问题
I have two data tables, DT1 and DT2:
set.seed(1)
DT1<-data.table(id1=rep(1:3,2),id2=sample(letters,6), v1=rnorm(6), key="id2")
DT1
## id1 id2 v1
## 1: 2 e 0.7383247
## 2: 1 g 1.5952808
## 3: 2 j 0.3295078
## 4: 3 n -0.8204684
## 5: 3 s 0.5757814
## 6: 1 u 0.4874291
DT2<-data.table(id2=c("n","u"), v1=0, key="id2")
DT2
## id2 v1
## 1: n 0
## 2: u 0
I would like to update DT1 based on a join with DT2, but only for a subset of DT1. For example, for DT1[id1==3], I would expect the value of v1 in row 4 to be updated as in the following result:
DT1
## id1 id2 v1
## 1: 2 e 0.7383247
## 2: 1 g 1.5952808
## 3: 2 j 0.3295078
## 4: 3 n 0
## 5: 3 s 0.5757814
## 6: 1 u 0.4874291
I know how to update a table (using the := assignment operator), how to join the tables (DT1[DT2]), and how to subset a table (DT1[id1==3]). However I'm not sure how to do all three at once.
EDIT: Note that the original example only attempts to update one column, but my actual data requires updating many columns. Consider the additional scenarios in DT1b and DT2b:
set.seed(2)
DT1b<-DT1[,v2:=rnorm(6)] # Copy DT1 and add a new column
setkey(DT1b,id2)
DT1b
## id1 id2 v1 v2
## 1: 2 e 0.7383247 -0.89691455
## 2: 1 g 1.5952808 0.18484918
## 3: 2 j 0.3295078 1.58784533
## 4: 3 n -0.8204684 -1.13037567
## 5: 3 s 0.5757814 -0.08025176
## 6: 1 u 0.4874291 0.13242028
DT2b<-rbindlist(list(DT2,data.table(id2="e",v1=0))) # Copy DT2 and add a new row
DT2b[,v2:=-1] # Add a new column to DT2b
setkey(DT2b,id2)
DT2b
## id2 v1 v2
## 1: e 0 -1
## 2: n 0 -1
## 3: u 0 -1
Based on the helpful answers from @nmel and @BlueMagister, I came up with this solution for the updated scenario:
DT1b[DT2b[DT1b[id1 %in% c(1,2)],nomatch=0],c("v1","v2"):=list(i.v1,i.v2)]
DT1b
## id1 id2 v1 v2
## 1: 2 e 0.0000000 -1.00000000
## 2: 1 g 1.5952808 0.18484918
## 3: 2 j 0.3295078 1.58784533
## 4: 3 n -0.8204684 -1.13037567
## 5: 3 s 0.5757814 -0.08025176
## 6: 1 u 0.0000000 -1.00000000
回答1:
The easiest way I can think of is to key by id1 as well.
eg
setkey(DT1, id2,id1)
DT2[, id1 := 3]
setkey(DT2, id2, id1)
# use i.v1 to reference v1 from the i component
DT1[DT2, v1 := i.v1 ]
DT1
id1 id2 v1
1: 2 e 0.7383247
2: 1 g 1.5952808
3: 2 j 0.3295078
4: 3 n 0.0000000
5: 3 s 0.5757814
6: 1 u 0.4874291
回答2:
This is similar to mnel's solution but uses ifelse instead of a second key.
DT1[DT2, v1 := ifelse(id1==3, i.v1, v1),nomatch=0]
来源：https://stackoverflow.com/questions/14720923/update-subset-of-data-table-based-on-join