purrr workshop

Slides to complement a hands-on workshop on the R package purrr (https://purrr.tidyverse.org)

Transcript

1. Jennifer Bryan 
 RStudio  @JennyBryan  @jennybc How to repeat yourself with purrr
2. This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. To view a copy of this license, visit 
 http://creativecommons.org/licenses/by-sa/4.0/
3. R installed? Pretty recent? • Current version: 3.5.1 RStudio installed? • Current Preview: 1.2.907 Have these packages? • tidyverse (includes purrr) • repurrrsive Get some help NOW if you need/want to do some setup during the intro!
4. rstd.io/purrr-latinr
5. bit.ly/jenny-live-code
6. Resources My purrr materials: https://jennybc.github.io/purrr-tutorial/ Charlotte Wickham's purrr materials: https://github.com/cwickham/purrr-tutorial My "row-oriented workflows" materials: rstd.io/row-work "Functionals" chapter of 2nd of Advanced R by Wickham https://adv-r.hadley.nz/functionals.html
7. 1. What is the harm with copy/paste and repetitive code? 2. What should I do instead? - write functions (R-Ladies Thursday) - use formal tools to iterate the R way 3. Hands-on practice with the purrr package for iteration
8. library(gapminder) library(tidyverse) gapminder #> # A tibble: 1,704 x 6 #> country continent year lifeExp pop gdpPercap #> <fct> <fct> <int> <dbl> <int> <dbl> #> 1 Afghanistan Asia 1952 28.8 8425333 779. #> 2 Afghanistan Asia 1957 30.3 9240934 821. #> 3 Afghanistan Asia 1962 32.0 10267083 853. #> 4 Afghanistan Asia 1967 34.0 11537966 836. #> 5 Afghanistan Asia 1972 36.1 13079460 740. #> 6 Afghanistan Asia 1977 38.4 14880372 786. #> 7 Afghanistan Asia 1982 39.9 12881816 978. #> 8 Afghanistan Asia 1987 40.8 13867957 852. #> 9 Afghanistan Asia 1992 41.7 16317921 649. #> 10 Afghanistan Asia 1997 41.8 22227415 635. #> # ... with 1,694 more rows
9. gapminder %>% count(continent) #> # A tibble: 5 x 2 #> continent n #> <fct> <int> #> 1 Africa 624 #> 2 Americas 300 #> 3 Asia 396 #> 4 Europe 360 #> 5 Oceania 24
12. africa <- gapminder[gapminder$continent == "Africa", ] africa_mm <- max(africa$lifeExp) - min(africa$lifeExp) americas <- gapminder[gapminder$continent == "Americas", ] americas_mm <- max(americas$lifeExp) - min(americas$lifeExp) asia <- gapminder[gapminder$continent == "Asia", ] asia_mm <- max(asia$lifeExp) - min(africa$lifeExp) europe <- gapminder[gapminder$continent == "Europe", ] europe_mm <- max(europe$lifeExp) - min(europe$lifeExp) oceania <- gapminder[gapminder$continent == "Oceania", ] oceania_mm <- max(europe$lifeExp) - min(oceania$lifeExp) cbind( continent = c("Africa", "Asias", "Europe", "Oceania"), max_minus_min = c(africa_mm, americas_mm, asia_mm, europe_mm, oceania_mm) )
13. What am I trying to do? Have I even done it?* * Can you find my mistakes?
14. How would you compute this? for each continent max life exp - min life exp put result in a data frame
15. gapminder %>% group_by(continent) %>% summarize(max_minus_min = max(lifeExp) - min(lifeExp)) #> # A tibble: 5 x 2 #> continent max_minus_min #> <fct> <dbl> #> 1 Africa 52.8 #> 2 Americas 43.1 #> 3 Asia 53.8 #> 4 Europe 38.2 #> 5 Oceania 12.1 Here's how I would do it. Conclusion: there are many ways to write a for loop in R!
