Is there a way to automatically propagate NULL in purrr::map?

Question

I have some data, some of which is replicated, some not. I can only fit my model for the replicated data.

library(tidyverse)
d <- tribble(
  ~env, ~val,
  "A",  1,
  "A",  2,
  "B",  3
)

I am using tidyr::nest() and purrr::map() functions to fit my model. However, in every function I use for map() I had to cater for the special case that a particular set of data is not modellable, which i archieved via calls in the style of

map(col, function(elem){ if(!is.null(elem)) DO_STUFF(elem) else NULL})

After a while, I managed to extract this behaviour to a purrr-style adverb function which takes another function and wraps it such that this behaviour for NULL elements is automatic:

maybe <- function(fun){
  function(val,...){ if(!is.null(val)) fun(val, ...) else NULL}
}

However, this left me wondering: Am I duplicating behaviour which is already archievable using tidyverse functions?

Bonus question: Is there a word in functional programming for a function like maybe?

This is an example to test my adverb:

Simple Model: A mean for data in environment A and no model for data in environment B (since the data is unreplicated:)

modelFuns <- list(A = mean, B = NULL)

Group data by environment and determine the appropriate model for each group

d <- d %>% group_by(env) %>% nest(.key = "data")
d %<>% mutate(model = modelFuns[env])
d

## # A tibble: 2 x 3
##   env   data             model 
##   <chr> <list>           <list>
## 1 A     <tibble [2 × 1]> <fn>  
## 2 B     <tibble [1 × 1]> <NULL>

Perform modelling:

d %<>% mutate(out = pmap(list(model, data), maybe(function(m,d) m(d$val))))
d

## # A tibble: 2 x 4
##   env   data             model  out      
##   <chr> <list>           <list> <list>   
## 1 A     <tibble [2 × 1]> <fn>   <dbl [1]>
## 2 B     <tibble [1 × 1]> <NULL> <NULL>

Which is equivalent to the following code which does not use my maybe adverb:

d %<>% mutate(out = pmap(list(model, data), function(m,d){if(!is.null(m)) m(d$val) else NULL}))
d

## # A tibble: 2 x 4
##   env   data             model  out      
##   <chr> <list>           <list> <list>   
## 1 A     <tibble [2 × 1]> <fn>   <dbl [1]>
## 2 B     <tibble [1 × 1]> <NULL> <NULL>

The fact that there might be a value or there might be NULL propagates to everything I want to do with the modelling results downstream, which is why the adverb maybe is useful. Does something like this already exist in the tidyverse?

isModelNice <- function(val) val > 0
d %<>% mutate(nice = map(out, maybe(isModelNice)))
d

## # A tibble: 2 x 5
##   env   data             model  out       nice     
##   <chr> <list>           <list> <list>    <list>   
## 1 A     <tibble [2 × 1]> <fn>   <dbl [1]> <lgl [1]>
## 2 B     <tibble [1 × 1]> <NULL> <NULL>    <NULL>

CJ Yetman · Accepted Answer · 2018-04-27 09:17:37Z

3

could you use purrr::possibly()?

library(tidyverse)

d <- tribble(
  ~env, ~val,
  "A",  1,
  "A",  2,
  "B",  3
)

modelFuns <- list(A = mean, B = NULL)

d %>% group_by(env) %>% 
  nest(.key = "data") %>% 
  mutate(model = modelFuns[env]) %>% 
  mutate(out = pmap(list(model, data), possibly(function(m,d) m(d$val), NULL)))

answered Apr 27, 2018 at 9:17

CJ Yetman

8,9382 gold badges31 silver badges63 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

akraf Over a year ago

OK, this works if my model function throws an error if I give a NULL value. Not exactly the behaviour I looked for (what if my model function returns numeric(0) or whatever given a NULL argument), but might be working for this particular case.

Collectives™ on Stack Overflow

Is there a way to automatically propagate NULL in purrr::map?

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related