Building a tree from a two columns data frame

[SamGG]

Hi,
I would like to build a tree from a data frame that looks like the following one. Is there a function or a piece of code? I would prefer not using a lot of dependencies. I already looked at ape and igraph, but didn't find my way. Any help is welcome.

data.frame(
    parent = c('Root', 'Root', 'Project A','Project A','Project B','Project B','Root'),
    child = c('Project A','Project B','Study A-A', 'Study A-B','Project B','Study B-A', 'Project C')
)

[trafficonese]

This function should do the trick:

treelist <- vector("list", length(unique(df$parent)))
for (i in unique(df$parent)) {
  childs <- as.list(rep("", length(df[df$parent == i, "child"])))
  names(childs) <- df[df$parent == i, "child"]
  treelist[[i]] <- childs
}

Full Shiny Example:

df <- data.frame(
  parent = c('Root', 'Root', 'Project A','Project A','Project B','Project B','Root'),
  child = c('Project A','Project B','Study A-A', 'Study A-B','Project B','Study B-A', 'Project C')
)


treelist <- vector("list", length(unique(df$parent)))
for (i in unique(df$parent)) {
  childs <- as.list(rep("", length(df[df$parent == i, "child"])))
  names(childs) <- df[df$parent == i, "child"]
  treelist[[i]] <- childs
}


library(shiny)
library(shinyTree)

ui <- fluidPage(shinyTree("tree"))

server <- function(input, output, session) {
  output$tree <- renderTree({
    treelist
  })
}

shinyApp(ui, server)

[SamGG]

Thanks for your reply.

The result is the following in my hands. There are 3 extra nodes at the top, and the result is not exactly indented as I am expecting.

Here is what I would like (Gimp edition).

I am stuck on the nested structure.

BTW I notice that I introduced an error in the example as Project B is repeated incorrectly. Please do find the correct example.

df <- data.frame(
  parent = c('Root', 'Root', 'Project A','Project A','Project B','Project B','Root'),
  child = c('Project A','Project B','Study A-A', 'Study A-B','Study B-A','Study B-B', 'Project C')
)

[trafficonese]

sry my bad, if you remove
treelist <- vector("list", length(unique(df$parent)))
with
treelist <- list()

you get the tree I intended. Apparently preallocating a list was no good idea. ;)

That still doesnt give you the tree you want, but with your new example, theres no easy logic to build the tree you want. If you can transform your data.frame to this for example it would be much easier to build a function and it would make more sense.

NOTE: I included Study C-A for Project C. If you really want it blank, you will have to fill it with "" for example and filter that out in the loop.

df <- data.frame(stringsAsFactors = FALSE,
                 parent = c('Root',      'Root',       'Root',      'Root',       'Root'),
                 child1 = c('Project A', 'Project A',  'Project B', 'Project B',  'Project C'),
                 child2 = c('Study A-A', 'Study A-B',  'Study B-A',  'Study B-B', 'Study C-A')
)

Then the function would be quite similar:

library(shiny)
library(shinyTree)

treelist <- list()
for (i in unique(df$child1)) {
  childs <- as.list(rep("", length(df[df$child1 == i, 1])))
  names(childs) <- df[df$child1 == i, "child2"]
  treelist[[i]] <- childs
}
treelist <- list("Root" = treelist)


## Shiny Example ####
ui <- fluidPage(shinyTree("tree"))
server <- function(input, output, session) {
  output$tree <- renderTree({
    treelist
  })
}

shinyApp(ui, server)

[SamGG]

Thanks for the update. Really helpful. I will work on that this evening.
Thanks for your time. I will update the issue when I find the exact solution.

[YsoSirius]

If you want a more general approach, where "Project-C" can be an empty node and that works with multiple "root"-nodes you could use this approach:

## Project-C empty node
df <- data.frame(stringsAsFactors = FALSE,
                 parent = c('Root',      'Root',       'Root',      'Root',       'Root'),
                 child1 = c('Project A', 'Project A',  'Project B', 'Project B',  'Project C'),
                 child2 = c('Study A-A', 'Study A-B',  'Study B-A',  'Study B-B', '')
)
## Multiple root-nodes
# df <- data.frame(stringsAsFactors = FALSE,
#                  parent = c('Root',      'Root',       'Root',      'Root',       'Root',       'Root1',      'Root1',    'Root1',      'Root1'),
#                  child1 = c('Project A', 'Project A',  'Project B', 'Project B',  'Project C',  'Project A',  'Project A','Project B', 'Project B' ),
#                  child2 = c('Study A-A', 'Study A-B',  'Study B-A',  'Study B-B', '',           'Study A-A',  'Study A-B','Study B-A',  'Study B-B')
# )

treelist <- list()
for (j in unique(df$parent)) {
  tmp <- df[df$parent == j, ]
  subtreelist <- list()
  for (i in unique(tmp$child1)) {
    if (length(tmp[tmp$child1 == i, "child2"]) == 1 && tmp[tmp$child1 == i, "child2"] == "") {
      childs <- i
    } else {
      childs <- as.list(rep("", length(tmp[tmp$child1 == i, 1])))
      names(childs) <- tmp[tmp$child1 == i, "child2"]    
    }
    subtreelist[[i]] <- childs
  }
  treelist[[j]] <- subtreelist
}

