Optionally strip `.recording` attribute

[aclark02-arcus]

Problem

Because of the .recording attribute stored in riskmetric assessments, this line of code will bloat the assessment object seen below. In my experience assessing all of CRAN (using 'pkg_cran_remote'), it bloats the assessment object from 847 MB to 1.5 GB.

pharmapkgs/R/PACKAGES.R

Line 159 in 7f7ce4c

assessment <- suppressMessages(riskmetric::pkg_assess(ref, assessments = metrics))

Here is what the attribute looks like for an example metric:

> x <- assessment$covr_coverage
> attributes(x)$.recording
$expr
{
    covr::coverage_to_list(x$covr_coverage)
}

$attributes
$attributes$names
[1] "filecoverage"  "totalcoverage"


$visible
[1] TRUE

$traceback
NULL

$output
character(0)

Resolution

This helper function may come in handy if you plan on populating the PACKAGES file with assessments instead of scores to make computation more nimble:

strip_recording <- function(assessment) {
  no_recording <-
    lapply(assessment, \(x) {
      structure(
        x,
        .recording = NULL,
        class = setdiff(class(x),
                        "with_eval_recording")
      )
    })
  class(no_recording) <- class(assessment)
  return(no_recording)
}

Usage

Note this solution only works if when pkg_ref() remains a list. The solution is a little "uglier" IMHO if you use the common as_tibble() shown below.

assess <- riskmetric::pkg_ref("askpass", source = "pkg_cran_remote") |>
  #   dplyr::as_tibble() |> # Recommend not using this line, which you aren't.
  riskmetric::pkg_assess()
no_recording <- assess |> strip_recording()

# Observe size diff:
object.size(assess)
> 110,480 bytes
object.size(no_recording)
> 22,608 bytes