Iteration with map

Published by onesixx on 17-06-1517-06-15

https://onesixx.com/13-iterating-with-purrr/

http://r4ds.had.co.nz/iteration.html#the-map-functions
https://www.rstudio.com/resources/videos/happy-r-users-purrr-tutorial/
https://github.com/rstudio/rstudio-conf/tree/master/2017/Happy_R_Users_Purrr-Charlotte_Wickham
http://purrr.tidyverse.org/reference/map.html
http://statkclee.github.io/parallel-r/ds-fp-purrr.html
http://r4ds.had.co.nz/iteration.html#the-map-functions

https://stackoverflow.com/questions/38403111/ways-to-add-multiple-columns-to-data-frame-using-plyr-dplyr-purrr
https://www.r-bloggers.com/rebuilding-map-example-with-apply-functions/

purrr에서 map()만 사용한다면 큰 의미가 없다.
https://jtr13.github.io/spring19/ss5593&fq2150.html

pacman::p_load("repurrrsive")
# sw_films 각 영화 7개의 list, 각 영화는 14개의 list로 이루어짐. 
sw_films %>% lapply(function(x){x[["director"]]})
sw_films %>% map(~ .x[["director"]])
sw_films %>% map(~ .[["director"]])

purrr의 장점

exlist <- list(1,2,3)
map(exlist, ~(.x+1))
# [[1]]
# [1] 2
# [[2]]
# [1] 3
# [[3]]
# [1] 4

일관된 문법 , 일관된 return type
(sapply() & lapply()는 return이 상황에 따라 다르다.)
Pipe사용가능
함수의 인수(mean()함수의 trim이나 na.rm )을 바로 붙여 사용할수 있다.

dt <- data.table(a=rnorm(10), b=sample(100,10), c=rchisq(10,2), d=rbinom(10,100,0.5))
map_dbl(dt, mean, trim=0.5)
#          a           b           c           d 
# 0.06995195 84.00000000  2.15738130 51.00000000

map functions 은 name을 유지한다.

z <- list(x=1:3, y=4:5)
map_int(z, length) 
# x y 
# 3 2 
lapply(z, length) %>% as.integer()
# [1] 3 2

간결한 shortcuts .
. 은 대명사 (for루프에서 i가 현재 index를 대신하듯이, . 은 현재 list요소를 대신한다)
function(x){ x+1 } 은 ~ .+1
즉 X %>% lapply(function(x) x+1) 은 X %>% map(~{.+1})

lapply(list, function(x) x[[2]])
map(list, function(x) x[[2]])
map(list, 2)

mtcars %>% split(.$cyl) %>%    
#map(function(df) lm(mpg ~ wt, data = df)) 
map(~ lm(mpg ~ wt, data=.)) %>% 
map(summary) %>% 
#extracting named components is a common operation models 
#map_dbl(~.$r.squared )      
map_dbl(  "r.squared")

가져올 요소의 위치(Position)을 Integer를 사용하여 대신할수 있다.

x <- list(list(1, 2, 3),
          list(4, 5, 6),
          list(7, 8, 9)) 
x %>% map_dbl(2) 
#> [1] 2 5 8

apply와 비교

sapply(iris[ ,-5], mean, na.rm=T)
apply(iris[ ,-5], 2, mean, na.rm=T)
iris[ ,-5] %>% apply(2,mean,na.rm=T)

iris[ ,-5] %>% map_dbl(mean,na.rm=T)

반복문

1:nrow(iris) %>% 
  map(function(rowNum){ 
        dataRow <- iris[rowNum,]
      })\t

Map() 의 종류

map_returnType

map function	makes a
map()	list	lapply(), sapply()
map_lgl()	logical vector
map_int()	integer vector
map_dbl()	double vector
map_chr()	character vector

각 map 함수는 vector를 input으로 받아서, 각 vector에 함수를 적용하고,
같은이름과 length의 새로운 vector를 output으로 return한다.
(map 함수의 접미사는 vector 유형을 결정한다.)

for루프의 속도는 많이 향상되었기 때문에, Map 함수는 속도보다는 코드의 간결/정확함을 위해 사용한다고 생각하면 된다.

map_if(), map_at()

numbers  <-  list(11, 12, 13, 14)
is_even  <-  function(x){ !as.logical(x %% 2) } # number가 짝수이면, TRUE

map_if(numbers, is_even, sqrt)
# [[1]]
# [1] 11
# [[2]]
# [1] 3.464102
# [[3]]
# [1] 13
# [[4]]
# [1] 3.741657

map_at(numbers, c(1,3), sqrt)
# [[1]]
# [1] 3.316625
# [[2]]
# [1] 12
# [[3]]
# [1] 3.605551
# [[4]]
# [1] 14

argument가 여러개일때 Mapping

map2(.x, .y, .f, ...)

set.seed(666)
options(digits=2)

mu    <- list(5,10,-3)
sigma <- list(1, 5,10)
n     <- list(1, 3, 5)

# map(mu, rnorm, n=5)
# map(seq_along(mu), ~rnorm(n=5, mean=mu[[.x]], sd=sigma[[.x]]))
map2(mu, sigma, rnorm, n=5)

pmap(.l, .f, …)

arg1 <- list(n, mu, sigma)
args2 <- list(mean = mu, sd = sigma, n = n)

pmap(arg1, rnorm)
pmap(arg2, rnorm)

example.

