数据表和列名

Question

I am using data.table to get some basic statistics in one column while filtering by another column(s).

This is the command

stats <- as.data.frame(mydata[, j = list(Sum = sum(as.numeric(get(selection))),
                                         Average = mean(as.numeric(get(selection))),
                                         Count = length(get(selection))), 
                                by = list(get(filters))])

where:

• mydata is a data.table with 20 or so columns
• selection is a column name which is passed programmatically
• filters is also a column name which is passed programmatically

If I limit myself to one filter (one column) everything works, but I would like to filter by more than one column.

It is possible to do:

by = list(get(filters[1]), get(filters[2]), ...) 

However, that requires I know how many filters will be used. That is a limitation I don't want to have.

How do I write the by = to take any number of filters (column names) - I just tried mget(filters) and that doesn't work.

Answer 1

Data.tables by argument accepts a character vector of column names (see the documentation: help("data.table") ). There is no need for get . Just use by = c(filters) .

Example:

library(data.table)
DT <- data.table(mtcars)

filters <- c("am", "gear")
DT[, mean(mpg), by=c(filters)]
#   am gear       V1
#1:  1    4 26.27500
#2:  0    3 16.10667
#3:  0    4 21.05000
#4:  1    5 21.38000