the source table where every projectIDs need to be processed*/

Projects contém todos os ID's ( projectID ) de T_Projects, os únicos campos que são usados ​​nesta tabela são o ID, projectcurrency, projectestablisheddate ( startdate ), enddate e syncmode ( deve ser =1, veremos isso mais tarde ). Podemos ignorar o status do sistema, pois é uma constante.

Os conjuntos de dados da tabela são descartados e carregados aqui. Nós só nos importamos com o gd_id da tabela Savix_Service_Group..Group_Data, mas ainda precisamos juntar as outras tabelas caso elas filtrem valores que não queremos. Essa lógica é usada novamente abaixo ao inserir registros.
IF exists ( select  * from    dbo.sysobjects where   id = object_id(N'[dbo].datasets]') and objectproperty(id, N'IsTable') = 1 ) DROP Table [dbo].datasets

SELECT mgd.Gd_ID, mg.Grp_ID, mgd.Gd_CollectionDate, mgd.Gd_IsDeleted, mgd.Gd_CreateDate, mg.Grp_Project, mg.Grp_IsDeleted , mg.Grp_Legacy_ID,  p.LegacyProjectId
INTO datasets
FROM Savix_Service_Group..Group_Data mgd 
        INNER JOIN Savix_Service_Group..Group_Cycle mgc ON mgc.Gc_ID = mgd.Gd_CycleID
        INNER JOIN Savix_Service_Group..Groups mg ON mg.Grp_ID = mgc.Gc_GroupID 
        inner join SG_Dynamic_Forms..v_projects p ON p.ProjectID = mg.Grp_Project

--DECLARE projects1 CURSOR LOCAL FOR Select distinct ProjectID, isnull([FacilitatingAgency], '00000000-0000-0000-0000-000000000000') from @projects P WHERE P.SystemStatus = 4 AND P.SyncMode = 1 
--/*First cursor - fetch the cursor from ProjectaTable*/

--OPEN projects1
--FETCH NEXT FROM projects1 INTO @ProjectID, @FAID
--WHILE @@FETCH_STATUS = 0

Aqui estamos inserindo em projects1 um valor distinto de id's de projeto. Sem mais dados, minha suposição é que o ID é distinto na tabela T_projects e, portanto, esta etapa é um pouco desnecessária, além de filtrar por SyncMode =1. Lembre-se de que SystemStatus é uma constante e a preenchemos com o valor '4', isso critérios podem ser removidos sem nenhum efeito.
INSERT INTO #projects1 ( ProjectID, FacilitatingAgency )
SELECT DISTINCT ProjectID, isnull([FacilitatingAgency], '00000000-0000-0000-0000-000000000000')  
FROM #projects p
WHERE SystemStatus = 4 AND SyncMode = 1

SELECT @MINIDprojects = MIN(ID), @MAXIDprojects = MAX(ID) FROM #projects1
WHILE @MINIDprojects < @MAXIDprojects + 1
BEGIN
--    BEGIN TRY
--        BEGIN TRAN

SELECT @ProjectID = ProjectID, @FAID = FacilitatingAgency FROM #projects1 WHERE ID = @MINIDprojects

        --DELETE FROM T_PROJECTGROUPSDATA WHERE T_PROJECTGROUPSDATA.UPLOAD_ID IN (SELECT ID FROM T_UPLOADS WHERE project_savix_ID = @ProjectID AND UPLOADFILENAME = 'Automatic upload from web MIS')

        --DECLARE datasets CURSOR LOCAL  FAST_FORWARD FOR SELECT Gd_ID, Grp_ID, Gd_CollectionDate, Gd_IsDeleted, Gd_CreateDate 
        --    FROM datasets
        --    WHERE LegacyProjectId = @ProjectID AND Grp_IsDeleted = 0 AND Gd_IsDeleted != 1 
        --    /*Second cursor - this will get the 'collectionDate'field from datasetsTable for every project fetched in above cursor and also get @dataID which is used to insert value in to other table-T_PROJECTGROUPSDATA*/
        --OPEN datasets

