github/langchain
1
0
Fork 0
mirror of https://github.com/hwchase17/langchain.git synced 2025-06-26 08:33:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

community[patch]: Refactoring PDF loaders: 01 prepare (#29062)

Browse Source 
- **Refactoring PDF loaders step 1**: "community: Refactoring PDF
loaders to standardize approaches"

- **Description:** Declare CloudBlobLoader in __init__.py. file_path is
Union[str, PurePath] anywhere
- **Twitter handle:** pprados

This is one part of a larger Pull Request (PR) that is too large to be
submitted all at once.
This specific part focuses to prepare the update of all parsers.

For more details, see [PR
28970](https://github.com/langchain-ai/langchain/pull/28970).

@eyurtsev it's the start of a PR series.
This commit is contained in:
Philippe PRADOS 2025-01-07 17:00:04 +01:00 committed by GitHub
parent a49448a7c9
commit 2921597c71
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 90 additions and 85 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
libs/community/langchain_community/document_loaders/__init__.py
View File

@ -87,6 +87,7 @@ if TYPE_CHECKING:
from langchain_community.document_loaders.blob_loaders import (
Blob,
BlobLoader,
CloudBlobLoader,
FileSystemBlobLoader,
YoutubeAudioLoader,
)
@ -574,6 +575,7 @@ _module_lookup = {
"CSVLoader": "langchain_community.document_loaders.csv_loader",
"CassandraLoader": "langchain_community.document_loaders.cassandra",
"ChatGPTLoader": "langchain_community.document_loaders.chatgpt",
"CloudBlobLoader": "langchain_community.document_loaders.blob_loaders",
"CoNLLULoader": "langchain_community.document_loaders.conllu",
"CollegeConfidentialLoader": "langchain_community.document_loaders.college_confidential", # noqa: E501
"ConcurrentLoader": "langchain_community.document_loaders.concurrent",
@ -781,6 +783,7 @@ __all__ = [
"CSVLoader",
"CassandraLoader",
"ChatGPTLoader",
"CloudBlobLoader",
"CoNLLULoader",
"CollegeConfidentialLoader",
"ConcurrentLoader",

34
libs/community/langchain_community/document_loaders/parsers/pdf.py
View File

@ -6,7 +6,6 @@ import warnings
from typing import (
TYPE_CHECKING,
Any,
Dict,
Iterable,
Iterator,
Mapping,
@ -23,15 +22,13 @@ from langchain_community.document_loaders.base import BaseBlobParser
from langchain_community.document_loaders.blob_loaders import Blob
if TYPE_CHECKING:
import fitz.fitz
import pdfminer.layout
import pdfplumber.page
import pypdf._page
import pypdfium2._helpers.page
from pypdf import PageObject
import fitz
import pdfminer
import pdfplumber
import pypdf
import pypdfium2
from textractor.data.text_linearization_config import TextLinearizationConfig
_PDF_FILTER_WITH_LOSS = ["DCTDecode", "DCT", "JPXDecode"]
_PDF_FILTER_WITHOUT_LOSS = [
"LZWDecode",
@ -90,7 +87,7 @@ class PyPDFParser(BaseBlobParser):
extract_images: bool = False,
*,
extraction_mode: str = "plain",
extraction_kwargs: Optional[Dict[str, Any]] = None,
extraction_kwargs: Optional[dict[str, Any]] = None,
):
self.password = password
self.extract_images = extract_images
@ -107,7 +104,7 @@ class PyPDFParser(BaseBlobParser):
"`pip install pypdf`"
)
def _extract_text_from_page(page: "PageObject") -> str:
def _extract_text_from_page(page: pypdf.PageObject) -> str:
"""
Extract text from image given the version of pypdf.
"""
@ -126,12 +123,13 @@ class PyPDFParser(BaseBlobParser):
Document(
page_content=_extract_text_from_page(page=page)
+ self._extract_images_from_page(page),
metadata={"source": blob.source, "page": page_number}, # type: ignore[attr-defined]
metadata={"source": blob.source, "page": page_number},
# type: ignore[attr-defined]
)
for page_number, page in enumerate(pdf_reader.pages)
]
def _extract_images_from_page(self, page: pypdf._page.PageObject) -> str:
def _extract_images_from_page(self, page: pypdf.PageObject) -> str:
"""Extract images from page and get the text with RapidOCR."""
if not self.extract_images or "/XObject" not in page["/Resources"].keys(): # type: ignore[attr-defined]
return ""
@ -307,9 +305,7 @@ class PyMuPDFParser(BaseBlobParser):
for page in doc
]
def _get_page_content(
self, doc: fitz.fitz.Document, page: fitz.fitz.Page, blob: Blob
) -> str:
def _get_page_content(self, doc: fitz.Document, page: fitz.Page, blob: Blob) -> str:
"""
Get the text of the page using PyMuPDF and RapidOCR and issue a warning
if it is empty.
@ -327,7 +323,7 @@ class PyMuPDFParser(BaseBlobParser):
return content
def _extract_metadata(
self, doc: fitz.fitz.Document, page: fitz.fitz.Page, blob: Blob
self, doc: fitz.Document, page: fitz.Page, blob: Blob
) -> dict:
"""Extract metadata from the document and page."""
return dict(
@ -344,9 +340,7 @@ class PyMuPDFParser(BaseBlobParser):
},
)
def _extract_images_from_page(
self, doc: fitz.fitz.Document, page: fitz.fitz.Page
) -> str:
def _extract_images_from_page(self, doc: fitz.Document, page: fitz.Page) -> str:
"""Extract images from page and get the text with RapidOCR."""
if not self.extract_images:
return ""
@ -558,7 +552,7 @@ class AmazonTextractPDFParser(BaseBlobParser):
textract_features: Optional[Sequence[int]] = None,
client: Optional[Any] = None,
*,
linearization_config: Optional["TextLinearizationConfig"] = None,
linearization_config: Optional[TextLinearizationConfig] = None,
) -> None:
"""Initializes the parser.

