isoweek()/isoyear(): use convertDate() instead of format()

[m-muecke]

Follows the implementation that was commented out and mentioned in the wiki article. I tried out a rough version in C that would speed it up by another 2x, not sure if it's worth it, perhaps as a follow-up.

master

library(data.table, warn.conflicts = FALSE)
setDTthreads(1L)

set.seed(349)
x = sample(as.Date("2026-10-17") - 0:5000, 1e6, replace = TRUE)
bench::mark(isoweek(x), isoyear(x), check = FALSE, min_iterations = 10)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 isoweek(x)    316ms    335ms      3.00     122MB     6.31
#> 2 isoyear(x)    364ms    368ms      2.72     122MB     5.43

^{Created on 2026-10-17 with reprex v2.1.1}

PR

library(data.table, warn.conflicts = FALSE)
setDTthreads(1L)

set.seed(349)
x = sample(as.Date("2026-10-17") - 0:5000, 1e6, replace = TRUE)
bench::mark(isoweek(x), isoyear(x), check = FALSE, min_iterations = 10)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 isoweek(x)  13.29ms  13.54ms      73.4    7.65MB     20.3
#> 2 isoyear(x)   7.64ms   7.81ms     127.     7.63MB     51.9

^{Created on 2026-10-17 with reprex v2.1.1}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (master@c838f30). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #7926   +/-   ##
=========================================
  Coverage          ?   99.02%           
=========================================
  Files             ?       88           
  Lines             ?    17388           
  Branches          ?        0           
=========================================
  Hits              ?    17218           
  Misses            ?      170           
  Partials          ?        0           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ben-schwen]

Change LGTM. How much code would we bring in with the C code?

Could you maybe compare it with fasttime and fastymd to see if its worth for proposing to add it to R?

[github-actions]

• HEAD=isoweek-arith much faster for isoweek improved in #7144
• HEAD=isoweek-arith slower P<0.001 for memrecycle regression fixed in #5463

Generated via commit a6f8c6a

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	1 minutes and 49 seconds
Installing different package versions	13 minutes and 39 seconds
Running and plotting the test cases	5 minutes and 38 seconds

[m-muecke]

It's only about 20 lines added to src/idatetime.c

pure C

library(data.table, warn.conflicts = FALSE)
setDTthreads(1L)

set.seed(349)
x <- sample(as.Date("2026-10-17") - 0:5000, 1e6, replace = TRUE)
bench::mark(isoweek(x), isoyear(x), check = FALSE, min_iterations = 10)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 isoweek(x)   7.02ms   7.09ms      140.    7.65MB     46.7
#> 2 isoyear(x)    5.9ms   5.94ms      168.    7.63MB     66.5

^{Created on 2026-10-18 with reprex v2.1.1}

I didn't know of fasttime and fastymd, but only the latter seems relevant. However, it also doesn't have isoyear() and isoweek(). But it is faster by a margin, since they're using Howard Hinnant's algorithm with a branchless version:

Basic idea one could copy as well:

const unsigned mp = (5 * doy + 2) / 153;          // month index, March = 0
const unsigned d  = doy - (153 * mp + 2) / 5 + 1; // day of month
const unsigned m  = mp + (mp < 10 ? 3 : -9);      // 1..12

library(data.table, warn.conflicts = FALSE)
setDTthreads(1L)

set.seed(349)
x = sample(as.Date("2026-10-17") - 0:5000, 1e6, replace = TRUE)
bench::mark(fastymd::get_year(x), year(x), min_iterations = 10)
#> # A tibble: 2 × 6
#>   expression                min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 fastymd::get_year(x)   5.09ms   5.21ms      190.    15.3MB    373. 
#> 2 year(x)                3.41ms   3.46ms      289.    7.65MB     73.5
bench::mark(fastymd::get_month(x), month(x), min_iterations = 10)
#> # A tibble: 2 × 6
#>   expression                 min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>            <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 fastymd::get_month(x)   5.13ms   5.22ms     191.    15.26MB    332. 
#> 2 month(x)               17.64ms  17.75ms      56.2    7.63MB     12.8
bench::mark(fastymd::get_mday(x), mday(x), min_iterations = 10)
#> # A tibble: 2 × 6
#>   expression                min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>           <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 fastymd::get_mday(x)    5.5ms   5.56ms     178.    15.26MB    207. 
#> 2 mday(x)                17.8ms  18.29ms      54.4    7.63MB     12.4

^{Created on 2026-10-18 with reprex v2.1.1}