"""Pytest configuration and fixtures for AI Creative Studio & Moldovan AI Media Studio.
"""

import glob
import json
import os
import sys
import tempfile
from pathlib import Path
import pytest

PROJECT_ROOT = Path(__file__).resolve().parent.parent
MOLDOVAN_DIR = PROJECT_ROOT / "moldovan-qwen"

# Ensure root and subproject are on sys.path
if str(PROJECT_ROOT) not in sys.path:
    sys.path.insert(0, str(PROJECT_ROOT))
if str(MOLDOVAN_DIR) not in sys.path:
    sys.path.insert(0, str(MOLDOVAN_DIR))

# Set test environment flags
os.environ["HF_SKIP_PRELOAD"] = "1"
os.environ["TESTING"] = "1"

# Isolate ALL generated-media writes away from the real repo data/ tree.
# MOLDOVAN_DATA_ROOT is read by media_creator.py, studio_server.py, and
# daily_album_cron.py at import time; a per-process scratch root keeps the
# repo clean while still exercising real file I/O.
_REPO_DATA = MOLDOVAN_DIR / "data"
_DATA_ROOT = os.environ.get("MOLDOVAN_DATA_ROOT") or os.path.join(
    tempfile.gettempdir(), "moldovan_qwen_test_data"
)
os.environ["MOLDOVAN_DATA_ROOT"] = _DATA_ROOT


def _seed_scratch_data_root() -> None:
    """Copy read-only seed sources from the repo data/ into the scratch root once.

    Modules write generated artifacts under MOLDOVAN_DATA_ROOT, but they also
    READ seed inputs (lexicon, regionalisms, corpus, phonetic dictionary,
    master dictionary). Copy them so tests see real data while the repo
    data/ tree stays untouched.
    """
    import shutil

    os.makedirs(_DATA_ROOT, exist_ok=True)
    seeds = [
        "moldovan_lexicon.json",
        "regionalisme_moldovenesc.jsonl",
        "moldovan_master_dictionary.json",
        "phonetic_dictionary.json",
    ]
    for name in seeds:
        src = _REPO_DATA / name
        dst = os.path.join(_DATA_ROOT, name)
        if src.exists() and not os.path.exists(dst):
            shutil.copy2(src, dst)
    corpus_src = _REPO_DATA / "youtube_corpus"
    corpus_dst = os.path.join(_DATA_ROOT, "youtube_corpus")
    if corpus_src.exists() and not os.path.exists(corpus_dst):
        shutil.copytree(corpus_src, corpus_dst)
    # Read-only storyboards: DOM closure tests drive the timeline UI from these
    # (44 sampled chips + selectable scenes). Server never writes here in tests.
    scripts_src = _REPO_DATA / "generated_video_scripts"
    scripts_dst = os.path.join(_DATA_ROOT, "generated_video_scripts")
    if scripts_src.exists() and not os.path.exists(scripts_dst):
        shutil.copytree(scripts_src, scripts_dst)
        # Sanitize: drop absolute image references so scratch data is
        # self-contained (no console 404s for images not copied to scratch).
        for sp in glob.glob(os.path.join(scripts_dst, "*.json")):
            try:
                with open(sp, "r", encoding="utf-8") as f:
                    board = json.load(f)
                for scene in board.get("scenes", []):
                    scene.pop("image_url", None)
                    scene.pop("image_path", None)
                with open(sp, "w", encoding="utf-8") as f:
                    json.dump(board, f, ensure_ascii=False)
            except Exception:
                pass


_seed_scratch_data_root()
os.makedirs(os.path.join(_DATA_ROOT, "generated_music"), exist_ok=True)


@pytest.fixture(autouse=True)
def setup_test_env(monkeypatch):
    """Automatically ensure test flags and paths are set for all tests."""
    monkeypatch.setenv("HF_SKIP_PRELOAD", "1")
    monkeypatch.setenv("TESTING", "1")
    monkeypatch.setenv("MOLDOVAN_DATA_ROOT", _DATA_ROOT)