CODEWITHSUNDEEP
r
dim fish
dim fish

Reputation: 479

R: Compute a new dataframe column from a function that calls functions

I'm trying to compute a new column for a dataframe using a multi-argument function. When that function is simple, everything is fine. When that functions calls another function, things go awry.

PART 1: The dataset of interest.

# The goal is to transform Foo into a new column.
df <- data.frame(
   Year =c(1901,1901,1900,1902),
   Month=c(   2,   1,   2,   1),
   Foo  =c(   1,   2,   3,   4)
)

# This simple multi-arg transform works fine.
Foo2Baz <- function(year, month, foo) {
  return(year + month + foo)
}
df$Baz <- Foo2Baz(df$Year, df$Month, df$Foo)
# Expected Baz column: [1904, 1904, 1905, 1907] 
df$Baz
[1] 1904 1904 1905 1907

PART 2: A lookup table

The more interesting transform that fails is like Foo2Baz above, except it will call another function that looks up a value from some other table.

# First we load a dataframe that backs the lookup operation. It's nice
# to keep as a dataframe, as opposed to matrix, for human inspection
# of cells.
lookup_table <- data.frame(
  Year=c(1900, 1901, 1902),
  Jan =c(  10,   20,   30),
  Feb =c( 100,  200,  200))

# Then we define a function to lookup a cell.
Lookup <- function(year, month) {
  return(lookup_table[
    year - 1899, # converts year to row index
    1 + month    # converts month to column, skipping 'Year'
  ])
}
# We expect a lookup of 1901-Feb to be 200
Lookup(1901, 2)
[1] 200

PART 3: The transform function that calls a function.

# The goal is to transform Foo into new column Bar
# by looking up that case's Year and Month.
Foo2Bar <- function(year, month, foo) {
  return(foo + Lookup(year, month))
}

# We expect case 1 (1901,Feb,Foo=1) to have Bar=201
Foo2Bar(1901,2,1)
[1] 201

# We expect case 4 (1902,Jan,Foo=4) to have Bar=34
Foo2Bar(1902,1,4)
[1] 34

PART 4: How to compute a new column for Bar?

Seems like we can now compute a Bar column using Foo2Bar just like we used the simpler Foo2Baz:

df$Bar <- Foo2Bar(df$Year, df$Month, df$Foo)
df$Bar
    Feb Jan Feb.1 Jan.1
2   201  21   201    21
2.1 202  22   202    22
1   103  13   103    13
3   204  34   204    34

No column? Instead a matrix where rows and columns look like various inputs we sent to the Lookup() function?!

I've tried variations with do.call, apply,lapply,sapply and dplyr mutate. I seem to be simply missing something fundamental here.

Upvotes: 1

Views: 88

Answers (1)

akrun
akrun

Reputation: 887881

We just need a cbind 

Lookup <- function(year, month) {
   lookup_table[cbind(
           year - 1899, 
             1 + month)   
     ]
 }

Lookup(1901, 2)
#[1] 200
Foo2Bar(1901,2,1)
#[1] 201

df$Bar <- Foo2Bar(df$Year, df$Month, df$Foo)
df
#  Year Month Foo Bar
#1 1901     2   1 201
#2 1901     1   2  22
#3 1900     2   3 103
#4 1902     1   4  34

Upvotes: 1

Related Questions