Idiom for conditionally selecting columns from a data.table

Question

I use the following idiom for conditionally selecting columns from a data.frame:

DF = data.frame(a = 1:3,b = letters[1:3],c = LETTERS[1:3])
someCondition <- FALSE

# use `if(someCondition)` to conditionally include column 'c'
DF[,c('a','b',if(someCondition)'c')] 
:>   a b
:> 1 1 a
:> 2 2 b
:> 3 3 c

but the equivalent does not work with data.table's b/c NULL values are not dropped from lists the same way they are dropped from concatenation:

DT = as.data.table(DF)
DT[,.(a,b,if(someCondition)c)]
:> Error in setnames(jval, jvnames) : 
:>   Can't assign 3 names to a 2 column data.table

I've defined a function called .. which is a work around:

.. <- function(...){
    x = list(...)
    x= x[!sapply(x,is.null)]
    x
}
DT[,..(a,b,if(someCondition)c)]
:>    V1 V2
:> 1:  1  a
:> 2:  2  b
:> 3:  3  c

but it seeks kind of kludgy to have to include my own function to accomplish an operation that is so common. Is there a more idiomatic way of conditionally selecting columns from a data.table?

Answer 1

I think the .SDcols argument Does what you want. In the above example for data.table DF,

DF[, .SD, .SDcols= c("a","b", if(someCondition) "c")]

Will act in the same manner as in your data.frame. You can also implement this as in the below example.

DF[, .SD, .SDcols=if(someCondition) c("a","b","c") else c("a","b")]

will perform the desired selection. In a previous line, you could set up more elaborate constructions of the true and false vectors (which might defeat the purpose of keeping thing succinct).