unnest errors when data.table's IDate class is combined with integers

[lpiep]

unnest errors when data.table's IDate class is combined with integers. unnest is capable of handling situations where nested tibbles contain dates and integers, but it fails when it attempts the same thing with data.table's IDate class.

library(tidyr)
library(data.table)

df_idates <- tibble(
  ids= c(1, 2),
  data = list(
    tibble(
      dates = c(as.IDate('2012-01-01'), as.IDate('2013-12-03'), as.IDate('2023-06-03')),
    ),
    tibble(
      numbers = 1:3
    )
  )
)

df_dates <- tibble(
  ids= c(1, 2),
  data = list(
    tibble(
      dates = c(as.Date('2012-01-01'), as.Date('2013-12-03'), as.Date('2023-06-03')),
    ),
    tibble(
      numbers = 1:3
    )
  )
)

unnest(df_idates, 'data')
#> Error in as.Date.numeric(e): 'origin' must be supplied
unnest(df_dates, 'data')
#> # A tibble: 6 × 3
#>     ids dates      numbers
#>   <dbl> <date>       <int>
#> 1     1 2012-01-01      NA
#> 2     1 2013-12-03      NA
#> 3     1 2023-06-03      NA
#> 4     2 NA               1
#> 5     2 NA               2
#> 6     2 NA               3

^{Created on 2023-11-01 by the reprex package (v2.0.1)}

I am using the latest GH version of tidyr.

[hadley]

Could you please rework your reproducible example to use the reprex package ? That makes it easier to see both the input and the output, formatted in such a way that I can easily re-run in a local session.

[lpiep]

@hadley done. thanks!

[hadley]

Hmmm, that works for me:

library(tidyr)
library(data.table)

df_idates <- tibble(
  ids= c(1, 2),
  data = list(
    tibble(
      dates = c(as.IDate('2012-01-01'), as.IDate('2013-12-03'), as.IDate('2023-06-03')),
    ),
    tibble(
      numbers = 1:3
    )
  )
)

df_dates <- tibble(
  ids= c(1, 2),
  data = list(
    tibble(
      dates = c(as.Date('2012-01-01'), as.Date('2013-12-03'), as.Date('2023-06-03')),
    ),
    tibble(
      numbers = 1:3
    )
  )
)

unnest(df_idates, 'data')
#> # A tibble: 6 × 3
#>     ids dates      numbers
#>   <dbl> <IDate>      <int>
#> 1     1 2012-01-01      NA
#> 2     1 2013-12-03      NA
#> 3     1 2023-06-03      NA
#> 4     2 1970-01-02       1
#> 5     2 1970-01-03       2
#> 6     2 1970-01-04       3
unnest(df_dates, 'data')
#> # A tibble: 6 × 3
#>     ids dates      numbers
#>   <dbl> <date>       <int>
#> 1     1 2012-01-01      NA
#> 2     1 2013-12-03      NA
#> 3     1 2023-06-03      NA
#> 4     2 NA               1
#> 5     2 NA               2
#> 6     2 NA               3

^{Created on 2023-11-01 with reprex v2.0.2}

[lpiep]

Hmm. Those 1970 dates in the IDate result shouldn't be there though, right?

[hadley]

Oh yeah, that is weird! Slightly more minimal reprex:

library(tidyr)
library(data.table)

df <- tibble(
  ids = c(1, 2),
  date = list(tibble(x = as.Date('2012-01-01')), tibble(y = 1)),
  idate = list(tibble(x = as.IDate('2012-01-01')), tibble(y = 1)),
)
unnest(df, c(date, idate), names_sep = "_")
#> # A tibble: 2 × 5
#>     ids date_x     date_y idate_x    idate_y
#>   <dbl> <date>      <dbl> <IDate>      <dbl>
#> 1     1 2012-01-01     NA 2012-01-01      NA
#> 2     2 NA              1 1970-01-02       1

unnest_wider(df, c(date, idate), names_sep = "_")
#> # A tibble: 2 × 5
#>     ids date_x     date_y idate_x    idate_y
#>   <dbl> <date>      <dbl> <IDate>      <dbl>
#> 1     1 2012-01-01     NA 2012-01-01      NA
#> 2     2 NA              1 NA               1

^{Created on 2023-11-01 with reprex v2.0.2}

Interesting that unnest_wider() doesn't have the same probelm.

[hadley]

Looks like something is going wrong deep in the bowels of chop().

[DavisVaughan]

Tracking in r-lib/vctrs#1892, it's a vctrs issue