Assigning string as a variable-CodePudding

Is it possible to assign a string containing variable names to a variable in R?

For example imagine I have the string "log(p_n) log((1 - p_n)) log(p_n * p_b) log(p_n * (1 - p_b)) log((1 - p_n) * p_g)". I'd like to make this function:

formula_function = function(p_n, p_b, p_g){
  x <- log(p_n)   log((1 - p_n))   log(p_n * p_b)   log(p_n * (1 - p_b))   log((1 - p_n) * p_g)
  return(x)
}

I have a function which generates the formula strings and I'm just copy-pasting them by hand at the moment, is there a more 'program-y' way of doing it?

CodePudding user response：

1) Assuming you know the parameters:

x <- "log(p_n)   log((1 - p_n))   log(p_n * p_b)   log(p_n * (1 - p_b))   log((1 - p_n) * p_g)"

f <- function(p_n, p_b, p_g) {}
body(f) <- parse(text = x)
f
## function (p_n, p_b, p_g) 
## log(p_n)   log((1 - p_n))   log(p_n * p_b)   log(p_n * (1 - p_b))  
##    log((1 - p_n) * p_g)

2) The gsubfn package has a function for converting formulas to functions. The order of the formal arguments will be in the order encountered in x or you can specify the order. Convert the string to a formula using reformulate and then use as.function.

library(gsubfn)

# args are in order encountered
f <- as.function(reformulate(x))
f
## function (p_n, p_b, p_g) 
## log(p_n)   log((1 - p_n))   log(p_n * p_b)   log(p_n * (1 - p_b))   
##    log((1 - p_n) * p_g)

# specify order of args
f <- as.function(reformulate(x, "p_n   p_g   p_b"))
f
## function (p_n, p_g, p_b) 
## log(p_n)   log((1 - p_n))   log(p_n * p_b)   log(p_n * (1 - p_b))   
##     log((1 - p_n) * p_g)

3) gsubfn also has match.funfn that is like match.fun in R except it also translates arguments passed as formulas to functions. e.g.

myfun <- function(a, b, c, fun) {
  f <- match.funfn(fun)
  f(a, b, c)
}
myfun(0.1, 0.1, 0.1, reformulate(x, "p_n   p_g   p_b"))
## [1] -11.82901