import numpy as np
import pandas as pd
from qlib.data.dataset import DatasetH

dates = pd.to_datetime(pd.date_range("2020-01-01", "2020-01-10"))
instruments = ["SH600000", "SH600001"]
index = pd.MultiIndex.from_product([dates, instruments], names=["datetime", "instrument"])

data = {
    "feature_1": np.random.randn(len(index)),
    "feature_2": np.random.randn(len(index)),
    "label": np.random.randn(len(index)) * 0.01
}
my_df = pd.DataFrame(data, index=index)
my_df.iloc[1, 0] = np.nan  # 人为制造一个缺失值
my_df.iloc[5, 2] = np.nan  # 人为制造一个标签缺失值

print("----------- 原始 DataFrame -----------")
print(my_df.head())

# 2. 创建包含 StaticDataLoader 和 Processors 的完整配置
data_handler_config = {
    "class": "DataHandlerLP",
    "module_path": "qlib.data.dataset.handler",
    "kwargs": {
        # 核心部分：配置数据加载器
        "data_loader": {
            "class": "StaticDataLoader",
            "module_path": "qlib.data.dataset.loader",
            "kwargs": {
                "config": my_df,  # <--- 在这里将你的DataFrame传入！
            }
        },

        "shared_processors": [

        ],
        "infer_processors": [
            # {"class": "DropnaLabel", "module_path": "qlib.data.dataset.processor"},

        ],
        "learn_processors": [
            {"class": "Fillna", "module_path": "qlib.data.dataset.processor", "kwargs": {"fill_value": 0}},
        ]
    },
}

from qlib.utils import init_instance_by_config

# 3. 使用配置初始化 DataHandler
# 这一步会自动加载 StaticDataLoader 的数据，并运行所有定义的处理器
dh = init_instance_by_config(data_handler_config)
ds = DatasetH(
    dh,
    segments={
        "train": ("20190101", "20221231"),
        "valid": ("20220101", "20231231"),
        "test": ("20240101", "20250101"),
    },
)
# 4. 验证结果
# DK_L (Learn) 数据经过了 DropnaLabel -> ZScoreNorm -> Fillna
learn_data = ds.prepare("all", data_key='learn', segments='train')
print("----------- train DataFrame -----------")

print(learn_data)