import sys
from pathlib import Path

# 获取项目根目录（train.py上级的上级）
root_path = Path(__file__).parent.parent
sys.path.append(str(root_path))
import torch
import os
import glob
import numpy as np
import h5py
from tqdm import tqdm
from model.prithvi_wxc import PrithviWxC
from onescience.utils.YParams import YParams
from onescience.datapipes.climate import ERA5Datapipe


def get_stats(data_dir, channels):
    """从新版 h5 中读取变量列表与归一化参数（均值/标准差）"""
    h5_files = sorted(glob.glob(os.path.join(data_dir, "data", "*.h5")))
    with h5py.File(h5_files[0], "r") as f:
        ds = f["fields"]
        all_variables = [v.decode() if isinstance(v, bytes) else v for v in ds.attrs["variables"]]
        mu = f["global_means"][:]   # [1, C, 1, 1]
        std = f["global_stds"][:]

    channel_indices = [all_variables.index(v) for v in channels]
    means = mu[:, channel_indices, :, :]
    stds = std[:, channel_indices, :, :]
    return means, stds


if __name__ == "__main__":
    current_path = os.getcwd()
    sys.path.append(current_path)

    ## Model config init
    config_file_path = os.path.join(current_path, "conf/config.yaml")
    cfg = YParams(config_file_path, "model")

    ## DataLoader init
    cfg_data = YParams(config_file_path, "datapipe")
    means, stds = get_stats(cfg_data.dataset.data_dir, cfg_data.dataset.channels)

    cfg['N_in_channels'] = len(cfg_data.dataset.channels)
    cfg['N_out_channels'] = len(cfg_data.dataset.channels)

    datapipe = ERA5Datapipe(
        dataset_dir=cfg_data.dataset.data_dir,
        used_variables=cfg_data.dataset.channels,
        used_years=cfg_data.dataset.test_time,
        distributed=False,
        input_steps=2,
        batch_size=1,
        num_workers=4,
    )
    test_dataloader, _ = datapipe.get_dataloader("test")

    device = "cuda:0" if torch.cuda.is_available() else "cpu"
    ckpt = torch.load(f"{cfg.checkpoint_dir}/model_bak.pth", map_location=device, weights_only=False)
    model = PrithviWxC(
        in_channels=cfg['N_in_channels'],
        input_size_time=cfg.input_size_time,
        in_channels_static=cfg.in_channels_static,
        n_lats_px=cfg.n_lats_px,
        n_lons_px=cfg.n_lons_px,
        patch_size_px=cfg.patch_size_px,
        mask_unit_size_px=cfg.mask_unit_size_px,
        mask_ratio_inputs=0.0,
        embed_dim=cfg.embed_dim,
        n_blocks_encoder=cfg.n_blocks_encoder,
        n_blocks_decoder=cfg.n_blocks_decoder,
        mlp_multiplier=cfg.mlp_multiplier,
        n_heads=cfg.n_heads,
        dropout=cfg.dropout,
        drop_path=cfg.drop_path,
        parameter_dropout=cfg.parameter_dropout,
        residual=cfg.residual,
        masking_mode=cfg.masking_mode,
        positional_encoding=cfg.positional_encoding,
        encoder_shifting=cfg.encoder_shifting,
        decoder_shifting=cfg.decoder_shifting,
    ).to(device)
    model.load_state_dict(ckpt["model_state_dict"])

    model.eval()
    os.makedirs('result/output/', exist_ok=True)
    print(f"📂 infer results will be generated to './result/output/'")
    H, W = int(cfg.n_lats_px), int(cfg.n_lons_px)
    static_path = os.path.join(cfg_data.dataset.data_dir, "static", "static.npy")
    static_base = torch.from_numpy(np.load(static_path)).to(device=device, dtype=torch.float32).unsqueeze(0)
    expected_static = (1, int(cfg.in_channels_static), H, W)
    if tuple(static_base.shape) != expected_static:
        raise ValueError(f"static data shape {tuple(static_base.shape)} != expected {expected_static}")
    with torch.no_grad():
        for data in tqdm(test_dataloader, desc="Inferring testset", unit="batch"):
            invar = data[0].to(device, dtype=torch.float32)      # [1, 2, C, H, W]
            filename = data[4][-1][0]
            B = invar.shape[0]
            static = static_base.expand(B, -1, -1, -1)
            lead_time = torch.full((B,), 6.0, device=device)
            pred_var = model(invar, static, lead_time=lead_time).cpu().numpy()
            pred_var = pred_var * stds + means
            np.save(f"result/output/{filename}.npy", pred_var)