Add warning when two different timezones are joined

[danielsjf]

This relates to this issue: #2643

The decision was taken to convert to UTC when two different posixct columns are joined. Indeed, the timezone is often only used for presentation, but some functions such as lubridate::year() work on the current timezone. Therefore such a decision could have an impact.

There is a warning if factors and characters are joined. It would be convenient to have a similar warning if the timezones don't match.

library(tidyverse)
#> Loading tidyverse: ggplot2
#> Loading tidyverse: tibble
#> Loading tidyverse: tidyr
#> Loading tidyverse: readr
#> Loading tidyverse: purrr
#> Loading tidyverse: dplyr
#> Conflicts with tidy packages ----------------------------------------------
#> filter(): dplyr, stats
#> lag():    dplyr, stats

tib_cet <- tibble(date = ISOdate(2016,1,1,0:2, tz = 'CET'), val_cet = runif(3))
tib_eet <- tibble(date = ISOdate(2016,1,1,0:2, tz = 'EET'), val_cet = runif(3))
tib_utc <- tibble(date = ISOdate(2016,1,1,0:2, tz = 'UTC'), val_utc = runif(3))

lubridate::year(tib_cet$date)
#> [1] 2016 2016 2016
lubridate::year(tib_eet$date)
#> [1] 2016 2016 2016
lubridate::year(tib_utc$date)
#> [1] 2016 2016 2016

tib_join <- left_join(tib_cet, tib_utc, by = 'date')
lubridate::year(tib_join$date) # UTC is the convention but I would assume that the convention is on the left table
#> [1] 2015 2016 2016

tib_join <- left_join(tib_cet, tib_eet, by = 'date')
lubridate::year(tib_join$date) # This is even stranger since the UTC assumption is used in neither of the two tables
#> [1] 2015 2016 2016

Such issues go unnoticed. Certainly since I would assume that the timezone of the left hand table would be dominant.

[romainfrancois]

Related to #2432

[hadley]

The timezone only affects printing so I think a warning would be too much. Regardless, I think we'll tackle systematically as part of #2432.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/