93
libs/community/langchain_community/document_loaders/pdf.py
View File

@ -6,17 +6,17 @@ import tempfile
import time
from abc import ABC
from io import StringIO
from pathlib import Path
from pathlib import Path, PurePath
from typing import (
TYPE_CHECKING,
Any,
Dict,
BinaryIO,
Iterator,
List,
Mapping,
Optional,
Sequence,
Union,
cast,
)
from urllib.parse import urlparse
@ -68,7 +68,7 @@ class UnstructuredPDFLoader(UnstructuredFileLoader):
https://unstructured-io.github.io/unstructured/bricks.html#partition-pdf
"""
def _get_elements(self) -> List:
def _get_elements(self) -> list:
from unstructured.partition.pdf import partition_pdf
return partition_pdf(filename=self.file_path, **self.unstructured_kwargs) # type: ignore[arg-type]
@ -81,7 +81,9 @@ class BasePDFLoader(BaseLoader, ABC):
clean up the temporary file after completion.
"""
def __init__(self, file_path: Union[str, Path], *, headers: Optional[Dict] = None):
def __init__(
self, file_path: Union[str, PurePath], *, headers: Optional[dict] = None
):
"""Initialize with a file path.
Args:
@ -154,7 +156,7 @@ class BasePDFLoader(BaseLoader, ABC):
class OnlinePDFLoader(BasePDFLoader):
"""Load online `PDF`."""
def load(self) -> List[Document]:
def load(self) -> list[Document]:
"""Load documents."""
loader = UnstructuredPDFLoader(str(self.file_path))
return loader.load()
@ -223,13 +225,13 @@ class PyPDFLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
password: Optional[Union[str, bytes]] = None,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
extract_images: bool = False,
*,
extraction_mode: str = "plain",
extraction_kwargs: Optional[Dict] = None,
extraction_kwargs: Optional[dict] = None,
) -> None:
"""Initialize with a file path."""
try:
@ -262,9 +264,9 @@ class PyPDFium2Loader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
*,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
extract_images: bool = False,
):
"""Initialize with a file path."""
@ -290,7 +292,7 @@ class PyPDFDirectoryLoader(BaseLoader):
def __init__(
self,
path: Union[str, Path],
path: Union[str, PurePath],
glob: str = "**/[!.]*.pdf",
silent_errors: bool = False,
load_hidden: bool = False,
@ -308,7 +310,7 @@ class PyPDFDirectoryLoader(BaseLoader):
def _is_visible(path: Path) -> bool:
return not any(part.startswith(".") for part in path.parts)
def load(self) -> List[Document]:
def load(self) -> list[Document]:
p = Path(self.path)
docs = []
items = p.rglob(self.glob) if self.recursive else p.glob(self.glob)
@ -334,9 +336,9 @@ class PDFMinerLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
*,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
extract_images: bool = False,
concatenate_pages: bool = True,
) -> None:
@ -374,7 +376,9 @@ class PDFMinerLoader(BasePDFLoader):
class PDFMinerPDFasHTMLLoader(BasePDFLoader):
"""Load `PDF` files as HTML content using `PDFMiner`."""
def __init__(self, file_path: str, *, headers: Optional[Dict] = None):
def __init__(
self, file_path: Union[str, PurePath], *, headers: Optional[dict] = None
):
"""Initialize with a file path."""
try:
from pdfminer.high_level import extract_text_to_fp # noqa:F401
@ -395,14 +399,14 @@ class PDFMinerPDFasHTMLLoader(BasePDFLoader):
output_string = StringIO()
with open_filename(self.file_path, "rb") as fp:
extract_text_to_fp(
fp,
cast(BinaryIO, fp),
output_string,
codec="",
laparams=LAParams(),
output_type="html",
)
metadata = {
"source": self.file_path if self.web_path is None else self.web_path
"source": str(self.file_path) if self.web_path is None else self.web_path
}
yield Document(page_content=output_string.getvalue(), metadata=metadata)
@ -412,9 +416,9 @@ class PyMuPDFLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
*,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
extract_images: bool = False,
**kwargs: Any,
) -> None:
@ -447,7 +451,7 @@ class PyMuPDFLoader(BasePDFLoader):
blob = Blob.from_path(self.file_path) # type: ignore[attr-defined]
yield from parser.lazy_parse(blob)
def load(self, **kwargs: Any) -> List[Document]:
def load(self, **kwargs: Any) -> list[Document]:
return list(self._lazy_load(**kwargs))
def lazy_load(self) -> Iterator[Document]:
@ -461,11 +465,11 @@ class MathpixPDFLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
processed_file_format: str = "md",
max_wait_time_seconds: int = 500,
should_clean_pdf: bool = False,
extra_request_data: Optional[Dict[str, Any]] = None,
extra_request_data: Optional[dict[str, Any]] = None,
**kwargs: Any,
) -> None:
"""Initialize with a file path.
@ -499,7 +503,7 @@ class MathpixPDFLoader(BasePDFLoader):
self.should_clean_pdf = should_clean_pdf
@property
def _mathpix_headers(self) -> Dict[str, str]:
def _mathpix_headers(self) -> dict[str, str]:
return {"app_id": self.mathpix_api_id, "app_key": self.mathpix_api_key}
