docling 0.3.0__tar.gz → 0.4.0__tar.gz

{docling-0.3.0 → docling-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling
-Version: 0.3.0
+Version: 0.4.0
 Summary: Docling PDF conversion package
 Home-page: https://github.com/DS4SD/docling
 License: MIT
@@ -31,11 +31,20 @@ Project-URL: Repository, https://github.com/DS4SD/docling
 Description-Content-Type: text/markdown
 
 <p align="center">
-  <a href="https://github.com/ds4sd/docling"> <img loading="lazy" alt="Docling" src="https://github.com/DS4SD/docling/raw/main/logo.png" width="150" /> </a>
+  <a href="https://github.com/ds4sd/docling"> <img loading="lazy" alt="Docling" src="https://github.com/DS4SD/docling/raw/main/logo.png" width="150" />
 </p>
 
 # Docling
 
+[![PyPI version](https://img.shields.io/pypi/v/docling)](https://pypi.org/project/docling/)
+![Python](https://img.shields.io/badge/python-3.11%20%7C%203.12-blue)
+[![Poetry](https://img.shields.io/endpoint?url=https://python-poetry.org/badge/v0.json)](https://python-poetry.org/)
+[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)
+[![Imports: isort](https://img.shields.io/badge/%20imports-isort-%231674b1?style=flat&labelColor=ef8336)](https://pycqa.github.io/isort/)
+[![Pydantic v2](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/pydantic/pydantic/main/docs/badge/v2.json)](https://pydantic.dev)
+[![pre-commit](https://img.shields.io/badge/pre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white)](https://github.com/pre-commit/pre-commit)
+[![License MIT](https://img.shields.io/github/license/ds4sd/deepsearch-toolkit)](https://opensource.org/licenses/MIT)
+
 Docling bundles PDF document conversion to JSON and Markdown in an easy, self-contained package.
 
 ## Features
@@ -44,25 +53,20 @@ Docling bundles PDF document conversion to JSON and Markdown in an easy, self-co
 * 📝 Extracts metadata from the document, such as title, authors, references and language
 * 🔍 Optionally applies OCR (use with scanned PDFs)
 
-## Setup
+## Installation
 
-For general usage, you can simply install `docling` through `pip` from the pypi package index.
-```
+To use Docling, simply install `docling` from your package manager, e.g. pip:
+```bash
 pip install docling
 ```
 
-**Notes**:
-* Works on macOS and Linux environments. Windows platforms are currently not tested.
+> [!NOTE]  
+> Works on macOS and Linux environments. Windows platforms are currently not tested.
 
 ### Development setup
 
-To develop for `docling`, you need Python 3.11 and `poetry`. Install poetry from [here](https://python-poetry.org/docs/#installing-with-the-official-installer).
-
-Once you have `poetry` installed and cloned this repo, create an environment and install `docling` from the repo root:
-
+To develop for Docling, you need Python 3.11 / 3.12 and Poetry. You can then install from your local clone's root dir:
 ```bash
-poetry env use $(which python3.11)
-poetry shell
 poetry install
 ```
 
@@ -75,25 +79,45 @@ python examples/convert.py
 ```
 The output of the above command will be written to `./scratch`.
 
-### Enable or disable pipeline features
+### Adjust pipeline features
 
-You can control if table structure recognition or OCR should be performed by arguments passed to `DocumentConverter` 
+**Control pipeline options**
+
+You can control if table structure recognition or OCR should be performed by arguments passed to `DocumentConverter`:
 ```python
 doc_converter = DocumentConverter(
     artifacts_path=artifacts_path,
-    pipeline_options=PipelineOptions(do_table_structure=False, # Controls if table structure is recovered. 
-                                     do_ocr=True), # Controls if OCR is applied (ignores programmatic content)
+    pipeline_options=PipelineOptions(
+        do_table_structure=False,  # controls if table structure is recovered 
+        do_ocr=True,  # controls if OCR is applied (ignores programmatic content)
+    ),
 )
 ```
 
-### Impose limits on the document size
+**Control table extraction options**
+
+You can control if table structure recognition should map the recognized structure back to PDF cells (default) or use text cells from the structure prediction itself.
+This can improve output quality if you find that multiple columns in extracted tables are erroneously merged into one.
+
 
