What solves my problem: Map, reduce or a recursion? - r

Question

I really need some help to write a recursion in R.
The function that I want changes a certain observation according to a set of comparisons between different rows in a data frame, which I shall call g. One of these comparisons depends on the previous value of this same observation.
Suppose first that I want to update the value of column index, row i in my data df in the following way:
j <- 1:4
g <- (df$dom[i] > 0 &
abs(df$V2009[i] - df$V2009[j]) <= w) |
df$index[i] == df$index[j]
df$index[i] <- ifelse(any(g), which(g)[[1]], df$index[[i]])
The thing is, the object w is actually a list:
w = list(0, 1, 2, df$age[i])
So, as you can see, I want to create a function foo() that updates df$index iteratively. It changes it by looping through w and comparisons depend on updated values.
Here is some data:
df <- data.frame(dom = c(0, 0, 6, 6),
V2009 = c(9, 11, 9, 11),
index = c(1, 2, 1, 2),
age = c(2, 2, 2, 2))
I am not sure if a recursive function is actually needed or if something like reduce or map would do it.
Thank you!

Answer 1

The following function uses a double for loop to change the values of column index according to the condition defining g. It accepts a data.frame as input and returns the updated data.frame.
foo <- function(x){
change_index <- function(x, i, w){
j <- seq_len(nrow(x))
(x$dom[i] > 0 & abs(x$V2009[i] - x$V2009[j]) <= w) |
x$index[i] == x$index[j]
}
for(i in seq_len(nrow(x))){
W <- list(0, 1, 2, x$age[i])
for(w in W){
g <- change_index(x, i, w)
if(any(g)) x$index[i] <- which(g)[1]
}
}
x
}
foo(df)
# dom V2009 index age
#1 0 9 1 2
#2 0 11 2 2
#3 6 9 1 2
#4 6 11 1 2

Answer 2

One can define w inside a function and use lexical scoping (closure).
Using your instructions, the function index_value calculates for any given i the index value.
correct_index_col returns the corrected df.
df <- data.frame(dom = c(0, 0, 6, 6),
V2009 = c(9, 11, 9, 11),
index = c(1, 2, 1, 2),
age = c(2, 2, 2, 2))
index_value <- function(df, i) {
j <- nrow(df)
w <- c(0, 1, 2, df$age[i])
g <- (df$dom[i] > 0 & abs(df$V2009[i] - df$V2009[j]) <= w) |
df$index[i] == df$index[j]
ifelse(any(g), which(g)[[1]], df$index[[i]])
}
correct_index_col <- function(df) {
indexes <- Vectorize(function(i) {
index_value(df, i)
})
df$index <- indexes(1:nrow(df))
df
}
# > correct_index_col(df)
# dom V2009 index age
# 1 0 9 1 2
# 2 0 11 1 2
# 3 6 9 3 2
# 4 6 11 1 2
#
If you want to really update (mutate) your df, then you have to do
df <- correct_index_col(df).

Answer 3

Here is an attempt of my own. I guess I figured out a way to use recursion over mutate:
test <- function(i, df, k){
j <- 1:nrow(df)
w <- list(0, 1, 2, df$age[i])
g <- (df$dom[i] > 0 & abs(df$V2009[i] - df$V2009[j]) <= w[k]) |
df$index[i] == df$index[j]
l <- ifelse(any(g), which(g)[1], df$index[i])
return(l)
}
loop <- function(data,
k = 1) {
data <- data %>%
mutate(index = map_dbl(seq(n()),
~ test(.x, df = cur_data(), k)))
if (k == 4) {
return(data)
} else {
return(loop(data, k + 1))
}
}
df %>% loop()
I welcome any comments in case this is inefficient considering large datasets