[ARROW-14200] [R] strftime on a date should not use or be confused by timezones

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 6.0.0
• Component/s: R
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/29784
• Assignee: Jonathan Keane
• Reporter: Jonathan Keane
• Created: 01/Oct/21 19:42
• Updated: 11/Jan/23 08:38
• Resolved: 05/Oct/21 22:28

Description

When the input to strftime is a date, timezones shouldn't be necessary or assumed.

What I think is going on below is the date 1992-01-01 is being interpreted as 1992-01-01 00:00:00 in UTC, and then when strftime() is being called it's displaying that timestamp as 1991-12-31 ... (since my system is set to an after UTC timezone), and then taking the year out of it. If I specify tz = "utc" in the strftime(), I get the expected result (though that shouldn't be necessary).

Run in the US central timezone:

library(arrow, warn.conflicts = FALSE)
library(dplyr, warn.conflicts = FALSE)
library(lubridate, warn.conflicts = FALSE)
Table$create(
  data.frame(
    x = as.Date("1992-01-01")
  )
) %>% 
  mutate(
    as_int_strftime = as.integer(strftime(x, "%Y")),
    strftime = strftime(x, "%Y"),
    as_int_strftime_utc = as.integer(strftime(x, "%Y", tz = "UTC")),
    strftime_utc = strftime(x, "%Y", tz = "UTC"),
    year = year(x)
  ) %>%
  collect()
#>            x as_int_strftime strftime as_int_strftime_utc strftime_utc year
#> 1 1992-01-01            1991     1991                1992         1992 1992

Issue Links

• links to: GitHub Pull Request #11301