Use strict behavior for aweek objects

[zkamvar]

This PR will fix #20 and fix #8

The aweek package is going through a bit of a change to make things more
consistent. The list of changes can be found in the NEWS.md. Please test out
this code and let me know if you encounter any problems. I plan to release
this the second week of June.

The breaking changes in this release are that it’s no longer possible to
combine aweek objects that have different week_start attributes and factors
no longer coerce factors (but will be converted to characters upon
combination.

In the previous release, I enabled global week_start parameters that allowed
the user to set the week_start variable globally.

Please test this out with:

# install.packages("remotes")
remotes::install_github("reconhub/aweek#21")

Below are some highlights:

You can now create aweek objects directly from data frames of week numbers!

library("aweek")
library("dplyr")
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
set_week_start("Sunday") # Set the global week_start (defaults to Monday)

You can use numeric week data to create weeks or dates with get_aweek()
and get_date(). By default, they will give you the first week of the
current year:

get_aweek()
#> <aweek start: Sunday>
#> [1] "2019-W01-1"
get_date()
#> [1] "2018-12-30"

They are vectorised so you can use a data frame with them:

# create a table with all days in the week
d    <- as.Date("2019-03-03") + 0:6
res  <- lapply(weekdays(d), function(i) date2week(d, week_start = i))
resn <- lapply(weekdays(d), function(i) date2week(d, week_start = i, numeric = TRUE))
datf <- data.frame(wday = rep(weekdays(d), each = 7), 
                   week = unlist(res), # note: unlist converts to character
                   week_number = unlist(resn),
                   year  = 2019,
                   stringsAsFactors = FALSE)
datf$day <- substring(datf$week, 10, 11)
datf <- as_tibble(datf)
datf
#> # A tibble: 49 x 5
#>    wday   week       week_number  year day  
#>    <chr>  <chr>            <dbl> <dbl> <chr>
#>  1 Sunday 2019-W10-1          10  2019 1    
#>  2 Sunday 2019-W10-2          10  2019 2    
#>  3 Sunday 2019-W10-3          10  2019 3    
#>  4 Sunday 2019-W10-4          10  2019 4    
#>  5 Sunday 2019-W10-5          10  2019 5    
#>  6 Sunday 2019-W10-6          10  2019 6    
#>  7 Sunday 2019-W10-7          10  2019 7    
#>  8 Monday 2019-W09-7           9  2019 7    
#>  9 Monday 2019-W10-1          10  2019 1    
#> 10 Monday 2019-W10-2          10  2019 2    
#> # … with 39 more rows

Here you can convert week definitions with different starts to the correct
dates (the week of 2019-03-03).

datf %>%
  mutate(aweek = get_aweek(week = week_number, 
                           year = year, 
                           day = day, 
                           start = wday)) %>%
  mutate(the_date = get_date(week = week_number, 
                             year = year, 
                             day = day, 
                             start = wday))
#> # A tibble: 49 x 7
#>    wday   week       week_number  year day   aweek      the_date  
#>    <chr>  <chr>            <dbl> <dbl> <chr> <aweek>    <date>    
#>  1 Sunday 2019-W10-1          10  2019 1     2019-W10-1 2019-03-03
#>  2 Sunday 2019-W10-2          10  2019 2     2019-W10-2 2019-03-04
#>  3 Sunday 2019-W10-3          10  2019 3     2019-W10-3 2019-03-05
#>  4 Sunday 2019-W10-4          10  2019 4     2019-W10-4 2019-03-06
#>  5 Sunday 2019-W10-5          10  2019 5     2019-W10-5 2019-03-07
#>  6 Sunday 2019-W10-6          10  2019 6     2019-W10-6 2019-03-08
#>  7 Sunday 2019-W10-7          10  2019 7     2019-W10-7 2019-03-09
#>  8 Monday 2019-W09-7           9  2019 7     2019-W10-1 2019-03-03
#>  9 Monday 2019-W10-1          10  2019 1     2019-W10-2 2019-03-04
#> 10 Monday 2019-W10-2          10  2019 2     2019-W10-3 2019-03-05
#> # … with 39 more rows

