Multiple levels of recoding

[yulric]

A new non-derived variable in the library can be created from a set of start variables that are in a database. Similarly, a new derived variable
can be created from a combination of start variable(s) and other derived variables.

However, the library does not allow the creation of new derived/non-derived variable that uses a new derived/non-derived variable. For example the following test case,

test_that("Test", {
  data <- data.frame(
    start_a = c(1)
  )
  variables <- data.frame(
    variable = c("a", "b", "c", "d"),
    label = c("", "", "", ""),
    labelLong = c("", "", "", ""),
    units = c("N/A", "N/A", "N/A", "N/A"),
    variableType = c("Continuous", "Continuous", "Continuous", "Continuous"),
    databaseStart = c("database_one", "database_one", "database_one", "database_one"),
    variableStart = c("[start_a]", "DerivedVar::[a]", "DerivedVar::[b]", "[c]")
  )
  database_name <- "database_one"
  variable_details <- data.frame(
    variable = c("a", "b", "c", "d"),
    typeEnd = c("cont", "cont", "cont", "cont"),
    databaseStart = c("database_one", "database_one", "database_one", "database_one"),
    variableStart = c("[start_a]", "DerivedVar::[a]", "DerivedVar::[b]", "[c]"),
    typeStart = c("cont", "cont", "cont", "cont"),
    recEnd = c("copy", "Func::b", "Func::c", "copy"),
    numValidCategories = c("N/A", "N/A", "N/A", "N/A"),
    recStart = c("else", "else", "else", "else"),
    catLabel = c("", "", "", ""),
    catLabelLong = c("", "", "", "")
  )
  .GlobalEnv$b <- function(a) {
    return(a + 1)
  }
  .GlobalEnv$c <- function(b) {
    return(b+1)
  }
  recoded_data <- rec_with_table(
    data, variables, database_name, variable_details)
  print(recoded_data)
})

prints

	a	b	c
1	1	2	2

when it should print,

	a	b	c	d
1	1	2	3	3

In summary, the library should support use cases beyond only one level deep of recoding.

[zargot]

I think the value of c is correct as-is, since it would otherwise be ambiguous if it (being a derived var) was referring to the start-value or the end-value of b. Since b = DerivedVar::[a] = [start_a], it makes sense that DerivedVar::[b] = DerivedVar::[a] = [start_a] as well, otherwise we'd have to add a way to specify if it should be derived from the start or end value.

I haven't wrapped my head around d yet.

[yulric]

Can you expand on what you mean by start-value and end-value of b? b has only one value in my mind, the one that is derived from a

[zargot]

As far as I can see, the variables are assigned a start-value in variableStart, and a transformation function in recEnd which would output the end-value, unless I have misunderstood something.

[yulric]

That's one of way of looking at variableStart but it actually means the variables that are used to create the new variable. The variableStart definition you mentioned would breakdown if there were multiple start variables in there since a variable can't have more than value at a time.

[zargot]

does that mean that DerivedVar::[a] always is referring to the recoded/end/final value of a?

[zargot]

alright, then I see the issue with c. This seems like two different issues then. I'll look into this first, and then figure out d later, unless it would be the same fix.