-You can limit the file size and number of pages which should be allowed to process per document.
 ```python
-paths = [Path("./test/data/2206.01062.pdf")]
 
-input = DocumentConversionInput.from_paths(
-    paths, limits=DocumentLimits(max_num_pages=100, max_file_size=20971520)
+pipeline_options = PipelineOptions(do_table_structure=True)
+pipeline_options.table_structure_options.do_cell_matching = False # Uses text cells predicted from table structure model
+
+doc_converter = DocumentConverter(
+    artifacts_path=artifacts_path,
+    pipeline_options=pipeline_options,
+)
+```
+
+### Impose limits on the document size
+
+You can limit the file size and number of pages which should be allowed to process per document:
+```python
+conv_input = DocumentConversionInput.from_paths(
+    paths=[Path("./test/data/2206.01062.pdf")],
+    limits=DocumentLimits(max_num_pages=100, max_file_size=20971520)
 )
 ```
 
@@ -103,12 +127,12 @@ You can convert PDFs from a binary stream instead of from the filesystem as foll
 ```python
 buf = BytesIO(your_binary_stream)
 docs = [DocumentStream(filename="my_doc.pdf", stream=buf)]
-input = DocumentConversionInput.from_streams(docs)
-converted_docs = doc_converter.convert(input)
+conv_input = DocumentConversionInput.from_streams(docs)
+converted_docs = doc_converter.convert(conv_input)
 ```
 ### Limit resource usage
 
-You can limit the CPU threads used by `docling` by setting the environment variable `OMP_NUM_THREADS` accordingly. The default setting is using 4 CPU threads.
+You can limit the CPU threads used by Docling by setting the environment variable `OMP_NUM_THREADS` accordingly. The default setting is using 4 CPU threads.
 
