The power of three: purrr-poseful iteration in R with map, pmap and imap

The purrr package is a functional programming superstar which provides useful tools for iterating through lists and vectors, generalizing code and removing programming redundancies. The purrr tools work in combination with functions, lists and vectors and results in code that is consistent and concise.

In this post we focus primarily on the map family of functions which, at their simplest, offer an alternative to the apply functions and an alternative to for loops. For more complex uses map functions can be used to tidily manipulate multi-dimensional datasets and apply statistical models, for example.

This post explores several of the map family workhorses: map, pmap and imap. It assumes a basic familiarity with the map functions and their variants (i.e. map_lgl, map_chr, etc). If you need a refresher we recommend this documentation. For advanced users looking to speed up their purrr calculations we recommend our previous blog post on parellelizing your computations using the furrr package.

Load the packages

library(purrr)        # Functional programming
library(dplyr)        # Data wrangling
library(tidyr)        # Tidy-ing data
library(stringr)      # String operations
library(repurrrsive)  # Game of Thrones data
library(tidygraph)    # Convert data into node/edge format
library(ggraph)       # Network graphing

Get the data

For our example, we’re using the “Game of Thrones” dataset from the repurrrsive package.

dat <- got_chars

The dataset is a list of 30 lists containing various information on Game of Thrones characters. The first list looks something like this:

glimpse(dat[[1]])
## List of 18
##  $ url        : chr "https://www.anapioficeandfire.com/api/characters/1022"
##  $ id         : int 1022
##  $ name       : chr "Theon Greyjoy"
##  $ gender     : chr "Male"
##  $ culture    : chr "Ironborn"
##  $ born       : chr "In 278 AC or 279 AC, at Pyke"
##  $ died       : chr ""
##  $ alive      : logi TRUE
##  $ titles     : chr [1:3] "Prince of Winterfell" "Captain of Sea Bitch" "Lord of the Iron Islands (by law of the green lands)"
##  $ aliases    : chr [1:4] "Prince of Fools" "Theon Turncloak" "Reek" "Theon Kinslayer"
##  $ father     : chr ""
##  $ mother     : chr ""
##  $ spouse     : chr ""
##  $ allegiances: chr "House Greyjoy of Pyke"
##  $ books      : chr [1:3] "A Game of Thrones" "A Storm of Swords" "A Feast for Crows"
##  $ povBooks   : chr [1:2] "A Clash of Kings" "A Dance with Dragons"
##  $ tvSeries   : chr [1:6] "Season 1" "Season 2" "Season 3" "Season 4" ...
##  $ playedBy   : chr "Alfie Allen"

The map function

The map function iteratively applies a function or formula to each element of a list or vector. The operation is similar to a for loop but with fewer keystrokes and cleaner code. The result of applying map will be the same length as the input. Since purrr functions are type-stable there is little guesswork in knowing which type of output will be returned. For example map_chr() returns character vectors, map_dbl() returns double vectors, etc.

Example 1: Extract a single element from a list

Extracting an element from a list can be done a number of ways. The following code chunks will produce the same results.

# Method 1: using the name of the list element, similar to dat[[1]]["name"], dat[[2]]["name"], etc
map(dat, "name")

# Method 2: using the `pluck` function
map(dat, pluck("name"))

# Method 3: using the index of the list element
map(dat, 3)
## [[1]]
## [1] "Theon Greyjoy"
## 
## [[2]]
## [1] "Tyrion Lannister"
## 
## [[3]]
## [1] "Victarion Greyjoy"
## 
## [[4]]
## [1] "Will"
## 
## [[5]]
## [1] "Areo Hotah"

Example 2a: Create a dataframe from a list (easier)

Create a dataframe from several of the list items. This method will only work if the element you’re requesting (in this case name, gender and culture) has a length of 1.

What this is doing is the equivalent to dat[[1]][c("name", "gender", "culture")], dat[[2]][c("name", "gender", "culture")] and so on. The _dfr piece tells map to convert the result to a data.frame by row.