        --FETCH NEXT FROM datasets INTO @DataID, @GroupID, @CollectionDate, @Status, @Createdate
        --WHILE @@FETCH_STATUS = 0

Aqui inserimos em #datasets filtrando por T_Project.ID AND Savix_Service_Group..Groups.Grp_IsDeleted =0 AND Savix_Service_Group..Group_Data.Gd_IsDeleted !=1
        INSERT INTO #datasets ( Gd_ID, Grp_ID, Gd_CollectionDate, Gd_IsDeleted, Gd_CreateDate )
        SELECT Gd_ID, Grp_ID, Gd_CollectionDate, Gd_IsDeleted, Gd_CreateDate FROM datasets WHERE LegacyProjectId = @ProjectID AND Grp_IsDeleted = 0 AND Gd_IsDeleted != 1 

        SELECT @MINIDdatasets = MIN(ID), @MAXIDdatasets = MAX(ID) FROM #datasets

        WHILE @MINIDdatasets < @MAXIDdatasets + 1
        BEGIN

            SELECT @DataID = Gd_ID, @GroupID = Grp_ID, @CollectionDate = Gd_CollectionDate, @Status = Gd_IsDeleted, @Createdate = Gd_CreateDate FROM #datasets WHERE ID = @MINIDdatasets

            --DECLARE period CURSOR LOCAL  FAST_FORWARD FOR SELECT ID, dbo.fn_GetEndOfPeriod(ID) FROM T_PERIODS
            --/* dbo.fn_GetEndOfPeriod(ID) - this function will give the end of the date of that specifc quarter for any given date*/
            --    WHERE DATEDIFF(dd,@CollectionDate,dbo.fn_GetEndOfPeriod(ID)) >= 0
            --    ORDER BY [YEAR],[Quarter]
            --    /*Third Cursor - this will process the records from another table called period with above fetched @collectionDate*/

            --OPEN period
            --FETCH NEXT FROM period INTO @PeriodID, @EndDate
            --WHILE @@FETCH_STATUS = 0

Este é provavelmente o pior uso de um cursor aqui. Carregamos todos os períodos e os percorremos. No final só nos preocupamos com os períodos que estão na tabela T_UPLOADS.
            INSERT INTO #period ( IDPeriod, EndDate ) SELECT ID, dbo.fn_GetEndOfPeriod(ID) FROM T_PERIODS WHERE DATEDIFF(dd,@CollectionDate,dbo.fn_GetEndOfPeriod(ID)) >= 0
            SELECT @MINIDperiods = MIN(ID), @MAXIDperiods = MAX(ID) FROM #period

            WHILE @MINIDperiods < @MAXIDperiods + 1
            BEGIN

                --IF EXISTS (SELECT * FROM Savix_Service_Group..Group_Data mgd 
                --    INNER JOIN Savix_Service_Group..Group_Cycle mgc ON mgc.Gc_ID = mgd.Gd_CycleID
                --    WHERE mgc.Gc_GroupID = @GroupID
                --        AND DATEDIFF(dd,mgd.Gd_CollectionDate,@EndDate) >= 0 
                --        AND (mgd.Gd_CollectionDate > @CollectionDate ) 
                --        AND mgd.Gd_IsDeleted != 1) 

                --BEGIN
                --    BREAK
                --END

Aqui pulamos quaisquer períodos em que UPLOADFILENAME !='Upload automático do web MIS'
                --IF EXISTS (SELECT ID FROM T_UPLOADS u WHERE  u.project_savix_ID = @ProjectID AND u.PERIOD_ID = @PeriodID AND u.STATUS = 3 AND UPLOADFILENAME != 'Automatic upload from web MIS') 
                --BEGIN
                --    FETCH NEXT FROM period INTO @PeriodID, @EndDate
                --    CONTINUE
                --END

