Purrr safely creating lists of lists

Question

I've used safely to catch an error which occurs in my code when I'm purring. However, the result from safely is much more complex than I anticipated.

First we create the necessary functions and example data.

#base functions.
SI_tall <- function(topheight,  age, si ){
  paramasi <- 25
  parambeta <- 7395.6
  paramb2 <- -1.7829
  refAge <- 100

  d <- parambeta*(paramasi^paramb2)

  r <- (((topheight-d)^2)+(4*parambeta*topheight*(age^paramb2)))^0.5

  ## height at reference age
  h2 <- (topheight+d+r)/ (2+(4*parambeta*(refAge^paramb2)) / (topheight-d+r))

  return(abs(h2 - si))
}


new.topheight <- function(my.si, my.age){
  optim(par = list(topheight = 10), ## this topheight is just an initial value
                       method = 'L-BFGS-B', fn = SI_tall, si = my.si, age = my.age, lower= 0, upper=100)$par
}

#Creating the function which will display errors.
safe_new.topheight <- safely(new.topheight)

#Creating data
my.age <- seq(0,100, by=0.2)
my.si <- c(15)

si.crossing <- tidyr::crossing(my.si, my.age) %>% data.frame()


#Creating the column to be unnested.
si.crossing2<- si.crossing %>% 
 mutate(height=map2(my.si,my.age, safe_new.topheight))

However, the result becomes to complex for me to unnest - I don't even know how far down the nesting of lists in the ´height´ column goes. Here's a dput of the first 5 rows in my data:

structure(list(my.si = c(15, 15, 15, 15, 15), my.age = c(0, 0.2, 
0.4, 0.6, 0.8), height = list(list(result = NULL, error = structure(list(
    message = "L-BFGS-B needs finite values of 'fn'", call = optim(par = list(topheight = 10), 
        method = "L-BFGS-B", fn = SI_tall, si = my.si, age = my.age, 
        lower = 0, upper = 100)), class = c("simpleError", "error", 
"condition"))), list(result = c(topheight = 0.000693170450744849), 
    error = NULL), list(result = c(topheight = 0.00205917508142004), 
    error = NULL), list(result = c(topheight = 0.00390099534708239), 
    error = NULL), list(result = c(topheight = 0.00639475141226834), 
    error = NULL))), row.names = c(NA, 5L), class = "data.frame")

Is there any way to flatten this down to the columns:

my.si, my.age, topheight, error.

Many thanks!

Answer 1

It might be helpful to distinguish between safely and possibly . safely is great to see what error occurred (and where). It is best used with transpose not inside of a tibble . possibly is used to run a map function even if it throughs an error. It lets you choose an alternative value otherwise if an error is thrown.

# use `transpose` on the height column to turn the list inside out
# which results in two lists `result` and `error`
# first lets have a look at the structure
si.crossing2$height %>% 
  transpose %>% 
  str 

# then `pluck` the `error` list and remove all elements which are NULL 
# with `compact` - here you can see the error that occurred
si.crossing2$height %>% 
  transpose %>% 
  pluck("error") %>% 
  compact

# `safely` is a great function to see what went wrong
# but its not very useful inside a tibbles list-column
# what you actually want to use is `possibly`

possib_new.topheight <- possibly(new.topheight, otherwise = NA)

# this will not tell you what went wrong, but instead yield `NA`
# when an error is thrown - important to use `otherwise = NA`, 
# the default is NULL, which makes the output list shorter and
# won't fit to your tibble
si.crossing3 <- si.crossing %>% 
  mutate(height = map2_dbl(my.si,my.age, possib_new.topheight))

si.crossing3

#> # A tibble: 501 x 3
#>    my.si my.age    height
#>    <dbl>  <dbl>     <dbl>
#>  1    15    0   NA       
#>  2    15    0.2  0.000693
#>  3    15    0.4  0.00206 
#>  4    15    0.6  0.00390 
#>  5    15    0.8  0.00639 
#>  6    15    1    0.00947 
#>  7    15    1.2  0.0131  
#>  8    15    1.4  0.0172  
#>  9    15    1.6  0.0218  
#> 10    15    1.8  0.0269  
#> # … with 491 more rows