# The `[` is the function here -- essentially telling it to apply [] to each list
# and the name, gender and culter are the argument passed to []
map_dfr(dat,`[`, c("name", "gender", "culture"))
## # A tibble: 30 x 3
##   name              gender culture 
##   <chr>             <chr>  <chr>   
## 1 Theon Greyjoy     Male   Ironborn
## 2 Tyrion Lannister  Male   ""      
## 3 Victarion Greyjoy Male   Ironborn
## 4 Will              Male   ""      
## 5 Areo Hotah        Male   Norvoshi
## 6 Chett             Male   ""      
## # ... with 24 more rows

Example 2b: Create a dataframe from a list (harder)

If you try running the code above but add “aliases” and “allegiances” you should get the following error: Error: Argument 4 must be length 1, not 4.

x <- map_dfr(dat,`[`, c("name", "gender", "culture", "aliases", "allegiances"))
## Error: Argument 4 must be length 1, not 4

Take a closer look at the list elements “aliases” and “allegiances”. You’ll notice that some inputs are character vectors of length > 1.

glimpse(map(dat, "aliases"))
## List of 6
##  $ : chr [1:4] "Prince of Fools" "Theon Turncloak" "Reek" "Theon Kinslayer"
##  $ : chr [1:11] "The Imp" "Halfman" "The boyman" "Giant of Lannister" ...
##  $ : chr "The Iron Captain"
##  $ : chr ""
##  $ : chr ""
##  $ : chr ""

In order to include these items in our dataframe we’ll need to create a list-column using map.

dat_m <- dat %>% {
  tibble(
    name = map_chr(., "name"),
    gender = map_chr(., "gender"),
    culture = map_chr(., "culture"),
    aliases = map(., "aliases"),
    allegiances = map(., "allegiances")
)}
## # A tibble: 30 x 5
##   name              gender culture  aliases    allegiances
##   <chr>             <chr>  <chr>    <list>     <list>     
## 1 Theon Greyjoy     Male   Ironborn <chr [4]>  <chr [1]>  
## 2 Tyrion Lannister  Male   ""       <chr [11]> <chr [1]>  
## 3 Victarion Greyjoy Male   Ironborn <chr [1]>  <chr [1]>  
## 4 Will              Male   ""       <chr [1]>  <NULL>     
## 5 Areo Hotah        Male   Norvoshi <chr [1]>  <chr [1]>  
## # ... with 25 more rows

Example 3: Apply a custom function to a list

Write a function that outputs a statement indicating whether a character is alive or dead. Note that we’re using map_chr which will output a character vector instead of a list.

Also note that with the final season of Game of Thrones we already know that not all of these are true anymore!

dead_or_alive <- function(x){
  ifelse(x[["alive"]], paste(x[["name"]], "is alive!"),
    paste(x[["name"]], "is dead :("))
}
map_chr(dat, dead_or_alive)
## [1] "Theon Greyjoy is alive!"     "Tyrion Lannister is alive!" 
## [3] "Victarion Greyjoy is alive!" "Will is dead :("            
## [5] "Areo Hotah is alive!"        "Chett is dead :("

Example 4: Bonus, apply a custom function and create a network graph

Create a plot of all of Jon Snow’s aliases. First we’ll use compact to remove elements from dat that have length zero or are NULL. For example if you look at Cersei’s aliases you’ll see an empty list. Using compact will remove this empty list.

