cross

The goal of cross is to rerun a single function across a configuration of package versions or git branches. The typical use case for this is creating a reprex containing a before and after benchmark, where you’d like to run the same code on both the main branch and your branch/my-fix branch, or using the CRAN version of a package against the dev version of a package.

Installation

You can install the development version of cross like so:

pak::pak("DavisVaughan/cross")

Example

cross::run() runs a function across various versions of the same package. The pkgs can be any remote specification supported by pak::pkg_install().

library(bench) # So bench classes print nicely

out <- cross::run(pkgs = c("vctrs", "r-lib/vctrs"), ~{
  library(vctrs)
  x <- c(1, NA, 2, 3, NA)
  bench::mark(missing = vec_detect_missing(x)) 
})

out
#> # A tibble: 2 × 2
#>   pkg         result             
#>   <chr>       <list>             
#> 1 vctrs       <bnch_mrk [1 × 13]>
#> 2 r-lib/vctrs <bnch_mrk [1 × 13]>

tidyr::unnest(out, result)
#> # A tibble: 2 × 14
#>   pkg         expression     min median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc
#>   <chr>       <bch:expr> <bch:t> <bch:>     <dbl> <bch:byt>    <dbl> <int> <dbl>
#> 1 vctrs       missing      816ns  968ns   942535.    2.23KB        0 10000     0
#> 2 r-lib/vctrs missing      760ns  959ns   936836.    2.25KB        0 10000     0
#> # ℹ 5 more variables: total_time <bch:tm>, result <list>, memory <list>,
#> #   time <list>, gc <list>

cross::run_branches() runs a function across various local branches. It assumes that:

• usethis::proj_get() points to an R package using git.
• There are no uncommitted git changes.

If those are true, then it will automatically run the function against the current branch and the main branch, but you can change this with the current and branches arguments.

# Not run here, but assuming we are in the vctrs repo this would run the
# function against whatever branch we are on and the `main` branch

cross::run_branches(~{
  library(vctrs)
  x <- c(1, NA, 2, 3, NA)
  bench::mark(missing = vec_detect_missing(x)) 
})