Run the fplida product pipeline inside a single slice worker — build_fplida_slice

Entry point called inside a PSOCK worker R process. Sets the run-directory option so existing generators find the slice's spine, then invokes each requested product generator in order. Skips spine, core, and blade — these are handled centrally by the orchestrator. Returns per-product timing and metadata.

Usage

build_fplida_slice_worker(
  slice_run_dir,
  slice_id,
  slice_seed,
  years,
  products,
  export_format = "parquet",
  mbs_pbs_chunk = 300000L,
  product_years = NULL,
  stp_zstd_level = NULL
)

Arguments

stp_zstd_level: Integer or NULL. Optional STP ZSTD compression level from 1 to 22. NULL preserves the native compression. Each file is verified before replacement. See build_fplida().
slice_run_dir: Character. Path to slice run directory (already contains _system/base-spine.parquet).
slice_id: Integer. 0-based slice id.
slice_seed: Integer. Seed for this slice (orchestrator pre-computes as base_seed + slice_id * 100000L).
years: Integer vector.
products: Character vector. Products to build (should already exclude spine and core).
export_format: Character. "parquet" or "csv".
mbs_pbs_chunk: Integer. MBS/PBS chunk_size override passed to the worker. The orchestrator sizes this to keep the \(K \times \text{chunk peak}\) total memory comfortably under machine RAM. Default 300000.
product_years: Named list or NULL. Per-product years already narrowed to each dataset's published period by the orchestrator (see plida_dataset_years()). A product missing from the list falls back to years, narrowed here instead.

Value

List with slice_id, elapsed_seconds, product_results.