import os
from torchvision.io import read_video
import torchvision.transforms.functional as TF
import json
import pdb

from PIL import Image, ImageChops
import numpy as np, hashlib, io, os, math


def md5(path):
    h = hashlib.md5()
    with open(path, "rb") as f:
        for chunk in iter(lambda: f.read(8192), b""):
            h.update(chunk)
    return h.hexdigest()


metadata_path = "metadata"
# json_list = os.listdir(metadata_path)
json_list = ["426eb9d9-d4cf-4acc-9efb-979f61a5be91.json"]

for json_file in json_list:
    with open(os.path.join("metadata", json_file), "r") as f:
        metadata = json.load(f)

    # load metadata
    video_id = metadata["video_id"]
    video_name = metadata["video_name"]
    start_pt = metadata["start_sec"]
    end_pt = metadata["end_sec"]
    num_output_frames = metadata["num_output_frames"]
    offset = metadata["offset"]
    stride = metadata["stride"]

    # save paths
    video_path = f"ego4d_full_videos/v2/full_scale/{video_id}.mp4"
    save_dir = f"frames/{video_name}"

    print("\n PROCESSING VIDEO ", video_name)

    video, _, info = read_video(
        video_path, start_pts=start_pt, end_pts=end_pt, pts_unit="sec"
    )
    video = video.permute(0, 3, 1, 2).float() / 255.0  # [T, C, H, W]

    os.makedirs(save_dir, exist_ok=True)
    cropped_vid = video[offset::stride, :][:num_output_frames]

    # Save each frame
    for idx, frame_tensor in enumerate(cropped_vid):
        frame = TF.to_pil_image(frame_tensor)  # Convert to PIL image
        # rgb_path = os.path.join(save_dir, "rgb_frames")
        os.makedirs(save_dir, exist_ok=True)
        frame.save(
            os.path.join(save_dir, f"{idx:06d}.jpg")
        )  # Zero-padded filenames

    print(f"Saved {len(cropped_vid)} frames to {save_dir}")

    # # TODO(ilona) - remove this later
    # # checks that current dataset aligns with the files that we have !!
    # for frame_number in [f"{i:06d}" for i in range(0, num_output_frames)]:
    #     paths = [
    #         f"/data/ilona/datasets/ego4d/ego4d_chosen_videos/{video_name}/rgb_frames/{frame_number}.jpg",
    #         # f"/data/ilona/datasets/go4d/frames/{video_name}/rgb_frames/{frame_number}.jpg",
    #         f"/data/ilona/datasets/itto_release/itto/ego4d/frames/{video_name}/{frame_number}.jpg",
    #     ]

    #     # Load images
    #     imgs = [Image.open(p).convert("RGB") for p in paths]

    #     # Basic metadata & hashes
    #     meta = [
    #         (os.path.basename(p), im.size, md5(p)) for p, im in zip(paths, imgs)
    #     ]

    #     # Ensure same size
    #     same_size = imgs[0].size == imgs[1].size

    #     # Pixel-wise difference
    #     diff_img = ImageChops.difference(imgs[0], imgs[1])
    #     diff_array = np.asarray(diff_img)
    #     nonzero_pixels = int(np.count_nonzero(diff_array))

    #     # Mean absolute error per channel
    #     arr0 = np.asarray(imgs[0], dtype=np.int16)
    #     arr1 = np.asarray(imgs[1], dtype=np.int16)
    #     mae = float(np.mean(np.abs(arr0 - arr1)))

    #     # Peak Signal-to-Noise Ratio (PSNR)
    #     mse = float(np.mean((arr0 - arr1) ** 2))
    #     if mse == 0:
    #         psnr = float("inf")
    #     else:
    #         PIXEL_MAX = 255.0
    #         psnr = 20 * math.log10(PIXEL_MAX / math.sqrt(mse))

    #     if mae > 0.01:
    #         print("file 1: ", meta[0][0])
    #         print("file 2: ", meta[1][0])
    #         print("stride: ", stride)
    #         print("mean abs error: ", mae)
    #         print("psnr: ", psnr)