BFM-decomp/tools/public_rewrite/build_commit_map.py

#!/usr/bin/env python3
"""build_commit_map.py — the public ordinal → new-hash map, and the private old → new map (P33 C7).

    tools/public_rewrite/build_commit_map.py [--new .run/public_rewrite/repo.git]

Writes docs/commit-map.tsv (PUBLIC): one row per commit of the OLD main in order — `ordinal  new_hash  author_date
committer_date  subject` — with NO old hash anywhere (asserted: the scrub over the file's own bytes makes 0
replacements). The pruned purge-only commit keeps its row with `new_hash` = 40 zeros so the ordinals stay dense and
every `commit:NNNN` token in history has a row. Dates/subjects are read from the NEW commits (already scrubbed by the
rewrite); the pruned row's from the old commit, scrubbed here.
Writes .run/public_rewrite/old-to-new.tsv (PRIVATE, for probe_github.sh): `ordinal  old_hash  new_hash`.
"""
import argparse
import sys

sys.path.insert(0, str(__import__("pathlib").Path(__file__).resolve().parent))
import common as C  # noqa: E402
import scrub  # noqa: E402


def main(argv):
    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
    ap.add_argument("--new", default=str(C.CLONE_DIR))
    ap.add_argument("--old", default=str(C.REPO))
    ap.add_argument("--out", default=str(C.COMMIT_MAP_PUBLIC), help="the public map path (a scratch path for a trial run)")
    ap.add_argument("--private-out", default=str(C.OLD_TO_NEW_FILE))
    a = ap.parse_args(argv)
    out_pub, out_priv = __import__("pathlib").Path(a.out), __import__("pathlib").Path(a.private_out)
    new_repo = __import__("pathlib").Path(a.new)
    import verify_rewrite
    cmap = verify_rewrite.commit_map(new_repo)
    d = C.load_dict()
    old_main = [h for h, e in sorted(d["commits"].items(), key=lambda kv: kv[1].get("ord", 0)) if e["kind"] == "main"]
    if len(old_main) != d["main_count"]:
        C.die(f"dictionary main count mismatch ({len(old_main)} vs {d['main_count']})")
    s = scrub.Scrubber()
    rows, priv = [], []
    for h in old_main:
        o = d["commits"][h]["ord"]
        new = cmap.get(h)
        if new is None:
            C.die(f"old main commit ordinal {o} ({h[:9]}) is not in the commit-map — the clone did not hold the same main")
        if new == C.ZEROS:
            ad, cd, subj = C.git(["log", "-1", "--format=%aI%x00%cI%x00%s", h], a.old).rstrip("\n").split("\0")
            subj = s.scrub_text(subj.encode()).decode("utf-8", "replace")
        else:
            ad, cd, subj = C.git(["log", "-1", "--format=%aI%x00%cI%x00%s", new], new_repo).rstrip("\n").split("\0")
        rows.append(f"{o:04d}\t{new}\t{ad}\t{cd}\t{subj}")
        priv.append(f"{o:04d}\t{h}\t{new}")
    header = ("# docs/commit-map.tsv — the public history's commit map (P33 C7). Generated by tools/public_rewrite/build_commit_map.py; never edit.\n"
              "# Before the public flip the history was rewritten (ROM-derived paths purged, personal addresses mapped, old commit hashes\n"
              "# in every historical doc replaced by the inert token commit:NNNN). NNNN is the ordinal of the commit on main in the ORIGINAL\n"
              "# order (1 = the first commit); this table maps it to the rewritten commit. A row of 40 zeros is the one commit that touched\n"
              "# only purged paths and was pruned. Documents at the tip cite the new hashes directly (tools/public_rewrite/resolve_tokens.py).\n"
              "ordinal\tnew_hash\tauthor_date\tcommitter_date\tsubject\n")
    text = header + "\n".join(rows) + "\n"
    # a commit the rewrite left byte-identical (old == new: e.g. the noreply-authored "Initial commit", which cites no
    # hash and touches no purge path) legitimately keeps its hash — exempt those from the old-hash assertion
    unchanged = {o for o, n in cmap.items() if o == n}
    d2 = {"commits": {h: e for h, e in d["commits"].items() if h not in unchanged}, "ambiguous": d["ambiguous"],
          "excluded": d["excluded"]}
    s2 = scrub.Scrubber(d2)
    out = s2.scrub_text(text.encode())
    if s2.stats["replaced"] or s2.stats["emails"]:
        C.die(f"the public map would carry {s2.stats['replaced']} old-hash tokens / {s2.stats['emails']} addresses — refusing")
    (C.SCRATCH / "unchanged_commits.txt").write_text("\n".join(sorted(unchanged)) + ("\n" if unchanged else ""), encoding="utf-8")
    out_pub.write_text(text, encoding="utf-8")
    out_priv.write_text("ordinal\told_hash\tnew_hash\n" + "\n".join(priv) + "\n", encoding="utf-8")
    zeros = sum(1 for r in rows if "\t" + C.ZEROS + "\t" in r)
    print(f"build_commit_map: {len(rows)} rows -> {out_pub} ({zeros} pruned rows); private old-to-new -> {out_priv}; "
          f"0 old-hash tokens in the public file (asserted; {len(unchanged)} commits unchanged by the rewrite keep their hash)")
    return 0


if __name__ == "__main__":
    sys.exit(main(sys.argv[1:]))