 
 ## Contributing
@@ -118,7 +142,7 @@ Please read [Contributing to Docling](https://github.com/DS4SD/docling/blob/main
 
 ## References
 
-If you use `Docling` in your projects, please consider citing the following:
+If you use Docling in your projects, please consider citing the following:
 
 ```bib
 @software{Docling,
@@ -133,6 +157,6 @@ year = {2024}
 
 ## License
 
-The `Docling` codebase is under MIT license.
+The Docling codebase is under MIT license.
 For individual model usage, please refer to the model licenses found in the original packages.
 

docling-0.4.0/README.md

@@ -0,0 +1,129 @@
+<p align="center">
+  <a href="https://github.com/ds4sd/docling"> <img loading="lazy" alt="Docling" src="https://github.com/DS4SD/docling/raw/main/logo.png" width="150" />
+</p>
+
+# Docling
+
+[![PyPI version](https://img.shields.io/pypi/v/docling)](https://pypi.org/project/docling/)
+![Python](https://img.shields.io/badge/python-3.11%20%7C%203.12-blue)
+[![Poetry](https://img.shields.io/endpoint?url=https://python-poetry.org/badge/v0.json)](https://python-poetry.org/)
+[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)
+[![Imports: isort](https://img.shields.io/badge/%20imports-isort-%231674b1?style=flat&labelColor=ef8336)](https://pycqa.github.io/isort/)
+[![Pydantic v2](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/pydantic/pydantic/main/docs/badge/v2.json)](https://pydantic.dev)
+[![pre-commit](https://img.shields.io/badge/pre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white)](https://github.com/pre-commit/pre-commit)
+[![License MIT](https://img.shields.io/github/license/ds4sd/deepsearch-toolkit)](https://opensource.org/licenses/MIT)
+
+Docling bundles PDF document conversion to JSON and Markdown in an easy, self-contained package.
+
+## Features
+* ⚡ Converts any PDF document to JSON or Markdown format, stable and lightning fast
+* 📑 Understands detailed page layout, reading order and recovers table structures
+* 📝 Extracts metadata from the document, such as title, authors, references and language
+* 🔍 Optionally applies OCR (use with scanned PDFs)
+
+## Installation
+
+To use Docling, simply install `docling` from your package manager, e.g. pip:
+```bash
+pip install docling
+```
+
+> [!NOTE]  
+> Works on macOS and Linux environments. Windows platforms are currently not tested.
+
+### Development setup
+
+To develop for Docling, you need Python 3.11 / 3.12 and Poetry. You can then install from your local clone's root dir:
+```bash
+poetry install
+```
+
+## Usage
+
+For basic usage, see the [convert.py](https://github.com/DS4SD/docling/blob/main/examples/convert.py) example module. Run with:
+
+```
+python examples/convert.py
+```
+The output of the above command will be written to `./scratch`.
+
+### Adjust pipeline features
+
+**Control pipeline options**
+
+You can control if table structure recognition or OCR should be performed by arguments passed to `DocumentConverter`:
+```python
+doc_converter = DocumentConverter(
+    artifacts_path=artifacts_path,
+    pipeline_options=PipelineOptions(
+        do_table_structure=False,  # controls if table structure is recovered 
+        do_ocr=True,  # controls if OCR is applied (ignores programmatic content)
+    ),
+)
+```
+
+**Control table extraction options**
+
+You can control if table structure recognition should map the recognized structure back to PDF cells (default) or use text cells from the structure prediction itself.
+This can improve output quality if you find that multiple columns in extracted tables are erroneously merged into one.
+
+
+```python
+
+pipeline_options = PipelineOptions(do_table_structure=True)
+pipeline_options.table_structure_options.do_cell_matching = False # Uses text cells predicted from table structure model
+
+doc_converter = DocumentConverter(
+    artifacts_path=artifacts_path,
+    pipeline_options=pipeline_options,
+)
+```
+
+### Impose limits on the document size
+
+You can limit the file size and number of pages which should be allowed to process per document:
+```python
+conv_input = DocumentConversionInput.from_paths(
+    paths=[Path("./test/data/2206.01062.pdf")],
+    limits=DocumentLimits(max_num_pages=100, max_file_size=20971520)
+)
+```
+
+### Convert from binary PDF streams 
+
+You can convert PDFs from a binary stream instead of from the filesystem as follows:
+```python
+buf = BytesIO(your_binary_stream)
+docs = [DocumentStream(filename="my_doc.pdf", stream=buf)]
+conv_input = DocumentConversionInput.from_streams(docs)
+converted_docs = doc_converter.convert(conv_input)
+```
+### Limit resource usage
+
+You can limit the CPU threads used by Docling by setting the environment variable `OMP_NUM_THREADS` accordingly. The default setting is using 4 CPU threads.
+
+
+## Contributing
+
+Please read [Contributing to Docling](https://github.com/DS4SD/docling/blob/main/CONTRIBUTING.md) for details.
+
+
+## References
+
+If you use Docling in your projects, please consider citing the following:
+
+```bib
+@software{Docling,
+author = {Deep Search Team},
+month = {7},
+title = {{Docling}},
+url = {https://github.com/DS4SD/docling},
+version = {main},
+year = {2024}
+}
+```
+
+## License
+
+The Docling codebase is under MIT license.
+For individual model usage, please refer to the model licenses found in the original packages.

{docling-0.3.0 → docling-0.4.0}/docling/datamodel/base_models.py

@@ -1,3 +1,4 @@
+import copy
 from enum import Enum, auto
 from io import BytesIO
 from typing import Any, Dict, List, Optional, Tuple, Union
@@ -47,6 +48,15 @@ class BoundingBox(BaseModel):