16. sidebar on %>%
17. child <- c("Reed", "Wesley", "Eli", "Toby") age <- c( 14, 12, 12, 1) s <- rep_len("", length(child)) for (i in seq_along(s)) { s[i] <- paste(child[i], "is", age[i], "years old") } s #> [1] "Reed is 14 years old" "Wesley is 12 years old" #> [3] "Eli is 12 years old" "Toby is 1 years old" New example: making strings
18. child <- c("Reed", "Wesley", "Eli", "Toby") age <- c( 14, 12, 12, 1) paste(child, "is", age, "years old") #> [1] "Reed is 14 years old" "Wesley is 12 years old" #> [3] "Eli is 12 years old" "Toby is 1 years old" glue::glue("{child} is {age} years old") #> Reed is 14 years old #> Wesley is 12 years old #> Eli is 12 years old #> Toby is 1 years old Here's how I would do it. Conclusion: maybe someone already wrote that for loop for you!
19. But what if you really do need to iterate?
20. https://purrr.tidyverse.org Part of the tidyverse A "core" package in the tidyverse meta-package install.packages("tidyverse") # <-- install purrr + much more install.packages("purrr") # <-- installs only purrr library(tidyverse) # <-- loads purrr + much more library(purrr) # <-- loads only purrr
21. purrr is an alternative to "apply" functions purrr::map() ≈ base::lapply()
22. library(purrr) library(repurrrsive) help(package = "repurrrsive")
23. Get comfortable with lists! atomic vectors are familar: logical, integer, double, character, etc a list = a generalized vector a list can hold almost anything
24. "working with lists"
25. How many elements are in got_chars? 
 Who is the 9th person listed in got_chars? What information is given for this person? 
 What is the difference between got_chars[9] and got_chars[[9]]? 
 Or ... do same for sw_people or the n-th person
26. List exploration str(x, list.len = ?, max.level = ?) x[i] x[[i]] str(x[[i]], ...) View(x), in RStudio
27. If list x is a train carrying objects: x[[5]] is the object in car 5 x[4:6] is a train of cars 4-6. -- Tweet by @RLangTip
28. from Subsetting chapter of 2nd ed Advanced R
29. from Subsetting chapter of 2nd ed Advanced R
30. x[[i]] x[i] x from http://r4ds.had.co.nz/vectors.html#lists-of-condiments
31. map(.x, .f, ...) purrr::
32. map(.x, .f, ...) purrr:: for every element of .x do .f
33. .x = minis
34. map(minis, antennate)
35. from Functionals chapter of 2nd ed Advanced R
36. map(.x, .f) purrr:: .x <- SOME VECTOR OR LIST out <- vector(mode = "list", length = length(.x)) for (i in seq_along(out)) { out[[i]] <- .f(.x[[i]]) } out
37. map(.x, .f) purrr:: .x <- SOME VECTOR OR LIST out <- vector(mode = "list", length = length(.x)) for (i in seq_along(out)) { out[[i]] <- .f(.x[[i]]) } out purrr::map() is a nice way to write a for loop.
38. How many aliases does each GoT character have?
39. map(got_chars, .f = ) map(sw_people, .f = ) or
40. Workflow: 1. Do it for one element. 2. Find the general recipe. 3. Drop into map() to do for all.
41. Step 1: Do it for one element daenerys <- got_chars[[9]] ## View(daenerys) daenerys[["aliases"]] #> [1] "Dany" "Daenerys Stormborn" #> [3] "The Unburnt" "Mother of Dragons" #> [5] "Mother" "Mhysa" #> [7] "The Silver Queen" "Silver Lady" #> [9] "Dragonmother" "The Dragon Queen" #> [11] "The Mad King's daughter" length(daenerys[["aliases"]]) #> [1] 11
42. Step 1: Do it for one element asha <- got_chars[[13]] ## View(asha) asha[["aliases"]] #> [1] "Esgred" "The Kraken's Daughter" length(asha[["aliases"]]) #> [1] 2
43. Step 2: Find the general recipe .x <- got_chars[[?]] length(.x[["aliases"]])
44. Step 2: Find the general recipe .x <- got_chars[[?]] length(.x[["aliases"]]) .x is a pronoun, like "it" means "the current element"
45. Step 3: Drop into map() to do for all map(got_chars, ~ length(.x[["aliases"]])) #> [[1]] #> [1] 4 #> #> [[2]] #> [1] 11 #> #> [[3]] #> [1] 1 #> #> [[4]] #> [1] 1 #> ...
