pmap strips `Date`

[AkhilGNair]

As title - purrr::pmap seems to strip date class.

print_date = function(b) b

b = as.Date("2012-03-22")

purrr::pmap(list(b), print_date)
#> [[1]]
#> [1] 15421

Not sure if this is intended? Caught me out though :)

[david-jankoski]

there were a couple of similar discussions around map and dates around here. this workaround does the job for me.

print_date = function(b) as.Date(b, origin = "1970-01-01")
b = as.Date("2012-03-22")
purrr::pmap(list(b), print_date)

[[1]]
[1] "2012-03-22"

hth.

[AkhilGNair]

Thanks @david-jankoski - That's what I'm currently doing. Just wanted to make sure it was known, I admitted didn't look around to see if anything similar has been asked.

[mungojam]

I hit the same problem with pwalk

[huftis]

The bug is not specific to dates; pmap() removes all classes. Note that the plain map() function does not have this bug. Simple reprex:

library(purrr)
x = "test"
class(x) = "foo"
pmap(list(x), identity)
#> [[1]]
#> [1] "test"
map(list(x), identity)
#> [[1]]
#> [1] "test"
#> attr(,"class")
#> [1] "foo"

And as @mungojam mentioned, pwalk() also has this bug:

> pwalk(list(x), ~ print(class(.)))
# [1] "character" # should have been "foo" ...

[lionel-]

This is more of a known limitation rather than a bug, but we have plans to fix this.

[mungojam]

I found today that flatten does this too and removed my Date class as well as doing the intended un-listing. Perhaps that is intentional?

[lionel-]

It is not intentional we've just been too busy with tidy eval and other projects :(

[mungojam]

No worries, I'm enjoying all your other work :)

[hadley]

The root cause is this behaviour:

x <- as.Date("2012-03-22")
x[[1]]
#> [1] "2012-03-22"
list(x)[[c(1,1)]]
#> [1] 15421

This seems like a base R problem?

[yutannihilation]

This seems like a base R problem?

It seems so to me; [[, or do_subset2(), tries to dispatch the method based only on the first argument: https://github.com/wch/r-source/blob/58aec3ee94f2a0769974520f6cb2d99c49fb829f/src/main/subset.c#L894-L914

If it fails, the default implementation of subset2 will be used for all subsets, but, IIUC, it doesn't care about the classes or attributes: https://github.com/wch/r-source/blob/58aec3ee94f2a0769974520f6cb2d99c49fb829f/src/main/subset.c#L1029-L1056