0

目前我正在使用 R 来转换我的表格数据：

ID    Code    Condition    WT
104   KEENTRAN  CON4      .30577
.     .         .          .
.     .         .          .

该链接应该适用于任何想要下载我的数据框的人，否则这里是一个子集：

>dput(head(df))

structure(list(ID = c(104L, 368L, 10632L, 20385L, 24361L, 34378L
), Code = c("KEENTRAN", "ALEXEXPR", "MINNEXPMN", "JACKMOVWI", 
            "FREICOIN", "JBXEXPGA"), Condition = c("CON4", "CON4", "CON2", 
                                                   "CON2", "CON6", "CON5"), WT = c(0.3057717456, 0.7909870604, 1, 
                                                                                   1, 0.4301040524, 0.5977268575)), .Names = c("ID", "Code", "Condition", 
                                                                                                                               "WT"), class = c("tbl_df", "data.frame"), row.names = c(NA, -6L
                                                                                                                               ))

背景

我的示例数据是长格式，其中 Condition 变量的范围从“CON1”到“CON6”，我想将我的数据重新转换为宽格式，其中ID和Code值将是主键，而Condition的级别将是其列值取该特定 ID、代码、条件分组的WT最大值（如果不存在此类配对，则为零）。dcast()这可以使用包中的函数在 R 中轻松完成reshape2：

library(reshape2)

Result <- df %>% group_by(ID, Condition) %>% 
  summarise(value = max(as.numeric(WT))) %>% 
  dcast(ID ~ Condition)

Result[is.na(Result)] <- 0

我想在 SQL Server 中复制这个数据操作过程，但我不确定如何最好地做到这一点。非常感谢任何帮助或见解。

4

2 回答 2

1

考虑 SQL Server 的PIVOT操作：

SELECT t.ID, 
       ISNULL(t.[CON1], 0) AS [CON1],
       ISNULL(t.[CON2], 0) AS [CON2],
       ISNULL(t.[CON3], 0) AS [CON3],
       ISNULL(t.[CON4], 0) AS [CON4],
       ISNULL(t.[CON5], 0) AS [CON5],
       ISNULL(t.[CON6], 0) AS [CON6]    
FROM RDataFrame As r                  
PIVOT 
  (
    MAX(r.[WT]) 
    FOR r.Condition IN ([CON1], [CON2], [CON3], [CON4], [CON5], [CON6])
  ) AS t

-- ID   CON1    CON2    CON3    CON4            CON5    CON6
-- 8    0       0       0       0.4394051665    0       0
-- 10   0       0       0       0.6013843825    0       0
-- 15   0       0       0       0.07231002554   0       0
-- 21   0       0       0       0.6013843825    0       0
-- 23   0       0       0       0.7720454793    0       0
-- 80   0       1       0       0               0       0
-- 104  0       0       0       0.3057717456    0       0
-- 144  0       0       0       0.1430937996    0       0.2646439667
-- 145  0       0       0       0.8276574       0       0
-- 155  0       1       0       0.8977280575    0       0
-- 156  0       0       0       0.8453629338    0       0
-- 158  0       0       0       0.5221399019    0       0
于 2016-10-29T04:50:34.963 回答
0

我能够通过在 SQL Server 2016 中执行以下操作来回答我的问题：

首先，我将发布的数据（这是存储过程的结果）转储到临时表中：

DROP TABLE IF EXISTS #InputDataFrame
CREATE TABLE #InputDataFrame ( ID int, Code varchar(25), Condition varchar(25), WT float)
INSERT INTO #InputDataFrame
exec dbo.[Stored_Proc_to_Create_Sample_Data]   -- stored proc to create my posted data

然后我能够重新创建所需的转换，如下所示：

DROP TABLE IF EXISTS #DistinctIDs
CREATE TABLE #DistinctIDs ( ID int, Code varchar(25) )
INSERT INTO #DistinctIDs ( ID, Code)
Select Distinct
I.ID,
I.Code
From #InputDataFrame I

Select
D.*,
CASE WHEN CON1.WT IS NULL THEN 0 ELSE CON1.WT END as CON1,
CASE WHEN CON2.WT IS NULL THEN 0 ELSE CON2.WT END as CON2,
CASE WHEN CON3.WT IS NULL THEN 0 ELSE CON3.WT END as CON3,
CASE WHEN CON4.WT IS NULL THEN 0 ELSE CON4.WT END as CON4,
CASE WHEN CON5.WT IS NULL THEN 0 ELSE CON5.WT END as CON5,
CASE WHEN CON6.WT IS NULL THEN 0 ELSE CON6.WT END as CON6

From #DistinctIDs D

Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON1' Group By I.ID) CON1 on CON1.ID = D.ID
Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON2' Group By I.ID) CON2 on CON2.ID = D.ID
Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON3' Group By I.ID) CON3 on CON3.ID = D.ID
Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON4' Group By I.ID) CON4 on CON4.ID = D.ID
Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON5' Group By I.ID) CON5 on CON5.ID = D.ID
Left Join ( Select I.ID, MAX( I.WT) as WT From #InputDataFrame I Where I.Condition = 'CON6' Group By I.ID) CON6 on CON6.ID = D.ID

我怀疑这是最优雅的解决方案，但它适用于我的目的，希望它对任何可能在环境中努力模仿 R dcast()功能的人有用SQL。

于 2016-10-28T21:36:32.857 回答