x <- list(1, 10, 100)
y <- list(1,  2,   3)
z <- list(5, 50, 500)
l <- list(a=x, b=y, c=z)

map2(x, y, `+`)
map2(x, y, ~{.x +.y})

pmap(list(x, y, z), sum)
# Matching arguments by position
pmap(list(x, y, z), function(a, b, c){ a/(b+c)})
# Matching arguments by name
pmap(l, function(a, b, c){ a/(b+c) })

# Split into pieces, fit model to each piece, then predict
by_cyl <- mtcars %>% split(.$cyl)
mods <- by_cyl %>% map(~ lm(mpg ~ wt, data = .))
map2(mods, by_cyl, predict)

# Vectorizing a function over multiple arguments
df <- data.frame(
  x = c("apple", "banana", "cherry"),
  pattern = c("p", "n", "h"),
  replacement = c("x", "f", "q"),
  stringsAsFactors = FALSE
)
pmap(df, gsub)
pmap_chr(df, gsub)

function이 여러개일때 Mapping

invoke_map() 함수

f <- c("runif", "rnorm", "rpois") 
param <- list(
 list(min = -1, max = 1), list(sd = 5), list(lambda = 10)
)
invoke_map(f, param, n = 5) %>% str()

sim <- tribble(
         ~f, ~params,
    "runif", list(min = -1, max = 1), 
    "rnorm", list(sd = 5),
    "rpois", list(lambda = 10)
)
sim %>% mutate(sim = invoke_map(f, params, n = 10))

SOLVE ITERATION PROBLEMS
FOR EACH __DO __
You are already solving them:
copy & paste, for loops, (l/s)apply()
I'll show you an alternative purrr::map() & friends

library(purrr)
library(tidyverse)

load("~/Dropbox/Rhome/purr/data/swapi.rda")

map( .x, .f, ...)
for each element of .x do .f

Other types of output : map_lgl( .x, .f, ...)

map_lgl() logical vector
map_int() integer vector
map_dbl() double vector
map_chr() character vecto

Other ways of specifying .f : map( .x, length, ...)
Other iteration functions : map2( .x, .y, .f, ...)

map() 항상 list를 return한다.

identical

map(.x, .f = some_function, ...)
map(.x, ~ some_function(.x, ...))

map(people, ~ .x[["hair_color"]])

map(people, .f = "hair_color")

map(people, "hair_color")

Dealing with failure

http://r4ds.had.co.nz/iteration.html#dealing-with-failure

여러 operation을 반복하기 위해 map 함수를 사용하다보면 그만큼 operation이 오류를 일으킬 확률이 높아진다.
하나때문에 모든 operation이 안돌아가고, 그게 어떤건지 찾는건 좀 짜증나는 일이다.

safely(.f, otherwise = NULL, quiet = TRUE)

quietly(.f)

possibly(.f, otherwise, quiet = TRUE)

safely()

뭔가 잘못되더라도, 함수가 실행되는 것을 멈추지 한는다.
Base R의 try와 비슷한 safely 는 함수를 감싸서 결과를 두가지로 알려준다.

result is the original result. If there was an error, this will be NULL.
error is an error object. If the operation was successful, this will be NULL.

log(10)   # 2.302585
safe_log <- safely(log)

정상적인 경우

str(safe_log(10))

c(10) %>% map(safely(log)) %>% str()

List of 2
 $ result: num 2.3
 $ error : NULL

에러상황 : 문자에 log를 취하려고 함

str(safe_log("a"))

c("a") %>% map(safely(log)) %>% str()

List of 2
 $ result: NULL
 $ error :List of 2
  ..$ message: chr "non-numeric argument to mathematical function"
  ..$ call   : language .f(...)
  ..- attr(*, "class")= chr [1:3] "simpleError" "error" "condition"

ex2> https://www.r-bloggers.com/lesser-known-purrr-tricks/

safely() 는 아래 possibly() 와 비슷하지만 list의 list를 return한다. (result리스트와 error 메세지)

numbers_with_error <- list(1, 2, 3, "spam", 4)

safe_sqrt <- safely(sqrt, otherwise = NA_real_)
map(numbers_with_error, safe_sqrt)

possible_sqrt <- possibly(sqrt, otherwise = NA_real_)
map(numbers_with_error, possible_sqrt)

possibly()

에러상황: 정의된 값으로 대체한다.

x <- list(1, 10, "a")
x %>% map_dbl(possibly(log, NA_real_))

[1] 0.00 2.30       NA

quietly()

safely()와 비슷하지만, 에러를 잡는대신 printed output, messages, warnings을 잡는다.

x <- list(1, -1)
x %>% map(quietly(log)) %>% str()

Iteration with map

https://onesixx.com/13-iterating-with-purrr/

purrr의 장점

apply와 비교

Map() 의 종류

map_returnType

map_if(), map_at()

argument가 여러개일때 Mapping

map2(.x, .y, .f, ...)

pmap(.l, .f, …)

example.

function이 여러개일때 Mapping

identical

Dealing with failure

http://r4ds.had.co.nz/iteration.html#dealing-with-failure

safely()

possibly()

quietly()

onesixx

factor

4. Model :: intro

Program :: Iteration with loop

Iteration with map

https://onesixx.com/13-iterating-with-purrr/

purrr의 장점

apply와 비교

Map() 의 종류

map_returnType

map_if(), map_at()

argument가 여러개일때 Mapping

map2(.x, .y, .f, ...)

pmap(.l, .f, …)

example.

function이 여러개일때 Mapping

identical

Dealing with failure

http://r4ds.had.co.nz/iteration.html#dealing-with-failure

safely()

possibly()

quietly()

onesixx

Related Posts

factor

4. Model :: intro

Program :: Iteration with loop