Skip to contents

Load the extracts into DuckDB and reduce them to an alias table

Source: R/schedi.R
pf_schedi_build.Rd

Three tables are created. grants_raw is the union of the yearly CSVs (union_by_name, because the pre-2019 files predate the RecipientTable column). recip is one normalized row per grant. alias is the payload: one row per distinct (ein, name_norm, city, state, zip5) with n_obs and n_filers.

Usage

pf_schedi_build(
  dir = pf_schedi_dir(),
  db = file.path(dir, "schedi_grants.duckdb"),
  rebuild = FALSE,
  threads = 4L
)

Arguments

dir

Directory holding the CSVs.

db

Path to the DuckDB file.

rebuild

Reload even if grants_raw already exists.

threads

DuckDB thread count.

Value

db, invisibly.

Details

n_filers is the trust signal and the reason this works — it counts independent observations, so it separates a real alias from one filer's typo. See the precision table in pf_schedi.