MuPDF Extractor

We provide a PDF line extractor built on top of PyMuPdf.

This extractor is the fastest but may not be as portable as that PdfMinerExtractor. However, it should also be relatively easy to install on a wide range of architectures, Linux, OS X and Windows.

License

Beware, PyMuPdf is distributed under the AGPL license, therefore so is this component, and any model depending on this component must be too.

Installation

For the licensing reason mentioned above, the mupdf component is distributed in a separate package edspdf-mupdf. To install it, use your favorite Python package manager :

poetry add edspdf-mupdf
# or
pip install edspdf-mupdf

Example

API-basedConfiguration-based

pipeline.add_pipe(
    "mupdf-extractor",
    config=dict(
        extract_style=False,
    ),
)

[components.mupdf-extractor]
@factory = "mupdf-extractor"
extract_style = false

and use it as follows:

from pathlib import Path

# Apply on a new document
pipeline(Path("path/to/your/pdf/document").read_bytes())

Parameters

PARAMETER	DESCRIPTION
pipeline	The pipeline object TYPE: Pipeline DEFAULT: None
name	Name of the component TYPE: str DEFAULT: 'mupdf_extractor'
extract_style	Extract style TYPE: bool DEFAULT: False
raise_on_error	Whether to raise an error when parsing a corrupted PDF (defaults to False) TYPE: bool DEFAULT: False
use_cropbox	Whether to use the cropbox instead of the mediabox (defaults to True) TYPE: bool DEFAULT: True
render_pages	Whether to extract the rendered page as a numpy array in the page.image attribute (defaults to False) TYPE: bool DEFAULT: False
render_dpi	DPI to use when rendering the page (defaults to 200) TYPE: int DEFAULT: 200
sort_mode	Box sorting mode "blocks": sort while keeping blocks of boxes intaxct. Use this mode if you trust the PDF to have been generated by a tool that produces blocks of text. "lines": sort by lines, without preserving the order of lines inside blocks "none": do not sort boxes TYPE: Literal['blocks', 'lines', 'none'] DEFAULT: 'none'

---