使用R

Question

我正在尝试将SQL代码转换为R代码。但是，数据大约有3500万条记录，每条记录有200列。所以我能找到的最佳选择是data.table包。

这是问题所在。 在SQL代码中，我能够执行这样的操作，

select order_date,sum(case when item in ("D","C","B") then col4 end)as col1
sum(case when item not in ("Z","X","Y") then col4 end) as col2 
from datatable 
where col3 <25
group by order_date;

以上查询允许我按每个日期分组。我无法在data.table中复制它。我的尝试如下。

grp1<- c("D","C","B")
grp2<- c("Z","X","Y")
d1 <- dat[item %in% grp1,.(col1 = sum(col4,na.rm = TRUE),by = Order_Date]
d2 <- dat[item %in% grp2,.(col2 = sum(col4,na.rm = TRUE),by = Order_Date]
d3 <- data.table(d1,d2)

现在，因为它subsets，我的分组在d1和d2

中的分组不同

Answer 1

您可以尝试以下操作：

DT[col3 < 25,
   .(col1 = sum(col4[item %in% c("D","C","B")]),
     col2 = sum(col4[!item %in% c("Z","X","Y")])),
   by = .(order_date)]

Answer 2

>     d <- "
+     order_date,item,col4,col3
+     2000-01-01,D,1,10
+     2000-01-01,C,1,10
+     2000-01-01,M,1,10
+     2000-01-01,N,1,50
+     2000-01-01,Z,1,10
+     2000-01-01,X,1,10
+     2001-01-02,Z,1,0
+     2001-01-02,X,1,50"
>
>     df = read.csv(textConnection(d))
>
>     # data.frame + plyr approach
>
>     require(plyr)
Loading required package: plyr
>     ddply(
+       df[df$col3<25,],
+       .(order_date),
+       summarize,
+       col1=sum(item %in% c("D","C","B") & col4),
+       col2=sum(!item %in% c("Z","X","Y") & col4)
+     )
      order_date col1 col2
1     2000-01-01    2    3
2     2001-01-02    0    0
>
>     # data.table approach, thanks to jangorecki
>
>     require(data.table)
Loading required package: data.table
>     dt = data.table(df)
>
>     dt[col3 < 25,
+        .(col1 = sum(col4[item %in% c("D","C","B")]),
+          col2 = sum(col4[!item %in% c("Z","X","Y")])),
+        by = .(order_date)]
       order_date col1 col2
1:     2000-01-01    2    3
2:     2001-01-02    0    0
>