Introduction to forestry

Jiena Gu McLellan

2023-11-20

Built on the top of data.tree, a Node (tree) is an R6 object that is especially useful when we are facing hierarchical data. The forestry package helps to reshape or create tree objects. This package is a series of utility functions to help with nested data. Since data.tree has the capability to convert a tree to JSON using toJSON() after converting to a list using as.list(), the forestry package is particularly useful when creating a specific JSON object for building htmlwidgets. The forestry package aims to reshape or create tree objects with a specific format.

Create a Node with Assigned Attributes

create_nodes() creates a Node object. tree_name is to assign the name of this Node. add_children_count is to assign the number of children to this Node, it will be listed in numerical order. To assign values to each node, simply put the appropriate variable as a parameter with a vector containing the values. The name of the parameter will be the variable name and the values in the vector will be assigned to each node respectively.

library(data.tree)
library(forestry)
new_node <- create_nodes(tree_name = "tree1", 
                         add_children_count = 3, 
                         class = c("A", "B", "C") )
print(new_node, "class")
#>   levelName class
#> 1     tree1      
#> 2      ¦--1     A
#> 3      ¦--2     B
#> 4      °--3     C

Fill Missing Values Across a Level

The fill_NA_level() function will fill missing values across the desired level with desired value (default as 0). For example, new_node is a tree with missing value in hc field.

new_node <- create_nodes(tree_name = "tree1", 
                         add_children_count = 3, 
                         hc = c(1, 2, NA))
print(new_node, "hc" )
#>   levelName hc
#> 1     tree1 NA
#> 2      ¦--1  1
#> 3      ¦--2  2
#> 4      °--3 NA

We apply fill_NA_level() to new_node, simply put new_node as input_node, assign the field_name with hc, and assign by_level = 2, we will fill the NA in hc field with 0 across level 2.

result <- fill_NA_level(input_node = new_node, 
                        field_name = "hc", 
                        by_level = 2, 
                        fill_with = 0)
print(result, "hc")
#>   levelName hc
#> 1     tree1 NA
#> 2      ¦--1  1
#> 3      ¦--2  2
#> 4      °--3  0

Create a Tree From a List

create_tree() creates a new tree from a list. It appends each item of the input list as a numbered child in the new tree. This is useful when we convert a Node to a JSON array.

For instance, let’s use test_node$children (a list) as an example. We can see a list of groupA, groupB and groupC.

data(test_df)
test_node <- data.tree::as.Node(test_df)
print(test_node$children)
#> $groupA
#>    levelName
#> 1 groupA    
#> 2  ¦--Male  
#> 3  °--Female
#> 
#> $groupB
#>    levelName
#> 1 groupB    
#> 2  ¦--Male  
#> 3  °--Female
#> 
#> $groupC
#>    levelName
#> 1 groupC    
#> 2  ¦--Male  
#> 3  °--Female

Now we see that this list is reshaped into a list, new_tree, with each item in test_node$children added as a child. The index of each item in the list is assigned as the name of each child.

library(data.tree)
test_node <- as.Node(test_df)
new_shape <- create_tree(test_node$children,"new_tree")
print(new_shape, "hc")
#>             levelName hc
#> 1  new_tree           NA
#> 2   ¦--1              NA
#> 3   ¦   °--groupA     NA
#> 4   ¦       ¦--Male   80
#> 5   ¦       °--Female 97
#> 6   ¦--2              NA
#> 7   ¦   °--groupB     NA
#> 8   ¦       ¦--Male   44
#> 9   ¦       °--Female 37
#> 10  °--3              NA
#> 11      °--groupC     NA
#> 12          ¦--Male   81
#> 13          °--Female 46

Expand Children Nodes

fix_items() creates a tree with fixed children nodes from another tree. It automatically copies fields to the tree and fills missing values with NA. Similar to left joining to a tree with certian children nodes.

This function is to make sure the tree has the desired children nodes.

See cell_node2, it has only B and C.

cell_node2 <- Node$new("cell2")
cell_node2$AddChild("B")
cell_node2$AddChild("C")
cell_node2$Set(class = c(NA, "B1", "C1"))
print(cell_node2, "class")
#>   levelName class
#> 1     cell2      
#> 2      ¦--B    B1
#> 3      °--C    C1

Now we put fix_vector = c("A", "B", "C", "D") and assign to a new tree, cell_fixed_items. We can see that cell_fixed_items has all of the nodes from fix_vector and still inherits the fields from cell_node2.

cell_fixed_items <- fix_items(fix_vector = c("A", "B", "C", "D"), 
                              input_node = cell_node2)
print(cell_fixed_items, "class")
#>   levelName class
#> 1     cell2      
#> 2      ¦--A      
#> 3      ¦--B    B1
#> 4      ¦--C    C1
#> 5      °--D

Sort Chidren Nodes

children_sort() function sorts the children nodes into a desired order. If there are children nodes not listed in the input_order, we can set the mismatch_last parameter (default is T) to put the mismatched children nodes to the top or bottom.

data(test_df)
test_node <- data.tree::as.Node(test_df)
sorted_node <- children_sort(
  input_node = test_node, 
  input_order = c("groupB", "groupA"),
  mismatch_last = T)
print(sorted_node)
#>         levelName
#> 1  tree1         
#> 2   ¦--groupB    
#> 3   ¦   ¦--Male  
#> 4   ¦   °--Female
#> 5   ¦--groupA    
#> 6   ¦   ¦--Male  
#> 7   ¦   °--Female
#> 8   °--groupC    
#> 9       ¦--Male  
#> 10      °--Female

Cumulative Sum Across a Level

cumsum_across_level() gets the cumulative value across a level, the cumulative value will be added to the cumsum_number field.

