Class DoclingThreadedPdfParser

docling_parse/pdf_parser.py:1393–1546

Threaded PDF parser that decodes pages from multiple documents in parallel.

class DoclingThreadedPdfParser:
    """Threaded PDF parser that decodes pages from multiple documents in parallel."""

    def __init__(
        self,
        parser_config: ThreadedPdfParserConfig | None = None,
        decode_config: DecodeConfig | None = None,
    ):
        if parser_config is None:
            parser_config = ThreadedPdfParserConfig()

        self._parser_config = parser_config
        if parser_config.render_config is not None:
            parser_config.render_config = _validated_render_config(
                parser_config.render_config
            )
        self._decode_config = (decode_config or DecodeConfig()).model_copy()
        self._content_config = (
            parser_config.page_content_config or ContentConfig()
        ).model_copy()
        self._batch_content_config = self._content_config.model_copy()
        # The threaded C++ parser decodes the whole batch with one fixed config.
        self._cpp_decode_config = _compile_decode_config(
            decode_config=self._decode_config,
            page_boundary=parser_config.boundary_type.value,
            content_config=self._batch_content_config,
        )
        self._page_counts: Dict[str, int] = {}
        self._scheduled_page_counts: Dict[str, int] = {}

        if parser_config.render_config is None:
            self._parser = _threaded_pdf_parser(
                loglevel=parser_config.loglevel,
                num_threads=parser_config.threads,
                max_concurrent_results=parser_config.max_concurrent_results,
                config=self._cpp_decode_config,
            )
        else:
            self._parser = _threaded_pdf_renderer(
                loglevel=parser_config.loglevel,
                num_threads=parser_config.threads,
                max_concurrent_results=parser_config.max_concurrent_results,
                decode_config=self._cpp_decode_config,
                render_config=parser_config.render_config,
            )

    def load(
        self,
        path_or_stream: Union[str, Path, BytesIO],
        password: str | None = None,
        page_numbers: Sequence[int] | None = None,
    ) -> str:
        """Load a document for parallel processing.

        Parameters:
            path_or_stream: File path or BytesIO object.
            password: Optional password for protected files.
            page_numbers: Optional 1-indexed physical pages to schedule.