Mutate value by using a value from a different row in a tibble

Question

I want to calculate the distance a node to the root dtr. All I have is a vector, that contains the parent node id for each node rel (in this example id == 7 is root):

library(tidyverse)

tmp <- tibble(
  id = 1:12,
  rel = c(2,7,4,2,4,5,7,7,10,8,7,7)
)

In the end I'm looking for this result:

tmp$dtr

[1] 2 1 3 2 3 4 0 1 3 2 1 1

So far I was able to write the following algorithm until I got stuck when trying to reference a different row in my code.

The algorithm should work like this (Pseudocode):

1. If not root, increment dtr: if(!equals(tid,trel)): dtr = dtr+1
2. Change tid to trel: tid = trel
3. Change trel to to the rel value where id == trel
4. If any !equals(tid,trel) GOTO 1., else END

First I added 2 helper columns to store temporary information:

tmp <- tmp %>%
  mutate(
    tid = id,
    trel = rel,
    dtr = 0
  )

The first two steps in the algorithm work like this:

tmp <- tmp %>%
  mutate(
    dtr = if_else(
      !equals(tid,trel),
      dtr + 1,
      dtr
    ),
    tid = trel
  ) 

The 3rd step I'm not sure about.... I tried to achieve it with the following code, but that does not work:

tmp <- tmp %>% 
  mutate(trel = rel[id == .$tid])

The result is (of course) wrong:

tmp$rel

[1] 7 7 7 7 7 7 7 7 7 7 7 7

But why not this? (Should be the right solution when running 3. the first time):

[1] 2 7 2 7 2 4 7 7 10 8 7 7

The 4th step is done by checking if I have more than one unique value in trel:

while(length(unique(tmp$trel)) > 1){
  ...
}

Thus the full algorithm should somewhat look like this:

get_dtr <- function(tib){
  tmp <- tib %>%
    mutate(
      tid = id,
      trel = rel,
      dtr = 0
    )
  
  while(length(unique(tmp$trel)) > 1){
    tmp <- tmp %>%
      mutate(
        dtr = if_else(
          !equals(tid,trel),
          dtr + 1,
          dtr
        ),
        tid = trel
      ) 
    
    ### Step 3
  }
  tmp
}

Any idea how to solve this or a simpler solution? Thanks in advance!

Answer 1

This is basically already implemented in the tidygraph package. If you are going to be working with graph-like data with the tidyverse you should look there first. you can do

library(tidygraph)
as_tbl_graph(tmp, directed=FALSE) %>% 
  activate(nodes) %>% 
  mutate(depth=bfs_dist(root=7)) %>% 
  as_tibble()
#     name depth
#    <chr> <int>
#  1     1     2
#  2     2     1
#  3     3     3
#  4     4     2
#  5     5     3
#  6     6     4
#  7     7     0
#  8     8     1
#  9     9     3
# 10    10     2
# 11    11     1
# 12    12     1

Answer 2

If you want to write a function yourself, you can use the following code:

library(tidyverse)

tmp <- tibble(
  id = 1:12,
  rel = c(2,7,4,2,4,5,7,7,10,8,7,7)
)


calc_dtr <- function(id, tmp){
  # find root
  root <- tmp$id[tmp$id == tmp$rel]

  # is this the root node? 
  if(id == root){return(0)}

  # initialize counter
  dtr <- 1
  trel <- tmp$rel[tmp$id == id]

  while(trel != root){
    dtr <- dtr + 1
    trel <- tmp$rel[tmp$id == trel]
  }

  return(dtr)
}

tmp %>% 
  mutate(
    dtr = map_dbl(id, calc_dtr, tmp)
  )

This produces the following output:

# A tibble: 12 x 3
      id   rel   dtr
   <int> <dbl> <dbl>
 1     1     2     2
 2     2     7     1
 3     3     4     3
 4     4     2     2
 5     5     4     3
 6     6     5     4
 7     7     7     0
 8     8     7     1
 9     9    10     3
10    10     8     2
11    11     7     1
12    12     7     1