map_* expects a vector of length 1

[rmflight]

This isn't really documented in the map_* that these functions expect a length 1 vector.

It is disconcerting, b/c the map_df* functions don't care how big the resulting objects are, they just get merged together.

Small pull request to update the documentation.

[rmflight]

For example, I really expect this to work:

l1 <- list(list(a = 1L),
           list(a = c(1L, 2L, 3L)))

purrr::map_int(l1, "a")
#> Error: Result 2 is not a length 1 atomic vector

purrr::map_chr(l1, function(x){class(x$a)})
#> [1] "integer" "integer"

[jennybc]

In terms of getting what you want, map() followed by a type-specific flatten_*() does it:

library(purrr)

l1 <- list(list(a = 1L),
           list(a = c(1L, 2L, 3L)))

map(l1, "a") %>% flatten_int()
#> [1] 1 1 2 3

[rmflight]

Fair enough, but the reason for the issue (and pull request) is that the documentation doesn't make this clear. The documentation seems to imply that the initial example should be possible. How can we change the documentation so that it is clearer the ??

…

On Wed, Jun 20, 2018, 4:29 PM Jennifer (Jenny) Bryan < ***@***.***> wrote: In terms of getting what you want, map() followed by a type-specific flatten_*() does it: library(purrr) l1 <- list(list(a = 1L), list(a = c(1L, 2L, 3L))) map(l1, "a") %>% flatten_int()#> [1] 1 1 2 3 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#518 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABcI-iDmphtAmyDyoK_jWFwUkp8CnLKFks5t-rCXgaJpZM4Uv5wT> .

[egnha]

@rmflight I hope I am not adding noise to the discussion, but what about the first line of the description of the map functions?:

The map functions transform their input by applying a function to each element and returning a vector the same length as the input.

Does this not imply that the function to be mapped by map_*() must return a scalar? (But you are right that it is not explicitly stipulated. If that is what's tripping up users, then I see your point.)

[jennybc]

I proposed some wording in the PR: #517 (comment)

[cderv]

Not about the wording but for reference about the use case needing flatten after map, there is an open PR #502 adding map_flat_* functions.

# devtools::install_github("tidyverse/purrr#502)
library(purrr)
l1 <- list(list(a = 1L),
           list(a = c(1L, 2L, 3L)))

map_flat_int(l1, "a")
#> [1] 1 1 2 3

Created on 2018-06-21 by the reprex package (v0.2.0).