In this example, it calculates the cumulative exercise_time field across level 3.

data(exercise_df)
exercise_node <- as.Node(exercise_df)
test <- forestry::cumsum_across_level(input_node = exercise_node, 
                              attri_name = "exercise_time", 
                              level_num = 3)
print(test, "cumsum_number", "exercise_time", "level")
#>      levelName cumsum_number exercise_time level
#> 1  Year                   NA            NA     1
#> 2   ¦--Q1                 NA            NA     2
#> 3   ¦   ¦--Jan          0.83          0.83     3
#> 4   ¦   ¦--Feb          1.14          0.31     3
#> 5   ¦   °--Mar          1.98          0.84     3
#> 6   ¦--Q2                 NA            NA     2
#> 7   ¦   ¦--Apr          2.17          0.19     3
#> 8   ¦   ¦--May          2.18          0.01     3
#> 9   ¦   °--Jun          2.45          0.27     3
#> 10  ¦--Q3                 NA            NA     2
#> 11  ¦   ¦--Jul          2.56          0.11     3
#> 12  ¦   ¦--Aug          3.54          0.98     3
#> 13  ¦   °--Sep          4.30          0.76     3
#> 14  °--Q4                 NA            NA     2
#> 15      ¦--Oct          4.49          0.19     3
#> 16      ¦--Nov          5.25          0.76     3
#> 17      °--Dec          5.54          0.29     3

In addition, level_num = "All" will get the cumulative value across all levels. Please note that there should be no missing values in the appropriate level when we apply cumsum_across_level().

data(exercise_df)
exercise_node <- as.Node(exercise_df)
exercise_node$Do(function(node) node$exercise_time <- Aggregate(node, 
                                                   attribute = "exercise_time", 
                                                   aggFun = sum), 
             traversal = "post-order")
print(exercise_node,  "exercise_time")
#>      levelName exercise_time
#> 1  Year                 5.54
#> 2   ¦--Q1               1.98
#> 3   ¦   ¦--Jan          0.83
#> 4   ¦   ¦--Feb          0.31
#> 5   ¦   °--Mar          0.84
#> 6   ¦--Q2               0.47
#> 7   ¦   ¦--Apr          0.19
#> 8   ¦   ¦--May          0.01
#> 9   ¦   °--Jun          0.27
#> 10  ¦--Q3               1.85
#> 11  ¦   ¦--Jul          0.11
#> 12  ¦   ¦--Aug          0.98
#> 13  ¦   °--Sep          0.76
#> 14  °--Q4               1.24
#> 15      ¦--Oct          0.19
#> 16      ¦--Nov          0.76
#> 17      °--Dec          0.29
exercise_node_test <- cumsum_across_level(input_node = exercise_node, 
                              attri_name = "exercise_time", 
                              level_num = "All")
print(exercise_node_test,"exercise_time", "cumsum_number", "level")
#>      levelName exercise_time cumsum_number level
#> 1  Year                 5.54            NA     1
#> 2   ¦--Q1               1.98          1.98     2
#> 3   ¦   ¦--Jan          0.83          0.83     3
#> 4   ¦   ¦--Feb          0.31          1.14     3
#> 5   ¦   °--Mar          0.84          1.98     3
#> 6   ¦--Q2               0.47          2.45     2
#> 7   ¦   ¦--Apr          0.19          2.17     3
#> 8   ¦   ¦--May          0.01          2.18     3
#> 9   ¦   °--Jun          0.27          2.45     3
#> 10  ¦--Q3               1.85          4.30     2
#> 11  ¦   ¦--Jul          0.11          2.56     3
#> 12  ¦   ¦--Aug          0.98          3.54     3
#> 13  ¦   °--Sep          0.76          4.30     3
#> 14  °--Q4               1.24          5.54     2
#> 15      ¦--Oct          0.19          4.49     3
#> 16      ¦--Nov          0.76          5.25     3
#> 17      °--Dec          0.29          5.54     3

Prepare for JSON array

The pre_get_array() function changes the numeric item name in a list into a format that is compatible with the JSON array standard. As mentioned earlier, when converting a tree to JSON, we need to save the tree as a list using as.list() then use htmlwidgets:::toJSON() to convert the list to JSON data.

For example, new_node is a tree with numeric children nodes.

new_node <- create_nodes(tree_name = "tree1", 
                         add_children_count = 3, 
                         class = c("A", "B", "C"))
print(as.list(new_node) )
#> $name
#> [1] "tree1"
#> 
#> $`1`
#> $`1`$class
#> [1] "A"
#> 
#> 
#> $`2`
#> $`2`$class
#> [1] "B"
#> 
#> 
#> $`3`
#> $`3`$class
#> [1] "C"

We can see the numeric children node names are listed. If we apply pre_get_array() to this list, we can change all numeric names so the nodes can be saved as a JSON array instead of JSON objects after we use htmlwidgets:::toJSON().

new_node <- create_nodes(tree_name = "tree1", 
                         add_children_count = 3, 
                         class = c("A", "B", "C"))
print(pre_get_array(as.list(new_node) ) )
#> [[1]]
#> [1] "tree1"
#> 
#> [[2]]
#> [[2]]$class
#> [1] "A"
#> 
#> 
#> [[3]]
#> [[3]]$class
#> [1] "B"
#> 
#> 
#> [[4]]
#> [[4]]$class
#> [1] "C"