"""Production-path regression harness for grow_articles_final.

Unlike _regress_grow.py (which calls _grow_article_downward directly with
dateline_starts=None and so over-estimates wrapping), this mirrors the REAL
pipeline tail: it computes dateline_starts via find_article_starts_by_dateline
and runs grow_articles_final exactly as production does. A diff of two snapshots
therefore isolates what the code change does ON THE SHIPPING PATH.

Usage:
    venv/bin/python3 _regress_final.py snapshot.json        # snapshot current code
    venv/bin/python3 _regress_final.py --diff a.json b.json # compare two snapshots
"""
import json
import sys
from collections import Counter
from pathlib import Path

OUT = Path("output")


def paper_of_height(h):
    if 7040 <= h <= 7060:
        return "sakshi"
    if h in (6833, 6549):
        return "andhra_jyothi"
    if h == 6422:
        return "namaste_telangana"
    return "unknown"


def union_bbox(member_ids, region_map):
    bs = [region_map[r]["bbox"] for r in member_ids if r in region_map]
    if not bs:
        return None
    return [min(b[0] for b in bs), min(b[1] for b in bs),
            max(b[2] for b in bs), max(b[3] for b in bs)]


def run():
    import copy
    from PIL import Image
    from smart_extractor import find_article_starts_by_dateline, grow_articles_final

    runs = sorted([d for d in OUT.iterdir() if d.is_dir()],
                  key=lambda d: d.stat().st_mtime, reverse=True)
    snapshot = {}
    papers = Counter()
    for run_dir in runs:
        pages_dir = run_dir / "pages"
        if not pages_dir.exists():
            continue
        for reg_path in sorted(pages_dir.glob("page_*.regions.json")):
            pol_path = reg_path.with_name(reg_path.name.replace(
                ".regions.json", ".political_result.json"))
            png = reg_path.with_name(reg_path.name.replace(".regions.json", ".png"))
            if not pol_path.exists() or not png.exists():
                continue
            regions = json.loads(reg_path.read_text())["regions"]
            region_map = {r["id"]: r for r in regions}
            political = json.loads(pol_path.read_text()).get("political_articles", [])
            if not political:
                continue
            try:
                paper = paper_of_height(Image.open(png).height)
                dateline_starts = find_article_starts_by_dateline(
                    regions, str(png), paper)
                grown = grow_articles_final(
                    copy.deepcopy(political), regions, dateline_starts)
            except Exception as e:
                print(f"  SKIP {run_dir.name}/{reg_path.stem}: {e}")
                continue

            for i, art in enumerate(grown):
                key = f"{run_dir.name}/{reg_path.stem}/art{i}"
                members = sorted(art.get("member_region_ids", []) or [])
                snapshot[key] = {
                    "paper": paper,
                    "headline": art.get("headline_english", "")[:50],
                    "members": members,
                    "bbox": union_bbox(members, region_map),
                }
                papers[paper] += 1
    print(f"  papers: {dict(papers)}")
    return snapshot


def diff(path_a, path_b):
    a = json.loads(Path(path_a).read_text())
    b = json.loads(Path(path_b).read_text())
    keys = sorted(set(a) | set(b))

    per_total = Counter()
    per_changed = Counter()
    grew = Counter()
    shrank = Counter()
    samples = []

    for k in keys:
        ra, rb = a.get(k), b.get(k)
        paper = (rb or ra).get("paper", "unknown")
        per_total[paper] += 1
        if ra is None or rb is None:
            per_changed[paper] += 1
            samples.append((k, paper, "ONLY IN ONE SNAPSHOT", None, None))
            continue
        sa, sb = set(ra["members"]), set(rb["members"])
        if sa != sb:
            per_changed[paper] += 1
            added = sorted(sb - sa)
            removed = sorted(sa - sb)
            if added:
                grew[paper] += 1
            if removed:
                shrank[paper] += 1
            samples.append((k, paper, rb["headline"], added, removed))

    print("\n=== PRODUCTION-PATH DIFF (A=baseline/OLD  ->  B=new) ===")
    print(f"{'paper':<18}{'articles':>10}{'changed':>9}{'grew':>7}{'shrank':>8}")
    for paper in sorted(per_total):
        print(f"{paper:<18}{per_total[paper]:>10}{per_changed[paper]:>9}"
              f"{grew[paper]:>7}{shrank[paper]:>8}")
    tot = sum(per_total.values())
    chg = sum(per_changed.values())
    print(f"{'TOTAL':<18}{tot:>10}{chg:>9}{sum(grew.values()):>7}{sum(shrank.values()):>8}")

    if samples:
        print(f"\n--- {len(samples)} changed article(s) ---")
        for k, paper, hl, added, removed in samples:
            print(f"\n[{paper}] {k}")
            print(f"  headline: {hl}")
            if added:
                print(f"  + added : {added}")
            if removed:
                print(f"  - removed: {removed}")
    else:
        print("\nNo differences — member sets identical across all fixtures.")
    return chg


if __name__ == "__main__":
    if len(sys.argv) >= 4 and sys.argv[1] == "--diff":
        diff(sys.argv[2], sys.argv[3])
    else:
        out_path = sys.argv[1] if len(sys.argv) > 1 else "/tmp/final_snapshot.json"
        snap = run()
        Path(out_path).write_text(json.dumps(snap, indent=2, ensure_ascii=False))
        print(f"wrote {len(snap)} article snapshots to {out_path}")