Question

I am having some trouble getting an eval within data.table in R to work with an expression. Here is some code:

dtb = data.table(a=1:100, b=100:1, id=1:10)
dtb[,`:=`(c=a+b, d=a/b),by=id] #this works fine
expr = expression({`:=`(c=a+b, d=a/b)}) #try to couch everything in an expression
dtb[,eval(expr),by=id] #this does not work
Error in `:=`(c = a + b, d = a/b) : 
   unused argument(s) (c = a + b, d = a/b)

expr = expression(`:=`(c=a+b, d=a/b)) #this works fine
dtb[,eval(expr),by=id] 

Why does including {} break this?

Answer 1

See the definition of :=:

function (LHS, RHS) 
stop(":= is defined for use in j only, and (currently) only once; i.e., DT[i,col:=1L] and DT[,newcol:=sum(colB),by=colA] are ok, but not DT[i,col]:=1L, not DT[i]$col:=1L and not DT[,{newcol1:=1L;newcol2:=2L}]. Please see help(\":=\"). Check is.data.table(DT) is TRUE.")

The assignment of a column doesn’t happen within a call of :=–the function itself doesn’t do anything besides produce an error. The assignment happens when [.data.table detectsj is an expression of the form `:=`(...) and then sets everything up for a call to the C code. When you enclose expr in brackets, you’re making the first part of the expression { instead of :=, which passes by the above detection and eventually results in an evaluation of := with arguments c and d.

I guess that leads to the question, why do you need to enclose it in { }?