Pass character string formula with comma to R data table

user_11770286 0
Viewed 96

I'm trying to assign new variables to a data.table based on formulas which I have in a pre-existing data.frame, by looping over the data.frame, like this:

# Setup
DT <- data.table(id = c("a", "b"), value1 = c(1, 3), value2 = c(2,4))
var_exp = data.frame(varname = c("ratio", "sum"), 
                     expr = c("(value1 / value2) * 100", 
                              'rowSums(.SD, na.rm=TRUE), .SDcols=c("value1", "value2")'),
                     stringsAsFactors=FALSE)

# Assign new variables
for (row in 1:nrow(var_exp)) {
  varname <- var_exp[row, "varname"]
  expr <- var_exp[row, "expr"]
  
  DT[, (varname) := eval(parse(text = expr))]
}

The first expression is understood without problems. However, the second expression returns an error because there is a comma in the expression:

Error in parse(text = expr) : <text>:1:29: unexpected ','
1: lapply(.SD, sum, na.rm=TRUE),

How can I pass this expression to data.table?

data.table dataframe r string
2 Answers

Usually it's not a good practice to evaluate string as codebut here is a way to do it.

library(data.table)

for (row in 1:nrow(var_exp)) {
  varname <- var_exp[row, "varname"]
  expr <- var_exp[row, "expr"]
  eval(parse(text = sprintf('DT[, %s := %s]', varname, expr)))
}

DT

#   id value1 value2 ratio sum
#1:  a      1      2    50   3
#2:  b      3      4    75   7
user_39629140

We can do this with lapply

DT[, (var_exp$varname) := setDT(var_exp)[, lapply(sprintf("DT[,%s]",
       expr), function(x) eval(parse(text = x)))]]
DT
   id value1 value2 ratio sum
1:  a      1      2    50   3
2:  b      3      4    75   7
user_37322710
Related