在R中使用合并定义data.table的子集

Question

I'm using multiple merges to define an ID variable in R (see this question for more context).

I want to merge variable v from data.table x to data.table y first according to, say, keys k1 in y .

Then for those observations that weren't matched in the first stage, I want to merge them according to table y keys k2 .

y[is.na(v),x,v:=v]

doesn't work, as data.table syntax expects a data.table first when merging.

y[is.na(v),][x,v:=v]

works in a sense, but doesn't save the results of the merge to y .

Here's a minimal example:

x<-data.table(v1=c("A","B","C"),v2=c("a","b","c"),v=rnorm(3),key=c("v1","v2"))
y<-data.table(v1=c("A","B","C"),v21=c("","b","c"),v22=c("a","",""))
setkey(y,v1,v21)
y[x,v:=v]

gives

> x
   v1 v2          v
1:  A  a  0.3316665
2:  B  b  0.8470424
3:  C  c -0.5955292
> y
   v1 v21 v22          v
1:  A       a         NA
2:  B   b      0.8470424
3:  C   c     -0.5955292

And of course what I want is:

> y
   v1 v21 v22          v
1:  A       a  0.3316665
2:  B   b      0.8470424
3:  C   c     -0.5955292

Answer 1

Try this:

setkey(y, v1, v22)
y[x, v := ifelse(is.na(v), i.v, v)]

i. can be used to distinguish same column name from the i-expression data.table .

Answer 2

An alternative to @eddi's solution that I'm working with that is somewhat more robust:

setkey(y, v1, v22)
y[x[!(v %in% y$v),],v:=i.v]

(basically, instead of subsetting y , subset x via y and join the subsetted x to y .