Com isso em mente, só nos importamos onde status =3 e UPLOADFILENAME ='Upload automático do web MIS'
                SET @UploadID = (SELECT ID FROM T_UPLOADS u WHERE  u.project_savix_ID = @ProjectID AND u.PERIOD_ID = @PeriodID AND u.[STATUS] = 3 AND UPLOADFILENAME = 'Automatic upload from web MIS')

                /*If T_uploads doesn't have appropirate period ID from cursor fetch then create a new entry in T_uploads with current projectID*/

                IF @UploadID IS NOT NULL
                        BEGIN

                            --declare @Project_ID_Legacy int = ISNULL((select distinct PROJECT_ID from T_UPLOADS where project_savix_ID = @ProjectID),@ProjectID)

Por que eles usam uma segunda tabela ( T_UPLOADFIRSTSTEP ) para gerar um registro onde eles usam apenas o id e depois excluem o registro usando esse ID na criação do registro T_Uploads está além do meu entendimento e me parece uma codificação terrível.
    --                            INSERT INTO T_UPLOADSFIRSTSTEP
    --                                   (PROJECT_ID
    --                                   --,UPLOADDATE
    --                                   --,UPLOADFILENAME
    --                                   --,UPLOADUSER_ID
    --                                   --,CURRENTSTEP
    --                                   ,[STATUS]
    --                                   ,Project_ID_MIS)
    --                             SELECT ISNULL((select distinct PROJECT_ID from T_UPLOADS where project_savix_ID = @ProjectID),@ProjectID),
    --                                    --GETDATE(),
    --                                    --'Automatic upload from web MIS',
    --                                    --2,
    --                                    --2,
    --                                    0,
    --                                    @ProjectID

A inserção em T_UPLOADS sempre falhará, pois há campos que não podem ser nulos que não estão na lista de inserções e não possuem um valor padrão atribuído a eles.
    --                            INSERT INTO T_UPLOADS ( ID, periodID, projectID,UPLOADDATE,UPLOADFILENAME,UPLOADUSER_ID )
    --                             SELECT uf.ID,
    --                                    @PeriodID,
    --                                    ISNULL((select distinct PROJECT_ID from T_UPLOADS where project_savix_ID = @ProjectID),@ProjectID),
    --                                    GETDATE(),
    --                                    'Automatic upload from web MIS',
    --                                    2
    --                             FROM T_UPLOADSFIRSTSTEP uf
    --                             INNER JOIN #projects mp ON  uf.Project_ID_MIS = mp.ProjectID
    --                             WHERE  uf.Project_ID_MIS = @ProjectID AND uf.[STATUS] = 0 
    --                                AND NOT EXISTS (SELECT * FROM T_UPLOADS u WHERE u.PROJECT_ID  = uf.PROJECT_ID AND u.PERIOD_ID = @PeriodID AND u.[STATUS] = 3)

    --                            DELETE FROM T_UPLOADSFIRSTSTEP WHERE STATUS = 0 AND PROJECT_ID = ISNULL((select distinct PROJECT_ID from T_UPLOADS where project_savix_ID = @ProjectID),@ProjectID)

    --                            --SET @UploadID = (SELECT ID FROM T_UPLOADS u WHERE  u.project_savix_ID = @ProjectID AND u.PERIOD_ID = @PeriodID AND u.[STATUS] = 3)
    --            END 
                --ELSE

