R: Combine two functions to aggregate data using data.table

Question

The following is the kind of data "types/structure" i'm working with, it includes 3 factor variables.

library(data.table)
library(ggplot2)

DT <- data.table(mtcars)
DT[["cyl"]] <- factor(DT[["cyl"]])
DT[["gear"]] <- factor(DT[["gear"]])
DT[["vs"]] <- factor(DT[["vs"]])
DT <- DT[, c("cyl", "gear", "vs"), with=F]
setkey(DT, cyl, gear, vs)

Context I've been using this function to aggregate data with data.table interactively and works just fine. The problem is when i try to include it in another function. I don't have much experience programming, so any guidence will be greatly appreciated. I imagine this has to do with enviroments, and how the arguments are passed but i don't really know how to solve it.

grp <- function(x) {
  percentage = as.numeric(table(x)/length(x))
  list(x = levels(x),
       percentage = percentage,
       label = paste( round( as.numeric(table(x)/length(x), 0 ) * 100 ), "%")
  )
}

This would be the expected output:

DT_agg <- DT[, grp(cyl), by=vs]

Question The idea of the second function is to take a data.frame/data.table object, apply the previous function including the option to use one or two grouping variables. The final idea, would be to include this last object in a ggplot() call, and use the grouping variables as facets, calling the agg() function from the ggplot() call.

agg <- function(data, x, groupby1, groupby2 = NULL,...){
  data = substitute(data)
  x = substitute(x)
  groupby1 = substitute(groupby1)
  groupby2 = substitute(groupby2)

  if(is.null(groupby2)){
    DT_agg = data[, grp(x), by=groupby1]
  } else {
    DT_agg = data[, grp(x), by=groupby1,groupby2]
  }
  DT_agg
}


agg(data = DT, x = cyl, groupby1 = vs)
Error in unique.default(x, nmax = nmax) : 
  unique() applies only to vectors

EDIT After agstudy's answer

agg <- function(data, x, groupby1, groupby2 = NULL,...){
  data = eval(substitute(data))
  x = substitute(data$x)  # changed this bit (it was producing an error)
  groupby1 = substitute(groupby1)
  groupby2 = substitute(groupby2)
  if(is.null(eval(substitute(groupby2)))) {
    eval(data)[, grp(eval(x)), by=groupby1]
  } else {
    eval(data)[, grp(eval(x)), by=list(eval((groupby1)),eval(groupby2))]
  }
}

For some reason the provided solution in the answer isn't working for me. Agstudy's agg() provides an answer, it runs but the output isn't. I've tried a few changes, but it's not working right.

Using the grp() function defined above i get this result that's correct:

ok = DT[, grp(cyl), by = vs]
print(ok)
#       vs x percentage label
#    1:  1 4 0.71428571   71%
#    2:  1 6 0.28571429   29%
#    3:  1 8 0.00000000    0%
#    4:  0 4 0.05555556    6%
#    5:  0 6 0.16666667   17%
#    6:  0 8 0.77777778   78%

Using agstudy version of agg() i get this that's not correct:

not_ok = agg(DT, cyl, vs)
print(not_ok)
#       groupby1 x percentage label
#    1:        1 4    0.34375   34%
#    2:        1 6    0.21875   22%
#    3:        1 8    0.43750   44%
#    4:        0 4    0.34375   34%
#    5:        0 6    0.21875   22%
#    6:        0 8    0.43750   44%

I wonder how can the function work correctly on it's own (1st case) and not inside the agg function.

Answer 1

Excellent question ! specially for somemone who don't have much experience programming.

Using eval and simplifying your function ( no need to assign the data.table):

agg <- function(data, x, groupby1, groupby2 = NULL,...){
  data = substitute(data)
  x = substitute(x)
  groupby1 = substitute(groupby1)
  groupby2 = substitute(groupby2)
  if(is.null(groupby2)) eval(data)[, grp(eval(x)), by=groupby1]
  else  eval(data)[, grp(eval(x)), 
              by=list(eval((groupby1)),eval(groupby2))]

}

testing it :

agg(data = DT, x = cyl, groupby1 = vs,groupby2 =  gear )
##     groupby1 groupby2 x percentage label
##  1:        1        3 4     0.3333  33 %
##  2:        1        3 6     0.6667  67 %
##  3:        1        3 8     0.0000   0 %
##  4:        1        4 4     0.8000  80 %
##  5:        1        4 6     0.2000  20 %
##  6:        1        4 8     0.0000   0 %
##  7:        0        5 4     0.2500  25 %
##  8:        0        5 6     0.2500  25 %
##  9:        0        5 8     0.5000  50 %
## 10:        1        5 4     1.0000 100 %
## 11:        1        5 6     0.0000   0 %
## 12:        1        5 8     0.0000   0 %
## 13:        0        4 4     0.0000   0 %
## 14:        0        4 6     1.0000 100 %
## 15:        0        4 8     0.0000   0 %
## 16:        0        3 4     0.0000   0 %
## 17:        0        3 6     0.0000   0 %
## 18:        0        3 8     1.0000 100 %

Answer 2

I believe this is working with the 1.8.11 version. Perhaps something changed in the dev version.

library(data.table)
library(ggplot2)

DT <- data.table(mtcars)
DT[["cyl"]] <- factor(DT[["cyl"]])
DT[["gear"]] <- factor(DT[["gear"]])
DT[["vs"]] <- factor(DT[["vs"]])
DT <- DT[, c("cyl", "gear", "vs"), with=F]
setkey(DT, cyl, gear, vs)

grp <- function(x) {
  percentage = as.numeric(table(x)/length(x))
  list(x = levels(x),
       percentage = percentage,
       label = paste( round( as.numeric(table(x)/length(x), 0 ) * 100 ), "%")
  )
}

agg <- function(data, x, groupby1, groupby2 = NULL,...){
  data = substitute(data)
  x = substitute(x)
  groupby1 = substitute(groupby1)
  groupby2 = substitute(groupby2)
  if(is.null(groupby2)) eval(data)[, grp(eval(x)), by=groupby1]
  else  eval(data)[, grp(eval(x)), 
                   by=list(eval((groupby1)),eval(groupby2))]

}

ok = DT[, grp(cyl), by = vs]
print(ok)


ok2 = agg(DT, cyl, vs)
print(ok2)