Files

168 lines
6.3 KiB
Python
Raw Permalink Normal View History

"""Provide service-layer logic for mineru parser."""
2026-05-14 15:07:34 +08:00
from typing import Optional, Dict
from dataclasses import dataclass, field
from loguru import logger
import os
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
@dataclass
class MinerUResult:
"""Represent the Miner U Result type."""
2026-05-14 15:07:34 +08:00
file_path: str
markdown_text: str
metadata: Dict[str, str] = field(default_factory=dict)
success: bool = True
error_message: str = ""
class MinerUParser:
"""Provide the Miner U Parser parser."""
2026-05-14 15:07:34 +08:00
def __init__(self):
"""Initialize the Miner U Parser instance."""
2026-05-14 15:07:34 +08:00
self.available = self._check_mineru_available()
def _check_mineru_available(self) -> bool:
"""Handle check mineru available for this module for the Miner U Parser instance."""
2026-05-14 15:07:34 +08:00
try:
from magic_pdf.pipe.UNIPipe import UNIPipe
return True
except ImportError:
logger.warning("MinerU (magic-pdf) 未安装,请运行: pip install magic-pdf[full]")
return False
def parse(self, file_path: str, output_dir: Optional[str] = None) -> MinerUResult:
"""Handle parse for the Miner U Parser instance."""
2026-05-14 15:07:34 +08:00
logger.info(f"尝试使用MinerU解析: {file_path}")
if not self.available:
return MinerUResult(
file_path=file_path,
markdown_text="",
success=False,
error_message="MinerU未安装"
)
try:
from magic_pdf.pipe.UNIPipe import UNIPipe
from magic_pdf.libs.MakeContentConfig import DropMode
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
if output_dir is None:
output_dir = os.path.dirname(file_path)
# Keep service responsibilities explicit so downstream behavior stays predictable.
# Keep service responsibilities explicit so downstream behavior stays predictable.
# Keep service responsibilities explicit so downstream behavior stays predictable.
# Keep service responsibilities explicit so downstream behavior stays predictable.
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
pipe = UNIPipe(file_path, output_dir)
# Keep service responsibilities explicit so downstream behavior stays predictable.
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
markdown_content = pipe.pipe_mk()
logger.success(f"MinerU解析成功")
return MinerUResult(
file_path=file_path,
markdown_text=markdown_content,
metadata=self._extract_metadata(pipe),
success=True
)
except Exception as e:
logger.error(f"MinerU解析失败: {e}")
return MinerUResult(
file_path=file_path,
markdown_text="",
success=False,
error_message=str(e)
)
def _extract_metadata(self, pipe) -> Dict[str, str]:
"""Handle extract metadata for this module for the Miner U Parser instance."""
2026-05-14 15:07:34 +08:00
metadata = {}
try:
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
if hasattr(pipe, 'pdf_mid_data') and pipe.pdf_mid_data:
mid_data = pipe.pdf_mid_data
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
metadata = {
"page_count": str(mid_data.get("page_count", "")),
"language": str(mid_data.get("language", "")),
"is_scanned": str(mid_data.get("is_scanned", "")),
}
except Exception as e:
logger.warning(f"提取MinerU元数据失败: {e}")
return metadata
def parse_to_markdown(self, file_path: str) -> str:
"""Parse to markdown for the Miner U Parser instance."""
2026-05-14 15:07:34 +08:00
result = self.parse(file_path)
return result.markdown_text if result.success else ""
class ParserOrchestrator:
"""Represent the Parser Orchestrator type."""
2026-05-14 15:07:34 +08:00
def __init__(self):
"""Initialize the Parser Orchestrator instance."""
2026-05-14 15:07:34 +08:00
from .pdf_parser import PDFParser
self.mineru_parser = MinerUParser()
self.pdf_parser = PDFParser()
self.mineru_available = self.mineru_parser.available
def parse_pdf(self, file_path: str, prefer_mineru: bool = True) -> str:
"""Parse pdf for the Parser Orchestrator instance."""
2026-05-14 15:07:34 +08:00
markdown_text = ""
if prefer_mineru and self.mineru_available:
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
result = self.mineru_parser.parse(file_path)
if result.success:
markdown_text = result.markdown_text
logger.info("使用MinerU解析成功")
return markdown_text
else:
logger.warning(f"MinerU解析失败回退到PyMuPDF: {result.error_message}")
# Keep service responsibilities explicit so downstream behavior stays predictable.
2026-05-14 15:07:34 +08:00
logger.info("使用PyMuPDF基础解析")
markdown_text = self.pdf_parser.parse_to_markdown(file_path)
return markdown_text
def parse_docx(self, file_path: str) -> str:
"""Parse docx for the Parser Orchestrator instance."""
2026-05-14 15:07:34 +08:00
from .docx_parser import DocxParser
docx_parser = DocxParser()
return docx_parser.parse_to_markdown(file_path)
def parse(self, file_path: str) -> str:
"""Handle parse for the Parser Orchestrator instance."""
2026-05-14 15:07:34 +08:00
ext = os.path.splitext(file_path)[1].lower()
if ext == ".pdf":
return self.parse_pdf(file_path)
elif ext in [".docx", ".doc"]:
return self.parse_docx(file_path)
else:
raise ValueError(f"不支持的文件类型: {ext}")
def parse_with_mineru(file_path: str) -> MinerUResult:
"""Parse with mineru."""
2026-05-14 15:07:34 +08:00
parser = MinerUParser()
return parser.parse(file_path)
def parse_pdf_smart(file_path: str) -> str:
"""Parse pdf smart."""
2026-05-14 15:07:34 +08:00
orchestrator = ParserOrchestrator()
return orchestrator.parse_pdf(file_path)