Build deterministic per-record person-year identifiers from source keys

Hashes the native filing/person key (OBJECTID, TABLE_ID) into a stable PYID- identifier. This is the durable back-link that ties a Part VII record to its source row across the raw, classified, and linked panels – distinct from the cross-year cluster id create_emp_ids() stamps as EMP_ID.

Usage

person_year_id(object_id, table_id, keyspec = "v1")

Arguments

object_id, table_id: Equal-length character vectors: the raw OBJECTID and TABLE_ID source columns.
keyspec: Version tag baked into the hash; identifies the recipe.

Value

A character vector of PYID-... identifiers, one per input row.

Details

OBJECTID is parser-independent (it comes from the IRS filing index, not the XML body parse); TABLE_ID is the stored per-person key within a filing. (OBJECTID, TABLE_ID) is unique and non-missing across the raw Part VII data, so unlike a name/title key it needs no disambiguation and is unaffected by downstream name or title cleaning. Both tokens are drawn from [A-Z0-9-], so the | field delimiter below can never occur inside them.

The recipe is frozen under keyspec; bump it if the inputs or delimiter ever change, so old and new ids never silently collide.

Future work – CONTENT_ID

PERSON_YEAR_ID depends on TABLE_ID, which is parser-derived and has broken on pathological XML in the past. A parser revision that renumbers TABLE_ID would shift every id. A parser-independent CONTENT_ID – a hash of normalized native content fields (name variants + raw title + compensation) – could be stored alongside as a recovery bridge to re-map old<->new ids for the records whose content is stable. Not implemented yet; note that a parser change severe enough to break TABLE_ID for a record would usually corrupt that record's content extraction too, so CONTENT_ID mainly insures against a global renumbering of otherwise-clean records. Persisting the raw OBJECTID and TABLE_ID columns (see dev/01_preprocess_year.R) keeps that bridge buildable after the fact.

Examples

person_year_id(c("OID-1", "OID-1"), c("TID-00001", "TID-00002"))
#> [1] "PYID-0028d1e6184c85432a1e870f935920a2"
#> [2] "PYID-0904de62938033fed380241f9265e1b3"