46. Step 3: Drop into map() to do for all map(got_chars, ~ length(.x[["aliases"]])) #> [[1]] #> [1] 4 #> #> [[2]] #> [1] 11 #> #> [[3]] #> [1] 1 #> #> [[4]] #> [1] 1 #> ... formula method of specifying .f .x means "the current element" concise syntax for anonymous functions a.k.a. lambda functions
47. Challenge (pick one or more!) How many x does each (GoT or SW) character have? (x = titles, allegiances, vehicles, starships) map(got_chars, ~ length(.x[["aliases"]]))
48. map_int(got_chars, ~ length(.x[["aliases"]])) #> [1] 4 11 1 1 1 1 1 1 11 5 16 #> [12] 1 2 5 3 3 3 5 0 3 4 1 #> [25] 8 2 1 5 1 4 7 3 Oh, would you prefer an integer vector? map() map_lgl() map_int() map_dbl() map_chr() type-specific variants of map()
49. Challenge: Replace map() with type-specific map() # What's each character's name? map(got_chars, ~.x[["name"]]) map(sw_people, ~.x[["name"]]) # What color is each SW character's hair? map(sw_people, ~ .x[["hair_color"]]) # Is the GoT character alive? map(got_chars, ~ .x[["alive"]]) # Is the SW character female? map(sw_people, ~ .x[["gender"]] == "female") # How heavy is each SW character? map(sw_people, ~ .x[["mass"]])
50. Review
51. Lists can be awkward Lists are necessary Get to know your list
52. map(.x, .f, ...) purrr:: for every element of .x do .f
53. map(.x, .f) purrr:: map(got_chars, ~ length(.x[["aliases"]])) quick anonymous functions via formula
54. map_lgl(sw_people, ~ .x[["gender"]] == "female") map_int(got_chars, ~ length(.x[["aliases"]])) map_chr(got_chars, ~ .x[["name"]])
55. Onwards!
56. Notice: We extract by name a lot # What's each character's name? map(got_chars, ~.x[["name"]]) # What color is each SW character's hair? map(sw_people, ~ .x[["hair_color"]]) # Is the GoT character alive? map(got_chars, ~ .x[["alive"]]) # How heavy is each SW character? map(sw_people, ~ .x[["mass"]])
57. map_chr(got_chars, ~ .x[["name"]]) map_chr(got_chars, "name") Shortcut! .f accepts a name or position
58. .x = minis
59. map(minis, "pants")
60. Challenge: Explore a GoT or SW list and find a new element to look at Extract it across the whole list with name and position shortcuts for .f Use map_TYPE() to get an atomic vector as output map_??(got_??, ??) map_??( sw_??, ??)
61. Common problem I'm using map_TYPE() but some individual elements aren't of length 1. They are absent or have length > 1.
62. Solutions Missing elements? Specify a .default value. Elements of length > 1? You can't make an atomic vector.* Get happy with a list or list-column. Or pick one element, e.g., the first. * You can, if you are willing to flatten() or squash().
