| 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455 |
- """数据处理工具的装配工厂 —— **唯一需要改动的替换入口**。
- 默认加载 ``pipeline.DefaultCustomsProcessor``。
- 如果要把整个数据处理工具换成另一套实现(例如改成 PySpark 版本、换成
- 流式处理、或换一个完全不同口径的统计引擎),只要满足 ``contract.CustomsProcessor``
- 协议,然后用环境变量指向它:
- .. code-block:: bash
- set CUSTOMS_PROCESSOR=my_company.customs:MyProcessor
- 平台其余模块(BFF、Agent 工具、前端 /api/process)都不需要改动。
- """
- from __future__ import annotations
- import importlib
- import os
- from .contract import CustomsProcessor
- from .pipeline import DefaultCustomsProcessor
- DEFAULT_SPEC = "toolsvc.customs.pipeline:DefaultCustomsProcessor"
- def _load(spec: str) -> CustomsProcessor:
- module_name, _, attribute = spec.partition(":")
- if not module_name or not attribute:
- raise ValueError(f"CUSTOMS_PROCESSOR 格式应为 '模块:类名',当前为 {spec!r}")
- module = importlib.import_module(module_name)
- target = getattr(module, attribute)
- instance = target() if isinstance(target, type) else target
- for method in ("process",):
- if not hasattr(instance, method):
- raise TypeError(f"{spec} 未实现 CustomsProcessor.{method}")
- return instance
- _cached: CustomsProcessor | None = None
- def load_processor() -> CustomsProcessor:
- """返回单例处理器(线程内复用规则库缓存)。"""
- global _cached
- if _cached is None:
- _cached = _load(os.environ.get("CUSTOMS_PROCESSOR", DEFAULT_SPEC))
- return _cached
- def reset_processor() -> None:
- """测试用:清掉缓存。"""
- global _cached
- _cached = None
|