Skip to contents

Generate a working sample from a full population.

Source: R/sample.R
build_working_sample.Rd

Build a working sample from a full (name, downloads) population, stratified by download popularity: a deterministic head of the top_n_head most-downloaded packages, plus a random draw of up to tail_size packages from the remaining long tail.

Usage

build_working_sample(
  downloads_tbl = NULL,
  top_n_head = 15000L,
  tail_size = 40000L,
  label = NULL
)

Arguments

downloads_tbl

A tibble with at least name and downloads columns, as returned by pypi_downloads_full() or npm_downloads_full().

top_n_head

Number of most-downloaded packages to include deterministically.

tail_size

Number of packages to randomly draw from the remaining long tail (outside the head). Capped at the size of that tail.

label

Optional string used to prefix a cli status message (e.g. "PyPI" or "npm"); no message is printed if NULL.

Value

A tibble combining the head and tail samples, with the same columns as downloads_tbl.

Examples

downloads_tbl <- tibble::tibble (
    name = paste0 ("pkg", seq_len (100)),
    downloads = round (100 * exp (-seq_len (100) / 15))
)
sample_tbl <- build_working_sample (
    downloads_tbl,
    top_n_head = 10L,
    tail_size = 20L
)
nrow (sample_tbl)
#> [1] 30