     def height(self):
         return abs(self.t - self.b)
 
+    def scaled(self, scale: float) -> "BoundingBox":
+        out_bbox = copy.deepcopy(self)
+        out_bbox.l *= scale
+        out_bbox.r *= scale
+        out_bbox.t *= scale
+        out_bbox.b *= scale
+
+        return out_bbox
+
     def as_tuple(self):
         if self.coord_origin == CoordOrigin.TOPLEFT:
             return (self.l, self.t, self.r, self.b)
@@ -241,6 +251,17 @@ class DocumentStream(BaseModel):
     stream: BytesIO
 
 
+class TableStructureOptions(BaseModel):
+    do_cell_matching: bool = (
+        True
+        # True:  Matches predictions back to PDF cells. Can break table output if PDF cells
+        #        are merged across table columns.
+        # False: Let table structure model define the text cells, ignore PDF cells.
+    )
+
+
 class PipelineOptions(BaseModel):
-    do_table_structure: bool = True
-    do_ocr: bool = False
+    do_table_structure: bool = True  # True: perform table structure extraction
+    do_ocr: bool = False  # True: perform OCR, replace programmatic PDF text
+
+    table_structure_options: TableStructureOptions = TableStructureOptions()

{docling-0.3.0 → docling-0.4.0}/docling/datamodel/document.py

@@ -117,9 +117,9 @@ class ConvertedDocument(BaseModel):
     errors: List[Dict] = []  # structure to keep errors
 
     pages: List[Page] = []
-    assembled: AssembledUnit = None
+    assembled: Optional[AssembledUnit] = None
 
-    output: DsDocument = None
+    output: Optional[DsDocument] = None
 
     def to_ds_document(self) -> DsDocument:
         title = ""

{docling-0.3.0 → docling-0.4.0}/docling/models/page_assemble_model.py

@@ -19,18 +19,6 @@ class PageAssembleModel:
     def __init__(self, config):
         self.config = config
 
-        # self.line_wrap_pattern = re.compile(r'(?<=[^\W_])- \n(?=\w)')
-
-    # def sanitize_text_poor(self, lines):
-    #     text = '\n'.join(lines)
-    #
-    #     # treat line wraps.
-    #     sanitized_text = self.line_wrap_pattern.sub('', text)
-    #
-    #     sanitized_text = sanitized_text.replace('\n', ' ')
-    #
-    #     return sanitized_text
-
     def sanitize_text(self, lines):
         if len(lines) <= 1:
             return " ".join(lines)

{docling-0.3.0 → docling-0.4.0}/docling/models/table_structure_model.py

@@ -1,7 +1,10 @@
-from typing import Iterable
+import copy
+import random
+from typing import Iterable, List
 
 import numpy
 from docling_ibm_models.tableformer.data_management.tf_predictor import TFPredictor
+from PIL import ImageDraw
 
 from docling.datamodel.base_models import (
     BoundingBox,
@@ -28,6 +31,21 @@ class TableStructureModel:
             self.tm_model_type = self.tm_config["model"]["type"]
 
             self.tf_predictor = TFPredictor(self.tm_config)
+            self.scale = 2.0  # Scale up table input images to 144 dpi
+
+    def draw_table_and_cells(self, page: Page, tbl_list: List[TableElement]):
+        image = page._backend.get_page_image()
+        draw = ImageDraw.Draw(image)
+
+        for table_element in tbl_list:
+            x0, y0, x1, y1 = table_element.cluster.bbox.as_tuple()
+            draw.rectangle([(x0, y0), (x1, y1)], outline="red")
+
+            for tc in table_element.table_cells:
+                x0, y0, x1, y1 = tc.bbox.as_tuple()
+                draw.rectangle([(x0, y0), (x1, y1)], outline="blue")
+
+        image.show()
 
     def __call__(self, page_batch: Iterable[Page]) -> Iterable[Page]:
 
@@ -36,16 +54,17 @@ class TableStructureModel:
             return
 
         for page in page_batch:
+
             page.predictions.tablestructure = TableStructurePrediction()  # dummy
 
             in_tables = [
                 (
                     cluster,
                     [
-                        round(cluster.bbox.l),
-                        round(cluster.bbox.t),
-                        round(cluster.bbox.r),
-                        round(cluster.bbox.b),
+                        round(cluster.bbox.l) * self.scale,
+                        round(cluster.bbox.t) * self.scale,
+                        round(cluster.bbox.r) * self.scale,
+                        round(cluster.bbox.b) * self.scale,
                     ],
                 )
                 for cluster in page.predictions.layout.clusters
@@ -65,20 +84,29 @@ class TableStructureModel:
                         ):
                             # Only allow non empty stings (spaces) into the cells of a table
                             if len(c.text.strip()) > 0:
-                                tokens.append(c.model_dump())
+                                new_cell = copy.deepcopy(c)
+                                new_cell.bbox = new_cell.bbox.scaled(scale=self.scale)
+
+                                tokens.append(new_cell.model_dump())
 
-            iocr_page = {
-                "image": numpy.asarray(page.image),
+            page_input = {
                 "tokens": tokens,
-                "width": page.size.width,
-                "height": page.size.height,
+                "width": page.size.width * self.scale,
+                "height": page.size.height * self.scale,
             }
+            # add image to page input.
+            if self.scale == 1.0:
+                page_input["image"] = numpy.asarray(page.image)
+            else:  # render new page image on the fly at desired scale
+                page_input["image"] = numpy.asarray(
+                    page._backend.get_page_image(scale=self.scale)
+                )
 
