Stage 3 — run the cascade and propagate to grant rows

Wraps pf_run_batches() (which applies pf_drop_self_matches() per batch) and then pf_propagate_matches(), so every grant row inherits the result resolved for its identity cluster. Out-of-scope rows are marked OOS.

Usage

pf_link(
  prepared,
  ref,
  dir = "data-dev/batches",
  batch_size = 25000L,
  passes = NULL,
  config = npmatch::np_config(),
  method = "hier",
  dyads = NULL,
  dyad_config = pf_dyad_config(),
  resume = TRUE,
  verbose = TRUE
)

Arguments

prepared

Output of pf_prepare().

ref

Output of pf_reference().

dir

Checkpoint directory. Completed batches are skipped on re-run.

batch_size

Identities per checkpoint.

passes, config, method

Passed to pf_run_batches().

dyads

Optional dyad table from pf_build_dyads(). When supplied, the substitutable pairs inside each candidate list are collapsed before selection, which is what removes ties a grant has with itself. NULL disables it. See pf_dyads and dev/DYAD-FINDINGS.md.

dyad_config

Settings for that collapse; see pf_dyad_config().

resume

Reuse checkpoints in dir whose run stamp matches this call. A checkpoint written under different settings stops the run; resume = FALSE recomputes and overwrites. See pf_run_batches(), Checkpoints are stamped.

verbose

Print progress.

Value

A list of class pf_linked with identity-level status, row-level grants, and timing.