@property
@ -515,7 +519,7 @@ class MathpixPDFLoader(BasePDFLoader):
return {"options_json": json.dumps(options)}
def send_pdf(self) -> str:
with open(self.file_path, "rb") as f:
with open(str(self.file_path), "rb") as f:
files = {"file": f}
response = requests.post(
self.url, headers=self._mathpix_headers, files=files, data=self.data
@ -562,7 +566,7 @@ class MathpixPDFLoader(BasePDFLoader):
# This indicates an error with the PDF processing
raise ValueError("Unable to retrieve PDF from Mathpix")
else:
print(f"Status: {status}, waiting for processing to complete") # noqa: T201
logger.info("Status: %s, waiting for processing to complete", status)
time.sleep(5)
raise TimeoutError
@ -572,8 +576,7 @@ class MathpixPDFLoader(BasePDFLoader):
response = requests.get(url, headers=self._mathpix_headers)
return response.content.decode("utf-8")
@staticmethod
def clean_pdf(contents: str) -> str:
def clean_pdf(self, contents: str) -> str:
"""Clean the PDF file.
Args:
@ -596,7 +599,7 @@ class MathpixPDFLoader(BasePDFLoader):
)
return contents
def load(self) -> List[Document]:
def load(self) -> list[Document]:
pdf_id = self.send_pdf()
contents = self.get_processed_pdf(pdf_id)
if self.should_clean_pdf:
@ -610,10 +613,10 @@ class PDFPlumberLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
text_kwargs: Optional[Mapping[str, Any]] = None,
dedupe: bool = False,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
extract_images: bool = False,
) -> None:
"""Initialize with a file path."""
@ -630,7 +633,7 @@ class PDFPlumberLoader(BasePDFLoader):
self.dedupe = dedupe
self.extract_images = extract_images
def load(self) -> List[Document]:
def load(self) -> list[Document]:
"""Load file."""
parser = PDFPlumberParser(
@ -669,13 +672,13 @@ class AmazonTextractPDFLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
textract_features: Optional[Sequence[str]] = None,
client: Optional[Any] = None,
credentials_profile_name: Optional[str] = None,
region_name: Optional[str] = None,
endpoint_url: Optional[str] = None,
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
*,
linearization_config: Optional["TextLinearizationConfig"] = None,
) -> None:
@ -743,7 +746,7 @@ class AmazonTextractPDFLoader(BasePDFLoader):
linearization_config=linearization_config,
)
def load(self) -> List[Document]:
def load(self) -> list[Document]:
"""Load given path as pages."""
return list(self.lazy_load())
@ -758,7 +761,7 @@ class AmazonTextractPDFLoader(BasePDFLoader):
if self.web_path and self._is_s3_url(self.web_path):
blob = Blob(path=self.web_path) # type: ignore[call-arg] # type: ignore[misc]
else:
blob = Blob.from_path(self.file_path) # type: ignore[attr-defined]
blob = Blob.from_path(self.file_path)
if AmazonTextractPDFLoader._get_number_of_pages(blob) > 1:
raise ValueError(
f"the file {blob.path} is a multi-page document, \
@ -792,7 +795,9 @@ class AmazonTextractPDFLoader(BasePDFLoader):
elif blob.mimetype in ["image/png", "image/jpeg"]: # type: ignore[attr-defined]
return 1
else:
raise ValueError(f"unsupported mime type: {blob.mimetype}") # type: ignore[attr-defined]
raise ValueError( # type: ignore[attr-defined]
f"unsupported mime type: {blob.mimetype}"
)
class DedocPDFLoader(DedocBaseLoader):
@ -887,7 +892,7 @@ class DedocPDFLoader(DedocBaseLoader):
from dedoc.utils.langchain import make_manager_pdf_config
return make_manager_pdf_config(
file_path=self.file_path,
file_path=str(self.file_path),
parsing_params=self.parsing_parameters,
split=self.split,
)
@ -898,10 +903,10 @@ class DocumentIntelligenceLoader(BasePDFLoader):
def __init__(
self,
file_path: str,
file_path: Union[str, PurePath],
client: Any,
model: str = "prebuilt-document",
headers: Optional[Dict] = None,
headers: Optional[dict] = None,
) -> None:
"""
Initialize the object for file processing with Azure Document Intelligence
@ -930,10 +935,10 @@ class DocumentIntelligenceLoader(BasePDFLoader):
... )
"""
self.parser = DocumentIntelligenceParser(client=client, model=model)
super().__init__(file_path, headers=headers)
self.parser = DocumentIntelligenceParser(client=client, model=model)
def load(self) -> List[Document]:
def load(self) -> list[Document]:
"""Load given path as pages."""
return list(self.lazy_load())
@ -964,7 +969,7 @@ class ZeroxPDFLoader(BasePDFLoader):
def __init__(
self,
file_path: Union[str, Path],
file_path: Union[str, PurePath],
model: str = "gpt-4o-mini",
**zerox_kwargs: Any,
) -> None:
@ -1005,7 +1010,7 @@ class ZeroxPDFLoader(BasePDFLoader):
# Directly call asyncio.run to execute zerox synchronously
zerox_output = asyncio.run(
zerox(file_path=self.file_path, model=self.model, **self.zerox_kwargs)
zerox(file_path=str(self.file_path), model=self.model, **self.zerox_kwargs)
)
# Convert zerox output to Document instances and yield them

