Files
DataMate/runtime/ops/formatter/unstructured_formatter/process.py
hhhhsc701 f1bffdcd61 bugfix: 创建清洗任务时修改数据集状态;无法删除已在模板/运行任务的算子
* bugfix: 创建清洗任务时修改数据集状态;无法删除已在模板/运行任务的算子
2025-11-27 17:34:53 +08:00

38 lines
1.2 KiB
Python

#!/user/bin/python
# -*- coding: utf-8 -*-
"""
Description: 非结构化文本抽取
Create: 2025/10/22 15:15
"""
import time
from typing import Dict, Any
from loguru import logger
from unstructured.partition.auto import partition
from datamate.core.base_op import Mapper
class UnstructuredFormatter(Mapper):
"""把输入的非结构化文本抽取为txt"""
def __init__(self, *args, **kwargs):
super(UnstructuredFormatter, self).__init__(*args, **kwargs)
def execute(self, sample: Dict[str, Any]) -> Dict[str, Any]:
start = time.time()
filepath = sample.get(self.filepath_key)
filename = sample.get(self.filename_key)
if not filename.lower().endswith((".ppt", ".pptx", "docx", "xlsx", ".csv")):
return sample
try:
elements = partition(filename=filepath)
sample[self.text_key] = "\n\n".join([str(el) for el in elements])
logger.info(f"fileName: {filename}, method: UnstructuredFormatter costs {(time.time() - start):6f} s")
except UnicodeDecodeError as err:
logger.exception(f"fileName: {filename}, method: UnstructuredFormatter causes decode error: {err}")
raise
return sample