Explode person profiles into a long blocking-key table

Turns each build_person_profile() row into the set of hash buckets it belongs to – the stage-1 input for cross-organization linkage. Because a profile keeps its variant sets, it emits the Cartesian product of its surname phonetic keys, states, and (truncated) NTEE codes for each requested pass, so a person with a maiden and a married surname, or activity in two states, lands in every matching bucket instead of a single one.

Usage

person_blocking_keys(
  profiles,
  passes = names(blocking_passes()),
  ntee_digits = 1L
)

Arguments

profiles

Output of build_person_profile() (needs last_name_keys; states/ntee list-columns for the geography/industry passes).

passes

Character vector naming the passes to emit (see blocking_passes()); default all four.

ntee_digits

Number of leading characters of each NTEE code to key on (default 1 = major group, e.g. "B25" -> "B").

Value

A data frame EMP_ID, pass, block_key (one row per bucket a person lands in). block_key is "<pass>|<last>[|<state>][|<ntee>]".

Details

Keys are namespaced by pass, so a strict bucket and a surname bucket never collide even when they share a surname key. A pass that needs a component the profile lacks (e.g. strict when states is empty) simply emits nothing for that profile.

See also

candidate_pairs() for the hash join over these keys.

Examples

if (FALSE) { # \dontrun{
profiles <- build_person_profile(linked, state = "state", ntee = "ntee")
keys <- person_blocking_keys(profiles)
table(keys$pass)
} # }