2
libs/community/tests/integration_tests/document_loaders/parsers/test_pdf_parsers.py
View File

@ -61,7 +61,7 @@ def _assert_with_parser(parser: BaseBlobParser, splits_by_page: bool = True) ->
assert metadata["source"] == str(LAYOUT_PARSER_PAPER_PDF)
if splits_by_page:
assert metadata["page"] == 0
assert int(metadata["page"]) == 0
def _assert_with_duplicate_parser(parser: BaseBlobParser, dedupe: bool = False) -> None:

42
libs/community/tests/integration_tests/document_loaders/test_pdf.py
View File

@ -1,3 +1,4 @@
import os
from pathlib import Path
from typing import Sequence, Union
@ -17,7 +18,7 @@ from langchain_community.document_loaders import (
def test_unstructured_pdf_loader_elements_mode() -> None:
"""Test unstructured loader with various modes."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = UnstructuredPDFLoader(str(file_path), mode="elements")
loader = UnstructuredPDFLoader(file_path, mode="elements")
docs = loader.load()
assert len(docs) == 2
@ -26,7 +27,7 @@ def test_unstructured_pdf_loader_elements_mode() -> None:
def test_unstructured_pdf_loader_paged_mode() -> None:
"""Test unstructured loader with various modes."""
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = UnstructuredPDFLoader(str(file_path), mode="paged")
loader = UnstructuredPDFLoader(file_path, mode="paged")
docs = loader.load()
assert len(docs) == 16
@ -35,7 +36,7 @@ def test_unstructured_pdf_loader_paged_mode() -> None:
def test_unstructured_pdf_loader_default_mode() -> None:
"""Test unstructured loader."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = UnstructuredPDFLoader(str(file_path))
loader = UnstructuredPDFLoader(file_path)
docs = loader.load()
assert len(docs) == 1
@ -44,26 +45,26 @@ def test_unstructured_pdf_loader_default_mode() -> None:
def test_pdfminer_loader() -> None:
"""Test PDFMiner loader."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = PDFMinerLoader(str(file_path))
loader = PDFMinerLoader(file_path)
docs = loader.load()
assert len(docs) == 1
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = PDFMinerLoader(str(file_path))
loader = PDFMinerLoader(file_path)
docs = loader.load()
assert len(docs) == 1
# Verify that concatenating pages parameter works
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = PDFMinerLoader(str(file_path), concatenate_pages=True)
loader = PDFMinerLoader(file_path, concatenate_pages=True)
docs = loader.load()
assert len(docs) == 1
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = PDFMinerLoader(str(file_path), concatenate_pages=False)
loader = PDFMinerLoader(file_path, concatenate_pages=False)
docs = loader.load()
assert len(docs) == 16
@ -72,13 +73,13 @@ def test_pdfminer_loader() -> None:
def test_pdfminer_pdf_as_html_loader() -> None:
"""Test PDFMinerPDFasHTMLLoader."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = PDFMinerPDFasHTMLLoader(str(file_path))
loader = PDFMinerPDFasHTMLLoader(file_path)