Então aqui está a lógica real que é feita dentro desses 3 cursores. Entendendo que não nos importamos com períodos, a lógica percorre todos eles. Nós só nos importamos que tenhamos que filtrar certos critérios ( status =3 e UPLOADFILENAME ='Upload automático do web MIS' ) e que eles tenham uma tabela T_Project.ID correspondente com os critérios T_Projects.SyncMode =1
                            UPDATE t SET 
                                    TOTALEXPENDITURES = CASE WHEN DATEDIFF(d,mp.StartDate,mp.EndDate) != 0 
                                            THEN mp.Cost_Until_Today*DATEDIFF(d,mp.StartDate,dbo.fn_GetEndOfPeriod(t.Period_ID))/DATEDIFF(d,mp.StartDate,mp.EndDate)/dbo.fn_RateAtDate(mp.Currency,dbo.fn_GetEndOfPeriod(t.Period_ID))
                                            ELSE 0 
                                            END,
                                    TotalExpendituresNative = CASE WHEN DATEDIFF(d,mp.StartDate,mp.EndDate) != 0 
                                            THEN mp.Cost_Until_Today*DATEDIFF(d,mp.StartDate,dbo.fn_GetEndOfPeriod(t.Period_ID))/DATEDIFF(d,mp.StartDate,mp.EndDate)
                                            ELSE 0 
                                            END
                            FROM T_UPLOADS t
                            JOIN #projects mp ON mp.ProjectID = t.project_savix_ID
                            WHERE 1=1 -- t.ID = @UploadID 
                            --AND t.project_savix_ID = @ProjectID AND t.PERIOD_ID = @PeriodID 
                            AND t.[STATUS] = 3 AND t.UPLOADFILENAME = 'Automatic upload from web MIS'

A inserção em T_PROJECTGROUPDATA duplicou a lógica usada para criar a variável @dataid derivada de #datasets. Que é derivado da tabela de conjuntos de dados que descartamos e criamos acima.
                        INSERT INTO T_PROJECTGROUPSDATA (uploadID, fieldA,fieldB,......fieldN )

                         SELECT @UploadID,p.fieldA,mg.fieldB,......mgc.fieldN
                         FROM #projects mp
                         inner join  SG_Dynamic_Forms..v_projects p ON p.LegacyProjectId = mp.projectID 
                         inner join Savix_Service_Group..Groups mg ON mg.Grp_Project = p.ProjectID 
                         INNER JOIN Savix_Service_Group..Group_Cycle mgc ON mgc.Gc_GroupID = mg.Grp_ID
                         INNER JOIN Savix_Service_Group..Group_Data mgd ON mgd.Gd_CycleID = mgc.Gc_ID
                         LEFT JOIN Savix_Service_Trainers..Trainers me ON me.Tr_ID = mgc.Gc_MonitoredBy
                         LEFT JOIN Savix_Service_Dictionaries..Dictionary mgt ON mgt.Dny_ID = me.Tr_Type 
                         --left join v1_Report_UDF_Data_UploadToSavix udf on udf.DataID = mgd.Gd_ID
                         WHERE mgd.Gd_ID = @DataID 

                --FETCH NEXT FROM period INTO @PeriodID, @EndDate
                SET @MINIDperiods = @MINIDperiods + 1
            END
            --CLOSE period
            --DEALLOCATE period

            --FETCH NEXT FROM datasets INTO @DataID, @GroupID, @CollectionDate, @Status, @Createdate
            SET @MINIDdatasets = @MINIDdatasets + 1
        END

        --CLOSE datasets
        --DEALLOCATE datasets

        --COMMIT
    --END TRY
    --BEGIN CATCH



        --SELECT ERROR_NUMBER(), ERROR_MESSAGE(),@PeriodID, @ProjectID, @UploadID,@DataID
        --IF CURSOR_STATUS('global' , 'period') >= 0
        --BEGIN
        --    CLOSE period
        --    DEALLOCATE uploadID
        --END

        --IF CURSOR_STATUS('global' , 'datasets') >= 0
        --BEGIN
        --    CLOSE datasets
        --    DEALLOCATE datasets
        --END

        --IF @@TRANCOUNT > 0
        --    ROLLBACK

        --INSERT INTO error_catch_UploadtoSavix

        --SELECT cast(ERROR_NUMBER() as nvarchar), ERROR_MESSAGE(),@PeriodID, @ProjectID, @UploadID,@DataID, getdate()
    --END CATCH

    SET @MINIDprojects = @MINIDprojects + 1
    --FETCH NEXT FROM projects1 INTO @ProjectID, @FAID
END


--CLOSE projects1
--DEALLOCATE projects1

--SELECT 1 as success

Aí está. Condensando quase 300 linhas e 3 cursores até 30 linhas e sem cursores.