# Empty list for Cersei Lannister's `aliases`
dat[[19]]$aliases
## list()
g <- map(dat, compact)

Write a function that pulls the name and aliases elements if aliases exists. Create a tibble and using the function tidygraph::as_tbl_graph convert the data into proper node and edge data that can be used for the plot.

also_known_as <- function(x){
  
  if ("aliases" %in% names(x)){
    g <- tibble(
      from = x$name,
      to = x$aliases)
      
    g <- as_tbl_graph(g)
  }
}
g <- map(g, also_known_as)

The tbl_graph format should look something like this:

g[[1]]
## # A tbl_graph: 5 nodes and 4 edges
## #
## # A rooted tree
## #
## # Node Data: 5 x 1 (active)
##   name           
##   <chr>          
## 1 Theon Greyjoy  
## 2 Prince of Fools
## 3 Theon Turncloak
## 4 Reek           
## 5 Theon Kinslayer
## #
## # Edge Data: 4 x 2
##    from    to
##   <int> <int>
## 1     1     2
## 2     1     3
## 3     1     4
## # ... with 1 more row

Finally plot the data using ggraph. Jon Snow’s index within our g object is 23 – so we’ll use g[[23]] for his data. Below are all of Jon Snow’s aliases!

ggraph(g[[23]], layout = "graphopt") + 
  geom_edge_link() + 
  geom_node_label(aes(label = name), 
    label.padding = unit(1, "lines"), 
    label.size = 0) +
  theme_graph()

The pmap function

The pmap function can be used on an arbitrary number of inputs and is great for doing row-wise iterations on a dataframe.

Example 1: Apply a function to each row

Using the pmap function we can apply a row-wise operation to our dataset. As a reminder here’s what our dat_m object looks like:

## # A tibble: 30 x 5
##   name              gender culture  aliases    allegiances
##   <chr>             <chr>  <chr>    <list>     <list>     
## 1 Theon Greyjoy     Male   Ironborn <chr [4]>  <chr [1]>  
## 2 Tyrion Lannister  Male   ""       <chr [11]> <chr [1]>  
## 3 Victarion Greyjoy Male   Ironborn <chr [1]>  <chr [1]>  
## 4 Will              Male   ""       <chr [1]>  <NULL>     
## 5 Areo Hotah        Male   Norvoshi <chr [1]>  <chr [1]>  
## # ... with 25 more rows

If we use pmap and apply the paste function but do not specify column names, the result will be all columns pasted together by row. Note that the aliases column for Theon and Tyrion includes 4 and 11 entries respectively – and each alias is pasted with gender, culture and allegiance so you end up with 4 and 11 strings

pmap(dat_m, paste)
## [[1]]
## [1] "Theon Greyjoy Male Ironborn Prince of Fools House Greyjoy of Pyke"
## [2] "Theon Greyjoy Male Ironborn Theon Turncloak House Greyjoy of Pyke"
## [3] "Theon Greyjoy Male Ironborn Reek House Greyjoy of Pyke"           
## [4] "Theon Greyjoy Male Ironborn Theon Kinslayer House Greyjoy of Pyke"
## 
## [[2]]
##  [1] "Tyrion Lannister Male  The Imp House Lannister of Casterly Rock"           
##  [2] "Tyrion Lannister Male  Halfman House Lannister of Casterly Rock"           
##  [3] "Tyrion Lannister Male  The boyman House Lannister of Casterly Rock"        
##  [4] "Tyrion Lannister Male  Giant of Lannister House Lannister of Casterly Rock"
##  [5] "Tyrion Lannister Male  Lord Tywin's Doom House Lannister of Casterly Rock" 
##  [6] "Tyrion Lannister Male  Lord Tywin's Bane House Lannister of Casterly Rock" 
##  [7] "Tyrion Lannister Male  Yollo House Lannister of Casterly Rock"             
##  [8] "Tyrion Lannister Male  Hugor Hill House Lannister of Casterly Rock"        
##  [9] "Tyrion Lannister Male  No-Nose House Lannister of Casterly Rock"           
## [10] "Tyrion Lannister Male  Freak House Lannister of Casterly Rock"             
## [11] "Tyrion Lannister Male  Dwarf House Lannister of Casterly Rock"

Example 2: Apply a function to each row using column names

Apply another function to the table but this time specify the column names.

