#!/usr/bin/env python3
"""Merge new languages' test material into the eval set and languages.json.

  python3 tools/ai/eval/merge_languages.py DIR [code...]

DIR holds one <code>.json per language: {lang, name, items: [20 x [partner,
target, words]], examples: {sentence: [...], reply: [...]}, notes}. Each one's
items go into items.json, and its examples into ../languages.json as a
language not yet tested ("not-yet", note "Not tested yet."): it is offered
only once run.mjs and the reviewers have passed it. A language already ready
keeps its status, note and examples. Checks the shape before writing anything.
"""

import json
import sys
from pathlib import Path

HERE = Path(__file__).resolve().parent
ITEMS = HERE / "items.json"
LANGS = HERE.parent / "languages.json"
EN_ITEMS = 20


def load(path):
    return json.loads(Path(path).read_text(encoding="utf-8"))


def check(code, d):
    problems = []
    if d.get("lang") != code:
        problems.append(f"lang {d.get('lang')!r}")
    if not str(d.get("name", "")).strip():
        problems.append("no name")
    items = d.get("items", [])
    if len(items) != EN_ITEMS or any(len(i) != 3 or not all(isinstance(x, str) and x.strip() for x in i) for i in items):
        problems.append(f"items: need {EN_ITEMS} rows of [partner, target, words]")
    ex = d.get("examples", {})
    sent, rep = ex.get("sentence", []), ex.get("reply", [])
    if len(sent) != 2 or any(len(s.get("out", [])) != 3 or not s.get("words") for s in sent):
        problems.append("sentence examples: 2, with words and 3 sentences each")
    if not any(s.get("partner") for s in sent) or not any(s.get("partner") is None for s in sent):
        problems.append("sentence examples: one with a partner line, one without")
    if len(rep) != 2 or any(len(r.get("out", [])) != 4 or not r.get("partner") for r in rep):
        problems.append("reply examples: 2, with a partner line and 4 replies each")
    return problems


def main(argv):
    if len(argv) < 2:
        print(__doc__, file=sys.stderr)
        return 2
    src = Path(argv[1])
    wanted = argv[2:] or sorted(p.stem for p in src.glob("*.json"))
    items, langs = load(ITEMS), load(LANGS)
    bad = False
    for code in wanted:
        d = load(src / f"{code}.json")
        problems = check(code, d)
        if problems:
            print(f"✗ {code}: {'; '.join(problems)}", file=sys.stderr)
            bad = True
            continue
        items[code] = d["items"]
        entry = langs["languages"].get(code)
        if entry and entry.get("status") == "ready":
            continue
        langs["languages"][code] = {
            "status": "not-yet",
            "name": d["name"],
            "note": "Not tested yet.",
            "examples": {
                "sentence": [{"partner": s.get("partner"), "words": s["words"], "out": s["out"]} for s in d["examples"]["sentence"]],
                "reply": [{"partner": r["partner"], "out": r["out"]} for r in d["examples"]["reply"]],
            },
        }
    if bad:
        return 1
    langs["languages"] = dict(sorted(langs["languages"].items()))
    ITEMS.write_text(json.dumps(items, indent=1, ensure_ascii=False) + "\n", encoding="utf-8")
    LANGS.write_text(json.dumps(langs, indent=1, ensure_ascii=False) + "\n", encoding="utf-8")
    print(f"✓ merged {', '.join(wanted)}")
    return 0


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