"""Test script for PR #4203: cross-repo LFS file copies.

Creates a model repo (with a txt file + LFS binary) and a dataset repo.
Copies files into the dataset repo, then verifies listing and content integrity.
"""

import os
import tempfile

from huggingface_hub import HfApi


api = HfApi(
    # endpoint="https://feat-cross-repo-lfs-78.us.dev.moon.huggingface.tech",
    # token="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",  # tmp token from GH ephemeral
    endpoint="http://localhost:5564",
    token="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",  # tmp local token
)
USERNAME = api.whoami()["name"]

MODEL_REPO = f"{USERNAME}/test-cross-repo-lfs-copy-model"
DATASET_REPO = f"{USERNAME}/test-cross-repo-lfs-copy-dataset"

# --- Generate random test data ---
txt_content = b"Hello, this is a plain text file for cross-repo copy testing!\n"
lfs_content = os.urandom(1024 * 1024)  # 1 MB random binary -> stored as LFS

# --- Cleanup: delete then recreate everything ---
print("=== Cleanup ===")
for repo_id, repo_type in [(MODEL_REPO, "model"), (DATASET_REPO, "dataset")]:
    try:
        api.delete_repo(repo_id, repo_type=repo_type)
        print(f"  Deleted {repo_type} repo '{repo_id}'")
    except Exception:
        pass

print("\n=== Create repos ===")
api.create_repo(MODEL_REPO, repo_type="model", exist_ok=True)
print(f"  Created model repo '{MODEL_REPO}'")
api.create_repo(DATASET_REPO, repo_type="dataset", exist_ok=True)
print(f"  Created dataset repo '{DATASET_REPO}'")

# --- Upload files to model repo ---
print("\n=== Upload to model repo ===")
api.upload_file(path_or_fileobj=txt_content, path_in_repo="readme.txt", repo_id=MODEL_REPO)
print("  Uploaded readme.txt (regular file)")
api.upload_file(path_or_fileobj=lfs_content, path_in_repo="weights.bin", repo_id=MODEL_REPO)
print("  Uploaded weights.bin (LFS file)")

# --- Copy model repo files to dataset (cross-repo copy!) ---
print("\n=== Copy model -> dataset (cross-repo) ===")
api.copy_files(
    f"hf://{MODEL_REPO}/",
    f"hf://datasets/{DATASET_REPO}/from_model/",
)
print("  Copied model repo files to dataset repo under 'from_model/'")

# --- Verify: list all files in dataset repo ---
print("\n=== Verify: list files in dataset repo ===")
files = list(api.list_repo_tree(DATASET_REPO, repo_type="dataset", recursive=True))
file_paths = sorted(f.path for f in files if hasattr(f, "size") and not f.path.endswith(".gitattributes"))
print(f"  Files found: {file_paths}")
expected = sorted(["from_model/readme.txt", "from_model/weights.bin"])
assert file_paths == expected, f"Expected {expected}, got {file_paths}"
print("  OK - file listing matches!")

# --- Verify: download and compare content ---
print("\n=== Verify: download and compare content ===")
with tempfile.TemporaryDirectory() as tmpdir:
    # 1. readme.txt (regular file from model)
    path = api.hf_hub_download(DATASET_REPO, "from_model/readme.txt", repo_type="dataset", local_dir=tmpdir)
    downloaded = open(path, "rb").read()
    assert downloaded == txt_content, f"readme.txt mismatch: {len(downloaded)} vs {len(txt_content)} bytes"
    print("  OK - from_model/readme.txt content matches")

    # 2. weights.bin (LFS file from model - the main cross-repo LFS copy test!)
    path = api.hf_hub_download(DATASET_REPO, "from_model/weights.bin", repo_type="dataset", local_dir=tmpdir)
    downloaded = open(path, "rb").read()
    assert downloaded == lfs_content, f"weights.bin mismatch: {len(downloaded)} vs {len(lfs_content)} bytes"
    print("  OK - from_model/weights.bin content matches (cross-repo LFS copy works!)")

print("\n=== ALL CHECKS PASSED ===")