The kinds of source

Every source in md_sources() has a kind, which decides how a new release of it changes the database and what one observation of it means (dev/IDENTITY.md section 2):

Usage

md_source_kinds()

Value

A character vector of kinds.

Details

cache

Compiled once; a new release is a correction. One observation is one row as delivered. A new release is diffed as a whole file.

snapshot

Republishes the current state, one row per EIN, overwriting old values (the monthly BMF, the e-Postcard). One observation is a spell: an unbroken run of releases with the same value.

seed

A snapshot that has already been folded into one row per EIN (bmf_unified). Loaded once; its date ranges bound the EIN, not the value.

series

An organisation describing itself once per period (the 990 header, 1023-EZ), or a cumulative list re-pulled whole. One observation is one period; corroboration is how many periods a value appeared in.

mentions

Third parties writing about an organisation as often as they like (Schedule I grants, Schedule R). One observation is one row, repeats included; corroboration is how many times, and by how many filers.

"Snapshot" rather than "roster": in this package a roster is a group-return roster (md_ingest_affiliate_listing()).