分析函式改寫自關聯

資料庫環境：SQL SERVER 2005

我們會看到這樣的需求，查詢部門的員工資訊和對應的部門人數，要求在同一行上展示。

在沒有分析函式之前，遇到需求，只能通過員工表自關聯來實現。

我這裡就有一個類似的例子

SELECT COUNT(*)
 FROM   dmrpout_03
        LEFT JOIN ctlm1210 ctlm1210_id_cgroup ON dmrpout_03. id_cgroup = ctlm1210_id_cgroup.id_cgroup
        LEFT JOIN ctlm7161 ctlm7161_id_schtype ON dmrpout_03. id_schtype = ctlm7161_id_schtype .id_schtype
        LEFT JOIN ctlm5003 ctlm5003_id_proc ON dmrpout_03. id_proc = ctlm5003_id_proc.id_proc
        LEFT JOIN ctlm1003 ctlm1003_id_dept ON dmrpout_03. id_dept = ctlm1003_id_dept.id_dept
        LEFT JOIN ctlm5111 ctlm5111_id_wkct ON dmrpout_03. id_wkct = ctlm5111_id_wkct.id_wkct
        LEFT JOIN ctlm1007 ctlm1007_id_clerk ON dmrpout_03. id_clerk = ctlm1007_id_clerk.id_clerk
        LEFT JOIN ctlm3006 ctlm3006_id_stockreas ON dmrpout_03.id_stockreas = ctlm3006_id_stockreas.id_stockreas
        LEFT JOIN ctlm3004 ctlm3004_id_keeper ON dmrpout_03. id_keeper = ctlm3004_id_keeper.id_keeper
        LEFT JOIN ctlm3001 ctlm3001_id_wh ON dmrpout_03. id_wh = ctlm3001_id_wh.id_wh
        LEFT JOIN ctlm5112 ctlm5112_id_pline ON dmrpout_03. id_pline = ctlm5112_id_pline.id_pline
        LEFT JOIN ctlm3001 ctlm3001_id_prodwh ON dmrpout_03. id_prodwh = ctlm3001_id_prodwh.id_wh
        LEFT JOIN ctlm1005 ctlm1005_id_recorder ON dmrpout_03.id_recorder = ctlm1005_id_recorder.id_user
        LEFT JOIN ctlm3002 ctlm3002_id_loc ON dmrpout_03. id_loc = ctlm3002_id_loc.id_loc
        LEFT JOIN ctlm3002 ctlm3002_id_prodloc ON dmrpout_03. id_prodloc = ctlm3002_id_prodloc .id_loc
        LEFT JOIN v_ctlm2004 v_ctlm2004_id_uom ON dmrpout_03. id_uom = v_ctlm2004_id_uom.id_uom
        LEFT JOIN v_ctlm2004 v_ctlm2004_id_secuom ON dmrpout_03.id_secuom = v_ctlm2004_id_secuom.id_uom
 WHERE  dmrpout_03.line_no = ( SELECT    MIN( line_no)
                               FROM     dmrpout_03 b
                               WHERE    b .dmrpout_no = dmrpout_03.dmrpout_no
                                        AND b .fiscal_year = '2015'
                                        AND b .fiscal_period = '6'
                                        AND b .id_dept IN (
                                        SELECT  id_dept
                                        FROM    ctlm1003 ctlm1003_id_dept
                                        WHERE   ctlm1003_id_dept.name_dept LIKE '%生產%' )
                                        AND b .id_wkct IN (
                                        SELECT  id_wkct
                                        FROM    ctlm5111 ctlm5111_id_wkct
                                        WHERE   ctlm5111_id_wkct.name_wkct LIKE '%儲運%' )
                             )
        AND dmrpout_03 .fiscal_year = '2015'
        AND dmrpout_03 .fiscal_period = '6'
        AND ctlm1003_id_dept .name_dept LIKE '% 生產%'
        AND ctlm5111_id_wkct .name_wkct LIKE '% 儲運%'

這條SQL要執行4分多鐘才出結果。我們可以從2方面進行優化，第一，SQL裡有很多LEFT JOIN，

都是dmrpout_03 LEFT JOIN其它表，如果是一對一關係，則LEFT JOIN後面的表可以省略；

第二，將自連線改寫成分析函式。

改寫後的SQL如下，只耗時12S就出結果了。

SELECT COUNT (*)
 FROM   ( SELECT    dmrpout_03.line_no ,
                    MIN(line_no ) OVER ( PARTITION BY dmrpout_03.dmrpout_no ) AS minline_no
          FROM      dmrpout_03
                    LEFT JOIN ctlm1003 ctlm1003_id_dept ON dmrpout_03.id_dept = ctlm1003_id_dept.id_dept
                    LEFT JOIN ctlm5111 ctlm5111_id_wkct ON dmrpout_03.id_wkct = ctlm5111_id_wkct.id_wkct
                    LEFT JOIN ctlm3002 ctlm3002_id_loc ON dmrpout_03.id_loc = ctlm3002_id_loc.id_loc
                    LEFT JOIN ctlm3002 ctlm3002_id_prodloc ON dmrpout_03.id_prodloc = ctlm3002_id_prodloc.id_loc
                    LEFT JOIN v_ctlm2004 v_ctlm2004_id_uom ON dmrpout_03.id_uom = v_ctlm2004_id_uom.id_uom
                    LEFT JOIN v_ctlm2004 v_ctlm2004_id_secuom ON dmrpout_03.id_secuom = v_ctlm2004_id_secuom.id_uom
          WHERE     dmrpout_03 .fiscal_year = '2015'
                    AND dmrpout_03 .fiscal_period = '6'
                    AND dmrpout_03 .id_dept IN (
                    SELECT  id_dept
                    FROM    ctlm1003 ctlm1003_id_dept
                    WHERE   ctlm1003_id_dept .name_dept LIKE '%生產%' )
                    AND dmrpout_03 .id_wkct IN (
                    SELECT  id_wkct
                    FROM    ctlm5111 ctlm5111_id_wkct
                    WHERE   ctlm5111_id_wkct .name_wkct LIKE '%儲運%' )
        ) t
 WHERE  t.line_no = minline_no

通過分析函式的改寫，可以減少表的掃描。以後如果遇到類似需求，不妨改成這樣試試。

（本文完）