Recent searches
Search options
A student in my intro #rstats class taught me something new today (by way of a cryptic "bug").
Suppose you're asked why this {purrr} code that should return the mean of each list element is not working as expected.
```
map(list(x=1:3, y=4:6), mean)
#> $x
#> [1] 1
#>
#> $y
#> [1] 4
```
What do you think is the simplest explanation for this behavior (in terms of the mistake that the student could've made)? It's not so obvious - there are multiple R "quirks" cascading!
Answer is in the images!
![library(purrr)
# !!THE MISTAKE!!
# - They calculated the mean of something in a prior problem
# and over-wrote the variable `mean`
set.seed(123)
mean <- mean(sample(2, 10, replace = TRUE))
mean
#> [1] 1.4
# Normally, this doesn't interfere with `mean`'s behavior as a function
# - See Section 6.4.2 of the "Advanced R" book
mean(1:5)
#> [1] 3
# In fact, the `lapply()` equivalent works fine. It only errors if you
# supply the value stored in the `mean` variable directly
lapply(list(x = 1:3, y = 4:6), mean)
#> $x
#> [1] 2
#>
#> $y
#> [1] 5
lapply(list(x = 1:3, y = 4:6), 1.4)
#> Error in match.fun(FUN): '1.4' is not a function, character or symbol
# But in `map()`, non-function values for the `.f` argument are allowed and
# used to "pluck" elements. This is convenient for wrangling deeply nested data.
fruits <- list(apple = list(color = "red"), grape = list(color = "purple"))
map(fruits, "color")
#> $apple
#> [1] "red"
#>
#> $grape
#> [1] "purple"](https://cdn.fosstodon.org/media_attachments/files/111/026/129/704/294/777/original/f935bdd440845362.png "library(purrr)
# !!THE MISTAKE!!
# - They calculated the mean of something in a prior problem
# and over-wrote the variable `mean`
set.seed(123)
mean <- mean(sample(2, 10, replace = TRUE))
mean
#> [1] 1.4
# Normally, this doesn't interfere with `mean`'s behavior as a function
# - See Section 6.4.2 of the \"Advanced R\" book
mean(1:5)
#> [1] 3
# In fact, the `lapply()` equivalent works fine. It only errors if you
# supply the value stored in the `mean` variable directly
lapply(list(x = 1:3, y = 4:6), mean)
#> $x
#> [1] 2
#>
#> $y
#> [1] 5
lapply(list(x = 1:3, y = 4:6), 1.4)
#> Error in match.fun(FUN): '1.4' is not a function, character or symbol
# But in `map()`, non-function values for the `.f` argument are allowed and
# used to \"pluck\" elements. This is convenient for wrangling deeply nested data.
fruits <- list(apple = list(color = \"red\"), grape = list(color = \"purple\"))
map(fruits, \"color\")
#> $apple
#> [1] \"red\"
#>
#> $grape
#> [1] \"purple\"")
![# The plucking is implemented in the `as_mapper()` generic. Interestingly,
# numeric (not strictly integer!) values are used to *index* into list elements
purrr:::as_mapper.numeric
# And we all know about the behavior of double => integer coercion...
as.integer(1.4)
#> [1] 1
# So that's why the following map code plucks out the first value
# of each element of the list!! Case closed!
map(list(x = 1:3, y = 4:6), mean)
#> $x
#> [1] 1
#>
#> $y
#> [1] 4
# --- Coda ---
# IMO: this should be a bug! Tidyverse is generally strict about using integers
# for subsetting in other places:
tibble::tibble(x = 1:3)[1]
#> # A tibble: 3 × 1
#> x
#> <int>
#> 1 1
#> 2 2
#> 3 3
tibble::tibble(x = 1:3)[1.3]
#> Error in `tibble::tibble(x = 1:3)[1.3]`:
#> ! Can't subset columns with `1.3`.
#> ✖ Can't convert from `j` <double> to <integer> due to loss of precision.](https://cdn.fosstodon.org/media_attachments/files/111/026/134/628/515/745/original/bf2809467a8850e9.png "# The plucking is implemented in the `as_mapper()` generic. Interestingly,
# numeric (not strictly integer!) values are used to *index* into list elements
purrr:::as_mapper.numeric
# And we all know about the behavior of double => integer coercion...
as.integer(1.4)
#> [1] 1
# So that's why the following map code plucks out the first value
# of each element of the list!! Case closed!
map(list(x = 1:3, y = 4:6), mean)
#> $x
#> [1] 1
#>
#> $y
#> [1] 4
# --- Coda ---
# IMO: this should be a bug! Tidyverse is generally strict about using integers
# for subsetting in other places:
tibble::tibble(x = 1:3)[1]
#> # A tibble: 3 × 1
#> x
#> <int>
#> 1 1
#> 2 2
#> 3 3
tibble::tibble(x = 1:3)[1.3]
#> Error in `tibble::tibble(x = 1:3)[1.3]`:
#> ! Can't subset columns with `1.3`.
#> ✖ Can't convert from `j` <double> to <integer> due to loss of precision.")