Read and prepare the sample grant CSVs

Usage

pf_read_grants(
  path = "inputs",
  pattern = "grants.*\\.csv$",
  allow_mixed = FALSE
)

Arguments

path: Directory or vector of CSV paths.
pattern: File pattern when path is a directory. The default matches the *_grants*.csv benchmark files and so will pull in every size variant sitting in the folder — pass an explicit pattern or file vector to pick one set.
allow_mixed: Set TRUE to suppress the vintage check described below. Only do this when the mixture is genuinely wanted.

Value

The prepared frame, with source_file recording provenance.

Vintage check

Grant extracts from different sources do not share a schema, and silently concatenating them produces a population that is neither. The two vintages in this project are distinguished by OBJECTID: the NCCS e-file rebuild carries it, the older GT-derived extracts do not. Reading both at once also double-counts, because the two sources were sampled from overlapping filings.

The default pattern matched 15 files across both vintages in inputs/ — 422,806 rows where either population alone is a fraction of that — so this function now stops when the matched files disagree on OBJECTID, naming the files on each side. The check is on the column, not on filenames, so it keeps working whatever the samples are called.