             table_clusters, table_bboxes = zip(*in_tables)
 
             if len(table_bboxes):
                 tf_output = self.tf_predictor.multi_table_predict(
-                    iocr_page, table_bboxes, do_matching=self.do_cell_matching
+                    page_input, table_bboxes, do_matching=self.do_cell_matching
                 )
 
                 for table_cluster, table_out in zip(table_clusters, tf_output):
@@ -91,6 +119,7 @@ class TableStructureModel:
                             element["bbox"]["token"] = text_piece
 
                         tc = TableCell.model_validate(element)
+                        tc.bbox = tc.bbox.scaled(1 / self.scale)
                         table_cells.append(tc)
 
                     # Retrieving cols/rows, after post processing:
@@ -111,4 +140,7 @@ class TableStructureModel:
 
                     page.predictions.tablestructure.table_map[table_cluster.id] = tbl
 
+                # For debugging purposes:
+                # self.draw_table_and_cells(page, page.predictions.tablestructure.table_map.values())
+
             yield page

{docling-0.3.0 → docling-0.4.0}/docling/pipeline/standard_model_pipeline.py

@@ -34,7 +34,7 @@ class StandardModelPipeline(BaseModelPipeline):
                     "artifacts_path": artifacts_path
                     / StandardModelPipeline._table_model_path,
                     "enabled": pipeline_options.do_table_structure,
-                    "do_cell_matching": False,
+                    "do_cell_matching": pipeline_options.table_structure_options.do_cell_matching,
                 }
             ),
         ]

{docling-0.3.0 → docling-0.4.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "docling"
-version = "0.3.0"  # DO NOT EDIT, updated automatically
+version = "0.4.0"  # DO NOT EDIT, updated automatically
 description = "Docling PDF conversion package"
 authors = ["Christoph Auer <cau@zurich.ibm.com>", "Michele Dolfi <dol@zurich.ibm.com>", "Maxim Lysak <mly@zurich.ibm.com>", "Nikos Livathinos <nli@zurich.ibm.com>", "Ahmed Nassar <ahn@zurich.ibm.com>", "Peter Staar <taa@zurich.ibm.com>"]
 license = "MIT"

docling-0.3.0/README.md

@@ -1,105 +0,0 @@
-<p align="center">
-  <a href="https://github.com/ds4sd/docling"> <img loading="lazy" alt="Docling" src="https://github.com/DS4SD/docling/raw/main/logo.png" width="150" /> </a>
-</p>
-
-# Docling
-
-Docling bundles PDF document conversion to JSON and Markdown in an easy, self-contained package.
-
-## Features
-* ⚡ Converts any PDF document to JSON or Markdown format, stable and lightning fast
-* 📑 Understands detailed page layout, reading order and recovers table structures
-* 📝 Extracts metadata from the document, such as title, authors, references and language
-* 🔍 Optionally applies OCR (use with scanned PDFs)
-
-## Setup
-
-For general usage, you can simply install `docling` through `pip` from the pypi package index.
-```
-pip install docling
-```
-
-**Notes**:
-* Works on macOS and Linux environments. Windows platforms are currently not tested.
-
-### Development setup
-
-To develop for `docling`, you need Python 3.11 and `poetry`. Install poetry from [here](https://python-poetry.org/docs/#installing-with-the-official-installer).
-
-Once you have `poetry` installed and cloned this repo, create an environment and install `docling` from the repo root:
-
-```bash
-poetry env use $(which python3.11)
-poetry shell
-poetry install
-```
-
-## Usage
-
-For basic usage, see the [convert.py](https://github.com/DS4SD/docling/blob/main/examples/convert.py) example module. Run with:
-
-```
-python examples/convert.py
-```
-The output of the above command will be written to `./scratch`.
-
-### Enable or disable pipeline features
-
-You can control if table structure recognition or OCR should be performed by arguments passed to `DocumentConverter` 
-```python
-doc_converter = DocumentConverter(
-    artifacts_path=artifacts_path,
-    pipeline_options=PipelineOptions(do_table_structure=False, # Controls if table structure is recovered. 
-                                     do_ocr=True), # Controls if OCR is applied (ignores programmatic content)
-)
-```
-
-### Impose limits on the document size
-
-You can limit the file size and number of pages which should be allowed to process per document.
-```python
-paths = [Path("./test/data/2206.01062.pdf")]
-
-input = DocumentConversionInput.from_paths(
-    paths, limits=DocumentLimits(max_num_pages=100, max_file_size=20971520)
-)
-```
-
-### Convert from binary PDF streams 
-
-You can convert PDFs from a binary stream instead of from the filesystem as follows:
-```python
-buf = BytesIO(your_binary_stream)
-docs = [DocumentStream(filename="my_doc.pdf", stream=buf)]
-input = DocumentConversionInput.from_streams(docs)
-converted_docs = doc_converter.convert(input)
-```
-### Limit resource usage
-
-You can limit the CPU threads used by `docling` by setting the environment variable `OMP_NUM_THREADS` accordingly. The default setting is using 4 CPU threads.
-
-
-## Contributing
-
-Please read [Contributing to Docling](https://github.com/DS4SD/docling/blob/main/CONTRIBUTING.md) for details.
-
-
-## References
-
-If you use `Docling` in your projects, please consider citing the following:
-
-```bib
-@software{Docling,
-author = {Deep Search Team},
-month = {7},
-title = {{Docling}},
-url = {https://github.com/DS4SD/docling},
-version = {main},
-year = {2024}
-}
-```
-
-## License
-
-The `Docling` codebase is under MIT license.
-For individual model usage, please refer to the model licenses found in the original packages.