Parse a raw source file into a ledger frame

Each parser reads one source's raw file, with the publisher's own column names, and returns the name of a temporary view in the standard frame shape md_apply_release() takes. Parsing is the whole job: which names and addresses a record states, under which role, with which key. Identity, diffing and writing are the engine's (dev/IDENTITY.md §5).

Usage

md_parse_efile_header(con, relation, view = "frame_efile_header")

md_parse_schedi(con, relation, view = "frame_schedi")

md_parse_bmf_active(con, relation, view = "frame_bmf_active")

md_parse_address_crosswalk(con, relation, view = "frame_address_crosswalk")

md_parse_bmf_unified(con, relation, view = "frame_bmf_unified")

md_parse_f1023ez(con, relation, year, view = "frame_f1023ez")

md_parse_efile_related(con, relation, view = "frame_efile_related")

md_parse_affiliate_listing(con, relation, view = "frame_affiliate_listing")

md_parse_e_postcard(con, relation, view = "frame_e_postcard")

md_parse_revocations(con, relation, view = "frame_revocations")

Arguments

con: A DuckDB connection.
relation: The raw file as a table, view or read_csv(...) expression.
view: Name for the temporary view returned.
year: For md_parse_f1023ez(): the approval year of the file.

Value

The view's name.

Details

Text is passed through as the source wrote it; only structure is parsed (an EIN to EIN2, a ZIP into its five and four digits, name lines joined). Case, punctuation and abbreviations are the normalizer's business, in the derived columns.

Phase 5 step 4 ports one source of each kind:

md_parse_efile_header(): series. One frame row per name on the return - legal, DBA, in care of - each with the return's address. record_key is OBJECTID.
md_parse_schedi(): mentions. One row per grant line: the recipient's name and address as the filer wrote them, asserted_by the filer. Identical lines in one return are kept, and the engine numbers them.
md_parse_bmf_active(): snapshot. Names and the address as separate rows, so a renamed organisation that did not move keeps its address spell, and the other way round.
md_parse_address_crosswalk(): cache. One row per address spell, with the range the crosswalk states as first_seen/last_seen.

Step 5 ports the remaining sources that write names or addresses:

md_parse_bmf_unified(): seed. Like the monthly BMF, names and the address are separate rows; each carries the EIN's first and last vintage, stored as an envelope ("the EIN existed over this range and ended with this value"), not as a spell.
md_parse_f1023ez(): series, one partition per approval year. The legal name with the address, keyed on the EIN.
md_parse_efile_related(): mentions. Schedule R Part II: each related organisation's name and address as the filer wrote them.
md_parse_affiliate_listing(): mentions. A group return's roster: each subordinate's own name, and from e-file v3_1 its address. A row needs a full nine-digit member EIN, and a parent listing itself is dropped, as in md_ingest_affiliate_listing().
md_parse_e_postcard(): snapshot. The legal name and up to three DBAs, and the address, as separate rows. The postcard's tax year is not kept: a spell runs across monthly releases, not across postcards.
md_parse_revocations(): series, one cumulative partition. The legal name and DBA, each with the address as of revocation, keyed on the EIN and revocation date; the year revoked is the period.