Initial prep before using pmap

To make this more interesting let’s filter to any character with a Stark or Lannister allegiance. We can do this by extracting “Lannister” or “Stark” from the list of allegiances. Here’s an example of what the allegiances column looks like:

dat_m$allegiances[16:18]
## [[1]]
## [1] "House Stark of Winterfell"
## 
## [[2]]
## [1] "House Baratheon of Storm's End" "House Stark of Winterfell"     
## [3] "House Tarth of Evenfall Hall"  
## 
## [[3]]
## [1] "House Stark of Winterfell" "House Tully of Riverrun"

We will use str_extract() from the stringr package to pull out the words “Lannister” or “Stark. Note that the ~ and .x are shorthand for function(x) and x as an example. You can see that it returns a vector with NA values if neither ”Lannister" nor “Stark” exist.

dat_p <- mutate(dat_m,
  stark_or_lannister = map(allegiances, ~str_extract(.x, "Lannister|Stark")))

dat_p$stark_or_lannister[16:18]
## [[1]]
## [1] "Stark"
## 
## [[2]]
## [1] NA      "Stark" NA     
## 
## [[3]]
## [1] "Stark" NA

We can drop the NA values with the handy discard() function.

dat_p <- mutate(dat_p, stark_or_lannister = map(stark_or_lannister, ~discard(.x, is.na)))

dat_p$stark_or_lannister[16:18]
## [[1]]
## [1] "Stark"
## 
## [[2]]
## [1] "Stark"
## 
## [[3]]
## [1] "Stark"

Finally, we can filter to those that have a Lannister or Stark allegiance and then use unnest() to essentially convert the stark_or_lannister column to a traditional character column. Careful though, if there happened to be a character with both a Lannister and Stark allegiance and thus the stark_or_lannister column had two entries unnest() would create two rows for that character.

dat_p <- filter(dat_p, stark_or_lannister %in% c("Lannister", "Stark")) %>%
  unnest(stark_or_lannister) 
## # A tibble: 11 x 6
##   name            gender culture   aliases    allegiances stark_or_lannist~
##   <chr>           <chr>  <chr>     <list>     <list>      <chr>            
## 1 Tyrion Lannist~ Male   ""        <chr [11]> <chr [1]>   Lannister        
## 2 Arya Stark      Female Northmen  <chr [16]> <chr [1]>   Stark            
## 3 Brandon Stark   Male   Northmen  <chr [3]>  <chr [1]>   Stark            
## 4 Brienne of Tar~ Female ""        <chr [3]>  <chr [3]>   Stark            
## 5 Catelyn Stark   Female Rivermen  <chr [5]>  <chr [2]>   Stark            
## 6 Cersei Lannist~ Female Westerman <NULL>     <chr [1]>   Lannister
## 7 Eddard Stark    Male   Northmen  <chr [3]>  <chr [1]>   Stark            
## 8 Jamie Lannist~  Male   Westerla~ <chr [4]>  <chr [1]>   Stark            
## 9 Jon Snow        Male   Northmen  <chr [8]>  <chr [1]>   Stark            
## 10 Kevin Lannist~ Male   ""        <chr [1]>  <chr [1]>   Stark            
## 11 Sansa Stark    Female Northmen  <chr [3]>  <chr [2]>   Stark            

Use pmap

Write a function called whom_can_you_trust that outputs a string using 3 columns from the table. Note that we are using the ... (ellipsis) which allows us to apply the function to our table which contains more than the three columns. If we were to remove the ellipsis we would get an error message.

