Python MyPy报incompatible types赋值类型不兼容错误如何修复

开发过程中编写了如下Python代码：

def get_tree(args):
    sensor_ids = argToList(args.get("sensor_ids"))
    process_ids = argToList(args.get("process_ids"))

    all_results = {}

    for sensor_id in sensor_ids:
        sensor_id_str = str(sensor_id)
        for process_id in process_ids:
            process_id_str = str(process_id)
            main_process = query_process(sensor_id_str, process_id_str)

            results = {}
            json_response = main_process.json()

            for vertex in json_response["resources"]:
                if vertex["vertex_type"] == "process":
                    results["main_process"] = {"sensor_id": sensor_id_str,
                                               "process_id": process_id_str,
                                               "properties": vertex["properties"]}

                    for edge in vertex["edges"]:
                        if edge == "child_process":
                            results["child_processes"] = []

对上述代码做静态检查时，flake8未提示任何异常，但MyPy抛出如下错误：

Incompatible types in assignment (expression has type
"List[]", target has type "Dict[str, Any]") [assignment]
results["child_processes"] = []
^

MyPy做静态类型检查时，会根据变量第一次被赋值的内容自动推断类型。代码中初始化results = {}后，第一次给results赋值的键是main_process，对应值为字典类型，MyPy会自动把results推断为「键为字符串、值为字典」的字典类型。后续给results["child_processes"]赋值列表类型的值时，就和MyPy之前推断的值类型不匹配，因此抛出类型不兼容的错误。
flake8不做严格的跨赋值语句的类型推断校验，因此不会提示该问题。

有两种常用的无副作用修复方式：

• 显式给results变量加类型标注，明确告知MyPy该字典的值可以是任意类型。Python 3.9及以上版本不需要额外导入模块，初始化时直接写results: dict[str, Any] = {}即可；低版本Python需要从typing模块导入Dict和Any，写为results: Dict[str, Any] = {}。
• 初始化results字典时就提前声明所有用到的键和对应类型的默认值，比如初始化为results = {"main_process": {}, "child_processes": []}，MyPy在初始化阶段就能识别到不同键对应的值类型，不会出现推断错误。

修复后的示例代码（采用显式类型标注方案）：

from typing import Any, Dict

def get_tree(args):
    sensor_ids = argToList(args.get("sensor_ids"))
    process_ids = argToList(args.get("process_ids"))

    all_results = {}

    for sensor_id in sensor_ids:
        sensor_id_str = str(sensor_id)
        for process_id in process_ids:
            process_id_str = str(process_id)
            main_process = query_process(sensor_id_str, process_id_str)

            # 显式标注类型，修正MyPy类型推断偏差
            results: Dict[str, Any] = {}
            json_response = main_process.json()

            for vertex in json_response["resources"]:
                if vertex["vertex_type"] == "process":
                    results["main_process"] = {"sensor_id": sensor_id_str,
                                               "process_id": process_id_str,
                                               "properties": vertex["properties"]}

                    for edge in vertex["edges"]:
                        if edge == "child_process":
                            results["child_processes"] = []

内容的提问来源于stack exchange，提问作者Neil