63. map(sw_vehicles, "pilots", .default = NA) #> [[1]] #> [1] NA #> #> ... #> #> [[19]] #> [1] "http://swapi.co/api/people/10/" "http://swapi.co/api/people/32/" #> #> [[20]] #> [1] "http://swapi.co/api/people/44/" #> #> ... #> #> [[37]] #> [1] "http://swapi.co/api/people/67/" #> #> [[38]] #> [1] NA #> #> [[39]] #> [1] NA
64. map_chr(sw_vehicles, list("pilots", 1), .default = NA) #> [1] NA NA #> [3] NA NA #> [5] "http://swapi.co/api/people/1/" NA #> [7] NA "http://swapi.co/api/people/13/" #> [9] NA NA #> [11] NA NA #> [13] "http://swapi.co/api/people/1/" NA #> [15] NA NA #> [17] NA NA #> [19] "http://swapi.co/api/people/10/" "http://swapi.co/api/people/44/" #> [21] "http://swapi.co/api/people/11/" "http://swapi.co/api/people/70/" #> [23] "http://swapi.co/api/people/11/" NA #> [25] NA "http://swapi.co/api/people/79/" #> [27] NA NA #> [29] NA NA #> [31] NA NA #> [33] NA NA #> [35] NA NA #> [37] "http://swapi.co/api/people/67/" NA #> [39] NA
65. Shortcut! .f accepts a name or position vector of names or positions or a list of names and positions map(got_chars, c(14, 1)) map(sw_vehicles, list("pilots", 1))
66. Names make life nicer! map_chr(got_chars, "name") #> [1] "Theon Greyjoy" "Tyrion Lannister" "Victarion Greyjoy" #> ... got_chars_named <- set_names(got_chars, map_chr(got_chars, "name")) got_chars_named %>% map_lgl("alive") #> Theon Greyjoy Tyrion Lannister Victarion Greyjoy #> TRUE TRUE TRUE #> ... Names propagate in purrr pipelines. Set them early and enjoy!
67. allegiances <- map(got_chars_named, "allegiances") tibble::enframe(allegiances, value = "allegiances") #> # A tibble: 30 x 2 #> name allegiances #> <chr> <list> #> 1 Theon Greyjoy <chr [1]> #> 2 Tyrion Lannister <chr [1]> #> 3 Victarion Greyjoy <chr [1]> #> 4 Will <NULL> #> 5 Areo Hotah <chr [1]> #> 6 Chett <NULL> #> 7 Cressen <NULL> #> 8 Arianne Martell <chr [1]> #> 9 Daenerys Targaryen <chr [1]> #> 10 Davos Seaworth <chr [2]> #> # ... with 20 more rows tibble::enframe() does this: named list → df w/ names & list-column
68. Review #2
69. Set list names for a happier life. There are many ways to specify .f. .default is useful for missing things. got_chars_named <- set_names(got_chars, map_chr(got_chars, "name")) map(got_chars, ~ length(.x[["aliases"]])) map_chr(got_chars, "name") map(sw_vehicles, list("pilots", 1)) map(sw_vehicles, "pilots", .default = NA) map_chr(sw_vehicles, list("pilots", 1), .default = NA)
70. Challenge: Create a named copy of a GoT or SW list with set_names(). Find an element with tricky presence/absence or length. Extract it many ways: - by name - by position - by list("name", pos) or c(pos, pos) - use .default for missing data - use map_TYPE() to coerce output to atomic vector
71. Challenge (pick one or more): Which SW film has the most characters? Which SW species has the most possible eye colors? Which GoT character has the most allegiances? Aliases? Titles? Which GoT character has been played by multiple actors?
72. Inspiration for your future purrr work
73. map(.x, .f, ...) books <- map(got_chars_named, "books") map_chr(books[1:2], paste, collapse = ", ") #> Theon Greyjoy #> "A Game of Thrones, A Storm of Swords, A Feast for Crows" #> Tyrion Lannister #> "A Feast for Crows, The World of Ice and Fire" map_chr(books[1:2], ~ paste(.x, collapse = ", ")) #> Theon Greyjoy #> "A Game of Thrones, A Storm of Swords, A Feast for Crows" #> Tyrion Lannister #> "A Feast for Crows, The World of Ice and Fire"
74. from Functionals chapter of 2nd ed Advanced R map(.x, .f, ...)
