Reinstate support for filtering and arranging lubridate Periods and Intervals

[richierocks]

Support for filter()ing and arrange()ing columns of class Period and Interval was recently dropped from dplyr. The ability to work with these types is useful, so I'd like the feature reinstating.

(By 'useful', I mean in particular that there are DataCamp courses depending upon this feature. For example, Working with Dates and Times in R.)

---

To reproduce:

library(dplyr)                        
library(lubridate)                    
x <- data_frame(                      
  start = as.Date("2000-01-01") + 0:19, 
  finish = as.Date("2001-01-01") + 19:0,
  interval = start %--% finish          
)                                     
x %>%                                 
  arrange(interval)                     
#> Error in arrange_impl(.data, dots): Column `interval` classes Period and Interval from lubridate are currently not supported.
                                      
x %>%                                 
  filter(interval > 0)                  
#> Error in filter_impl(.data, quo): Column `interval` classes Period and Interval from lubridate are currently not supported.

[krlmlr]

Are you sure these features worked as expected? How is comparison defined between intervals?

We should definitely introduce proper support for these and other data types, #2432 is the canonical issue.

[richierocks]

@krlmlr Comparison for intervals was on the length of the interval (in seconds) – I think this is reasonable. You can use unclass(lubridate::int_length(x)).

[romainfrancois]

There are a few things we need to figure out before we can make these things work again. Some code will need to be factored out of dplyr or reimplemented so that 📦 like lubridate can implement a method that does the right thing without necessarily depend on dplyr.

This is probably for the 0.9.0 horizon at least

[hadley]

filter() should work in 0.8.0, arrange() will need to wait for 0.9 & vctrs support

[romainfrancois]

They both work for intervals, but the comparison is only done through the @.Data slot I guess.

library(dplyr, warn.conflicts = FALSE)                        
library(lubridate, warn.conflicts = FALSE)                    
x <- data_frame(                      
  start = as.Date("2000-01-01") + 0:3, 
  finish = as.Date("2001-01-01") + 3:0,
  interval = start %--% finish          
) 
x$interval
#> [1] 2000-01-01 UTC--2001-01-04 UTC 2000-01-02 UTC--2001-01-03 UTC
#> [3] 2000-01-03 UTC--2001-01-02 UTC 2000-01-04 UTC--2001-01-01 UTC
str(x$interval)
#> Formal class 'Interval' [package "lubridate"] with 3 slots
#>   ..@ .Data: num [1:4] 31881600 31708800 31536000 31363200
#>   ..@ start: POSIXct[1:4], format: "2000-01-01" ...
#>   ..@ tzone: chr "UTC"

x %>%                                 
  arrange(interval)                     
#> # A tibble: 4 x 3
#>   start      finish     interval                      
#>   <date>     <date>     <S4: Interval>                
#> 1 2000-01-04 2001-01-01 2000-01-04 UTC--2001-01-01 UTC
#> 2 2000-01-03 2001-01-02 2000-01-03 UTC--2001-01-02 UTC
#> 3 2000-01-02 2001-01-03 2000-01-02 UTC--2001-01-03 UTC
#> 4 2000-01-01 2001-01-04 2000-01-01 UTC--2001-01-04 UTC

x %>%                                 
  arrange(interval@.Data)                     
#> # A tibble: 4 x 3
#>   start      finish     interval                      
#>   <date>     <date>     <S4: Interval>                
#> 1 2000-01-04 2001-01-01 2000-01-04 UTC--2001-01-01 UTC
#> 2 2000-01-03 2001-01-02 2000-01-03 UTC--2001-01-02 UTC
#> 3 2000-01-02 2001-01-03 2000-01-02 UTC--2001-01-03 UTC
#> 4 2000-01-01 2001-01-04 2000-01-01 UTC--2001-01-04 UTC

x %>%                                 
  filter(interval > 0)  
#> # A tibble: 4 x 3
#>   start      finish     interval                      
#>   <date>     <date>     <S4: Interval>                
#> 1 2000-01-01 2001-01-04 2000-01-01 UTC--2001-01-04 UTC
#> 2 2000-01-02 2001-01-03 2000-01-02 UTC--2001-01-03 UTC
#> 3 2000-01-03 2001-01-02 2000-01-03 UTC--2001-01-02 UTC
#> 4 2000-01-04 2001-01-01 2000-01-04 UTC--2001-01-01 UTC

x %>%                                 
  filter(interval@.Data > 0)  
#> # A tibble: 4 x 3
#>   start      finish     interval                      
#>   <date>     <date>     <S4: Interval>                
#> 1 2000-01-01 2001-01-04 2000-01-01 UTC--2001-01-04 UTC
#> 2 2000-01-02 2001-01-03 2000-01-02 UTC--2001-01-03 UTC
#> 3 2000-01-03 2001-01-02 2000-01-03 UTC--2001-01-02 UTC
#> 4 2000-01-04 2001-01-01 2000-01-04 UTC--2001-01-01 UTC

^{Created on 2018-12-14 by the reprex package (v0.2.1.9000)}

[hadley]

Duplicate 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/