"""
Single entry point for the study "second_origin_search": regenerates every number, table and figure.

    python3 run_all.py              # uses cached raw downloads when present (a few minutes)
    python3 run_all.py --refresh    # re-downloads every public count (MGnify pages take about 40 minutes; the counts will have grown)
    python3 run_all.py --skip-model # reuse ../data/model_samples*.npz instead of re-running Model 4.1

Steps
  1 fetch_public_counts.py   MGnify, GTDB, NCBI, ENA, PRIDE, MetaboLights counts and MGnify per-analysis summaries -> ../data/db_counts.csv ...
  2 fetch_pubmed_counts.py   PubMed record counts by method class, plus union queries that count each record once -> ../data/pubmed_method_counts.csv
  3 earth_stats.py           unassigned share of sequence by biome (with study counts, reweighting and pooled share), least-annotated studies,
                             genome catalogue                                                                  -> ../data/earth_*.csv/json
  4 model_samples.py         copies models/model4.py to scratch (never modified) and runs it, 60,000 samples per run, under nice -n 10
  5 completeness.py          itemised completeness from ../data/completeness_factors.csv (the only hand-authored input: 76 rows of structured
                             judgement, none of them a measurement) and the null likelihood under three structures (per habitat; per origin
                             for persistence; per origin for every factor = the model's own functional form)
  6 update.py                re-weights the model's samples by the likelihood of the record under each structure; scores the Cheyava Falls
                             potential biosignature over a range of likelihood ratios and finds the break-even; forecasts, claims, independence
  7 figures.py               the five figures

Revision after referee review (2026-10-12): see the "Response to referee" section of report.md for what changed and why.

Raw downloads go to $SOS_RAW (default fermi_raw/second_origin_search). Nothing outside this study directory is written,
except the scratch copy of the model under $SOS_RAW/model_copy.
"""
import os
import subprocess
import sys
import time

HERE = os.path.dirname(os.path.abspath(__file__))
REFRESH = ["--refresh"] if "--refresh" in sys.argv else []
SKIP_MODEL = "--skip-model" in sys.argv

STEPS = [
    (["fetch_public_counts.py"] + REFRESH, False),
    (["fetch_pubmed_counts.py"] + REFRESH, False),
    (["earth_stats.py"], False),
    (["model_samples.py", "60000"], True),
    (["completeness.py"], False),
    (["update.py"], False),
    (["figures.py"], False),
]


def main():
    t00 = time.time()
    for args, nice in STEPS:
        if SKIP_MODEL and args[0] == "model_samples.py":
            print("== skipping model_samples.py (--skip-model)"); continue
        cmd = (["nice", "-n", "10"] if nice else []) + [sys.executable, os.path.join(HERE, args[0])] + args[1:]
        print("==", " ".join(cmd), flush=True)
        t0 = time.time()
        subprocess.run(cmd, check=True, cwd=HERE, env=dict(os.environ, OMP_NUM_THREADS="1", OPENBLAS_NUM_THREADS="1", MKL_NUM_THREADS="1"))
        print(f"== done in {time.time() - t0:.0f} s", flush=True)
    print(f"all steps finished in {time.time() - t00:.0f} s")


if __name__ == "__main__":
    main()