Or, if you have ISO week formatted strings (YYYY-Www-d), then you can use
as.aweek():

datf %>%
  mutate(aweek = as.aweek(week, start = wday)) %>%
  mutate(the_date = as.Date(aweek))
#> # A tibble: 49 x 7
#>    wday   week       week_number  year day   aweek      the_date  
#>    <chr>  <chr>            <dbl> <dbl> <chr> <aweek>    <date>    
#>  1 Sunday 2019-W10-1          10  2019 1     2019-W10-1 2019-03-03
#>  2 Sunday 2019-W10-2          10  2019 2     2019-W10-2 2019-03-04
#>  3 Sunday 2019-W10-3          10  2019 3     2019-W10-3 2019-03-05
#>  4 Sunday 2019-W10-4          10  2019 4     2019-W10-4 2019-03-06
#>  5 Sunday 2019-W10-5          10  2019 5     2019-W10-5 2019-03-07
#>  6 Sunday 2019-W10-6          10  2019 6     2019-W10-6 2019-03-08
#>  7 Sunday 2019-W10-7          10  2019 7     2019-W10-7 2019-03-09
#>  8 Monday 2019-W09-7           9  2019 7     2019-W10-1 2019-03-03
#>  9 Monday 2019-W10-1          10  2019 1     2019-W10-2 2019-03-04
#> 10 Monday 2019-W10-2          10  2019 2     2019-W10-3 2019-03-05
#> # … with 39 more rows

change_week_start() will allow you to re-calculate week numbers based on a
different week_start attribute

w <- as.aweek(datf$week, start = datf$wday)
head(w)
#> <aweek start: Sunday>
#> [1] "2019-W10-1" "2019-W10-2" "2019-W10-3" "2019-W10-4" "2019-W10-5"
#> [6] "2019-W10-6"
head(change_week_start(w, "Monday"))
#> <aweek start: Monday>
#> [1] "2019-W09-7" "2019-W10-1" "2019-W10-2" "2019-W10-3" "2019-W10-4"
#> [6] "2019-W10-5"

factor_aweek() will aggregate your weeks with levels that span the range.

factor_aweek(as.aweek(as.Date(c("2019-03-03", "2019-05-31"))))
#> <aweek start: Sunday>
#> [1] 2019-W10 2019-W22
#> 13 Levels: 2019-W10 2019-W11 2019-W12 2019-W13 2019-W14 ... 2019-W22

^{Created on 2019-05-30 by the reprex package (v0.3.0)}

[zkamvar]

@scottyaz, since you've used aweek in the past, would you mind testing out this version?

# install.packages("remotes")
remotes::install_github("reconhub/aweek#21")

[scottyaz]

Sure but it may be a few days.

…

On May 31, 2019, at 2:37 PM, Zhian N. Kamvar ***@***.***> wrote: @scottyaz, since you've used aweek in the past, would you mind testing out this version? # install.packages("remotes") remotes::install_github("reconhub/aweek#21") — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[zkamvar]

Sure but it may be a few days.

Thank You!

[dirkschumacher]

Just a quick note before I can look at it more in depth: I thought ISO weeks always start on Monday. Wouldn't a call to set_week_start invalidate all ISO weeks generated?

[zkamvar]

Just a quick note before I can look at it more in depth: I thought ISO weeks always start on Monday. Wouldn't a call to set_week_start invalidate all ISO weeks generated?

Yes, ISO weeks all start on Monday, but all aweek objects contain their own week_start attribute, which only changes if the user specifies, so using set_week_start() will not invalidate any previously created aweek objects.

[dirkschumacher]

Installed and tested it. I don't have the time to go through the code really but it looks good to me and I really like the package.