pmap discards attributes

[kendonB]

Is this a feature or a bug?

seq(as.Date("1990-01-01"),
    as.Date("1990-01-02"), by = "days") %>% 
  list() %>% 
  purrr::pmap(function(x){
    print(class(x))
  })
# [1] "numeric"
# [1] "numeric"
# [[1]]
# [1] "numeric"
# 
# [[2]]
# [1] "numeric"

purrr::map yields a date object.

[picasa]

Apparently, it's the same for POSIXct class.

library(lubridate)
library(dplyr)
library(purrr)

# data frame containing model inputs
inputs <- data_frame(a=1, t0=ymd("2016-10-12"))

# simplified model function
model <- function(t0, ...) {return(data_frame(time=format(t0, format="%d/%m")))}

# direct call to model fonction : time is the expected character string
model(t0=ymd("2016-10-12"))$time

# wished call to model function : time is a numeric value encoded in character string
inputs %>% pmap_df(., model) %>% .$time

[lionel-]

It's a known non-feature that we will resolve with the vctrs package (we need C-level methods for joining vectors).

[shrektan]

bite by this issue as well...

[hadley]

This is a complicated issue and the root cause is not within purrr. We are aware of the problem and working to fix it in the long-term.

[Ax3man]

Factors are also coerced to numeric:

pmap(list(1:5, factor(letters)[1:5]), paste)

[Deleetdk]

Issue still here as of purrr 0.2.4, at least for factors being converted to integers instead of the more obvious character alternative.

[yutannihilation]

@Deleetdk FYI: the further discussion continues here.
#358 (comment)