whom_can_you_trust <- function(name, allegiances, stark_or_lannister, ...) {
  y <- glue::glue("{name} has an allegiance to the {stark_or_lannister} family")
    ifelse(length(allegiances) > 1, 
    glue::glue("{y} but also has {length(allegiances)-1} other allegiance(s)."), 
    glue::glue("{y} and no other allegiances."))
}
dat_p %>% pmap_chr(whom_can_you_trust)
##  [1] "Tyrion Lannister has an allegiance to the Lannister family and no other allegiances."      
##  [2] "Arya Stark has an allegiance to the Stark family and no other allegiances."                
##  [3] "Brandon Stark has an allegiance to the Stark family and no other allegiances."             
##  [4] "Brienne of Tarth has an allegiance to the Stark family but also has 2 other allegiance(s)."
##  [5] "Catelyn Stark has an allegiance to the Stark family but also has 1 other allegiance(s)."   
##  [6] "Cersei Lannister has an allegiance to the Lannister family and no other allegiances."      
##  [7] "Eddard Stark has an allegiance to the Stark family and no other allegiances."              
##  [8] "Jaime Lannister has an allegiance to the Lannister family and no other allegiances."       
##  [9] "Jon Snow has an allegiance to the Stark family and no other allegiances."                  
## [10] "Kevan Lannister has an allegiance to the Lannister family and no other allegiances."       
## [11] "Sansa Stark has an allegiance to the Stark family but also has 1 other allegiance(s)."

The imap function

The imap function can generally be thought of as indexed map. The function uses 2 arguments: the first is the value, the second is the position or the name.

  • For objects that do not have names it is shorthand for map2(x, seq_along(x), ...)
  • For objects that have names it is shorthand for map2(x, names(x), ...)

Example 1: Apply imap to character vector

Create a character vector of 10 names. Using imap ouput the name (.x) and the index of name (.y).

v <- sample(dat_m$name, 10)

imap_chr(v, ~ paste0(.y, ": ", .x))
##  [1] "1: Jon Snow"           "2: Asha Greyjoy"      
##  [3] "3: Daenerys Targaryen" "4: Eddard Stark"      
##  [5] "5: Brienne of Tarth"   "6: Melisandre"        
##  [7] "7: Kevan Lannister"    "8: Davos Seaworth"    
##  [9] "9: Victarion Greyjoy"  "10: Sansa Stark"

Example 2: Apply imap to list with names

For this example we create a new list called dat_i. The new list has named elements but no longer contains the name field for demonstration purposes.

dat_i <- dat_m %>% split(.$name) %>%
  map(., ~select(.x, -name))
## $`Aeron Greyjoy`
## # A tibble: 1 x 4
##   gender culture  aliases   allegiances
##   <chr>  <chr>    <list>    <list>     
## 1 Male   Ironborn <chr [2]> <chr [1]>  
## 
## $`Areo Hotah`
## # A tibble: 1 x 4
##   gender culture  aliases   allegiances
##   <chr>  <chr>    <list>    <list>     A tibble: 6 x 6  
## 1 Male   Norvoshi <chr [1]> <chr [1]>  
## 
## $`Arianne Martell`
## # A tibble: 1 x 4
##   gender culture aliases   allegiances
##   <chr>  <chr>   <list>    <list>     
## 1 Female Dornish <chr [1]> <chr [1]>

Using imap and mutate we can apply the name of the list element (.y) as a new column in our table (.x).

dat_i <- imap(dat_i, ~mutate(.x, name = .y))
## $`Aeron Greyjoy`
## # A tibble: 1 x 5
##   gender culture  aliases   allegiances name         
##   <chr>  <chr>    <list>    <list>      <chr>        
## 1 Male   Ironborn <chr [2]> <chr [1]>   Aeron Greyjoy
## 
## $`Areo Hotah`
## # A tibble: 1 x 5
##   gender culture  aliases   allegiances name      
##   <chr>  <chr>    <list>    <list>      <chr>     
## 1 Male   Norvoshi <chr [1]> <chr [1]>   Areo Hotah
## 
## $`Arianne Martell`
## # A tibble: 1 x 5
##   gender culture aliases   allegiances name           
##   <chr>  <chr>   <list>    <list>      <chr>          
## 1 Female Dornish <chr [1]> <chr [1]>   Arianne Martell

Summary

The purrr family of functions are an excellent choice for streamlining your code and removing programming redundancies. In this post we highlighted three of our favorite purrr functions: map, pmap and imap, plus some bonus functions like discard and compact. For a complete list of purrr functions check out the purrr cheat sheet. Happy coding!



2 responses

Leave a Reply to reyhan Cancel reply

Your email address will not be published. Required fields are marked *