docs = loader.load()
assert len(docs) == 1
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = PDFMinerPDFasHTMLLoader(str(file_path))
loader = PDFMinerPDFasHTMLLoader(file_path)
docs = loader.load()
assert len(docs) == 1
@ -87,13 +88,13 @@ def test_pdfminer_pdf_as_html_loader() -> None:
def test_pypdfium2_loader() -> None:
"""Test PyPDFium2Loader."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = PyPDFium2Loader(str(file_path))
loader = PyPDFium2Loader(file_path)
docs = loader.load()
assert len(docs) == 1
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = PyPDFium2Loader(str(file_path))
loader = PyPDFium2Loader(file_path)
docs = loader.load()
assert len(docs) == 16
@ -102,13 +103,13 @@ def test_pypdfium2_loader() -> None:
def test_pymupdf_loader() -> None:
"""Test PyMuPDF loader."""
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = PyMuPDFLoader(str(file_path))
loader = PyMuPDFLoader(file_path)
docs = loader.load()
assert len(docs) == 1
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = PyMuPDFLoader(str(file_path))
loader = PyMuPDFLoader(file_path)
docs = loader.load()
assert len(docs) == 16
@ -123,20 +124,21 @@ def test_pymupdf_loader() -> None:
assert len(docs) == 1
@pytest.mark.skipif(
not os.environ.get("MATHPIX_API_KEY"), reason="Mathpix API key not found"
)
def test_mathpix_loader() -> None:
file_path = Path(__file__).parent.parent / "examples/hello.pdf"
loader = MathpixPDFLoader(str(file_path))
loader = MathpixPDFLoader(file_path)
docs = loader.load()
assert len(docs) == 1
print(docs[0].page_content) # noqa: T201
file_path = Path(__file__).parent.parent / "examples/layout-parser-paper.pdf"
loader = MathpixPDFLoader(str(file_path))
loader = MathpixPDFLoader(file_path)
docs = loader.load()
assert len(docs) == 1
print(docs[0].page_content) # noqa: T201
@pytest.mark.parametrize(
@ -187,8 +189,8 @@ def test_mathpix_loader() -> None:
1,
False,
),
(str(Path(__file__).parent.parent / "examples/hello.pdf"), ["FORMS"], 1, False),
(str(Path(__file__).parent.parent / "examples/hello.pdf"), [], 1, False),
(Path(__file__).parent.parent / "examples/hello.pdf", ["FORMS"], 1, False),
(Path(__file__).parent.parent / "examples/hello.pdf", [], 1, False),
(
"s3://amazon-textract-public-content/langchain/layout-parser-paper.pdf",
["FORMS", "TABLES", "LAYOUT"],
@ -222,7 +224,7 @@ def test_amazontextract_loader(
@pytest.mark.skip(reason="Requires AWS credentials to run")
def test_amazontextract_loader_failures() -> None:
# 2-page PDF local file system
two_page_pdf = str(
two_page_pdf = (
Path(__file__).parent.parent / "examples/multi-page-forms-sample-2-page.pdf"
)
loader = AmazonTextractPDFLoader(two_page_pdf)

1
libs/community/tests/unit_tests/document_loaders/test_imports.py
View File

@ -43,6 +43,7 @@ EXPECTED_ALL = [
"CassandraLoader",
"CSVLoader",
"ChatGPTLoader",
"CloudBlobLoader",
"CoNLLULoader",
"CollegeConfidentialLoader",
"ConcurrentLoader",