[SamGG]

Thanks for your reply. In fact, I want the inverse transformation of the get_FlatList function, without any constraint. Here is how I exploit this function to get a data.frame view of the tree. This allows me to link the tree to some other information. Related to question #64, I would to use this data.frame structure to update the tree presented by shinyTree.

# This is the toy tree
> dput(tree)
list(root1 = structure(0, id = "1"), root2 = structure(list(SubListA = structure(list(
    leaf1 = structure(0, id = "4"), leaf2 = structure(0, id = "5"), 
    leaf3 = structure(0, id = "6")), id = "3"), SubListB = structure(list(
    leafA = structure(0, id = "8"), leafB = structure(0, id = "9")), id = "7"), 
    SubListC = structure(list(leaf1 = structure(0, id = "11"), 
        leaf2 = structure(0, stselected = TRUE, id = "12"), leaf3 = structure(0, id = "13")), stopened = TRUE, id = "10")), stopened = TRUE, id = "2"))
# One line conversion to data.frame
> as.data.frame(t(sapply(shinyTree:::get_flatList(tree), unlist)), stringsAsFactors = FALSE)
   id     text parent state.opened state.selected data.id
1   1    root1      #        FALSE          FALSE       1
2   2    root2      #         TRUE          FALSE       2
3   3 SubListA      2        FALSE          FALSE       3
4   4    leaf1      3        FALSE          FALSE       4
5   5    leaf2      3        FALSE          FALSE       5
6   6    leaf3      3        FALSE          FALSE       6
7   7 SubListB      2        FALSE          FALSE       7
8   8    leafA      7        FALSE          FALSE       8
9   9    leafB      7        FALSE          FALSE       9
10 10 SubListC      2         TRUE          FALSE      10
11 11    leaf1     10        FALSE          FALSE      11
12 12    leaf2     10        FALSE           TRUE      12
13 13    leaf3     10        FALSE          FALSE      13
# Is there any function to do the reverse conversion?

[msgoussi]

May I ask why you use :::: instead of :: and why there is no documentation help for this function?

[trafficonese]

The function get_flatList is not exported and just used internally by updateTree. Therefore there is no documentation for it.

You can still access the internal package functions with :::.

[patzaw]

inBranch <- function(branch){
   if(identical(branch, "")){
      return(NULL)
   }else{
      toRet <- unique(unlist(lapply(branch, inBranch)))
      toRet <- unique(c(toRet, names(branch)))
      return(toRet)
   }
}

getTree <- function(children){
   stopifnot(
      all(c("parent", "child") %in% colnames(children)),
      is.character(children$parent),
      is.character(children$child),
      all(children$child != children$parent)
   )
   leaf <- children %>% filter(!child %in% parent)
   leaf <- split(leaf$child, leaf$parent)
   tree <- lapply(
      leaf,
      function(x){
         toRet <- as.list(rep("", length(x)))
         names(toRet) <- x
         return(toRet)
      }
   )
   children <- children %>% filter(child %in% parent)
   while(nrow(children)>0){
      leaf <- children %>% filter(!child %in% parent)
      leaf <- split(leaf$child, leaf$parent)
      newTree <- lapply(
         names(leaf),
         function(x) c(tree[leaf[[x]]], tree[[x]])
      )
      names(newTree) <- names(leaf)
      tree <- c(tree[setdiff(names(tree), names(newTree))], newTree)
      children <- children %>% filter(child %in% parent)
   }
   toRm <- unique(unlist(lapply(tree, inBranch)))
   tree <- tree[setdiff(names(tree),toRm)]
   return(tree)
}

This function works in my hand.
However, your original example has a loop (loop: an ancestor is also a descendant of its descendant): there is one row in which child=parent (Project B). This case is not handled by my function and I don't think it should be. The parent/child table should be a table of identifiers to avoid ambiguity that cannot be handled if you have complex nested trees with descendants that can have multiple ancestors.

[woec]

Is it possible, to use the saved result of get_Flatlist to restore the selections of a tree by means of R scripting?

[trafficonese]

It is not possible with R only, but with some JavaScript it would be.
I wrote down one option here: https://stackoverflow.com/questions/57952813/save-and-restore-selections-in-a-shinytree/57959221#57959221