75. map(.x, .f, ...) books <- map(got_chars_named, "books") map_chr(books[1:2], paste, collapse = ", ") #> Theon Greyjoy #> "A Game of Thrones, A Storm of Swords, A Feast for Crows" #> Tyrion Lannister #> "A Feast for Crows, The World of Ice and Fire" map_chr(books[1:2], ~ paste(.x, collapse = ", ")) #> Theon Greyjoy #> "A Game of Thrones, A Storm of Swords, A Feast for Crows" #> Tyrion Lannister #> "A Feast for Crows, The World of Ice and Fire"
76. So, yes, there are many ways to specify .f. map(got_chars, ~ length(.x[["aliases"]])) map_chr(got_chars, "name") map_chr(books[1:2], paste, collapse = ", ") map(sw_vehicles, list("pilots", 1))
77. library(tidyverse) library(gapminder) countries <- c("Argentina", "Brazil", "Canada") gap_small <- gapminder %>% filter(country %in% countries, year > 1996) gap_small #> # A tibble: 9 x 6 #> country continent year lifeExp pop gdpPercap #> <fct> <fct> <int> <dbl> <int> <dbl> #> 1 Argentina Americas 1997 73.3 36203463 10967. #> 2 Argentina Americas 2002 74.3 38331121 8798. #> 3 Argentina Americas 2007 75.3 40301927 12779. #> 4 Brazil Americas 1997 69.4 168546719 7958. #> 5 Brazil Americas 2002 71.0 179914212 8131. #> 6 Brazil Americas 2007 72.4 190010647 9066. #> 7 Canada Americas 1997 78.6 30305843 28955. #> 8 Canada Americas 2002 79.8 31902268 33329. #> 9 Canada Americas 2007 80.7 33390141 36319. write_one <- function(x) { filename <- paste0(x, ".csv") dataset <- filter(gap_small, country == x) write_csv(dataset, filename) } walk(countries, write_one) list.files(pattern = "*.csv") #> [1] "Argentina.csv" "Brazil.csv" "Canada.csv" walk() is map() but returns no output
78. library(tidyverse) csv_files <- list.files(pattern = "*.csv") csv_files #> [1] "Argentina.csv" "Brazil.csv" "Canada.csv" map_dfr(csv_files, ~ read_csv(.x)) #> # A tibble: 9 x 6 #> country continent year lifeExp pop gdpPercap #> <fct> <fct> <int> <dbl> <int> <dbl> #> 1 Argentina Americas 1997 73.3 36203463 10967. #> 2 Argentina Americas 2002 74.3 38331121 8798. #> 3 Argentina Americas 2007 75.3 40301927 12779. #> 4 Brazil Americas 1997 69.4 168546719 7958. #> 5 Brazil Americas 2002 71.0 179914212 8131. #> 6 Brazil Americas 2007 72.4 190010647 9066. #> 7 Canada Americas 1997 78.6 30305843 28955. #> 8 Canada Americas 2002 79.8 31902268 33329. #> 9 Canada Americas 2007 80.7 33390141 36319. map_dfr() rowbinds a list of data frames
79. mapping over 2 or more things in parallel
80. .y = hair .x = minis
81. map2(minis, hair, enhair)
82. .y = weapons .x = minis
83. map2(minis, weapons, arm)
84. minis %>% map2(hair, enhair) %>% map2(weapons, arm)
85. from Functionals chapter of 2nd ed Advanced R
86. df <- tibble(pants, torso, head) embody <- function(pants, torso, head) insert(insert(pants, torso), head)
87. pmap(df, embody)
88. from Functionals chapter of 2nd ed Advanced R
89. map_dfr(minis, `[`, c("pants", "torso", "head")
90. For much more on this: rstd.io/row-work
91. from Functionals chapter of 2nd ed Advanced R You have the basis for exploring the world of purrr now!