fmtr.tools 1.0.31__tar.gz → 1.0.33__tar.gz

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: fmtr.tools
-Version: 1.0.31
+Version: 1.0.33
 Summary: Collection of high-level tools to simplify everyday development tasks, with a focus on AI/ML
 Home-page: https://github.com/fmtr/fmtr.tools
 Author: Frontmatter
@@ -25,6 +25,7 @@ Provides-Extra: metric
 Provides-Extra: netrc
 Provides-Extra: openai.api
 Provides-Extra: parallel
+Provides-Extra: pdf
 Provides-Extra: process
 Provides-Extra: profiling
 Provides-Extra: semantic

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr/tools/__init__.py

@@ -143,6 +143,11 @@ try:
 except ImportError as exception:
     caching = MissingExtraMockModule('caching', exception)
 
+try:
+    from fmtr.tools import pdf_tools as pdf
+except ImportError as exception:
+    pdf = MissingExtraMockModule('pdf', exception)
+
 
 __all__ = [
     'config',

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr/tools/api_tools.py

@@ -83,6 +83,11 @@ class ApiBase:
 
     @classmethod
     def launch(cls):
+        """
+
+        Initialise self and launch.
+
+        """
         self = cls()
         logger.info(f'Launching API {cls.TITLE}...')
         uvicorn.run(self.app, host=self.HOST, port=self.PORT)

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr/tools/path_tools.py

@@ -96,6 +96,29 @@ class Path(type(Path())):
         """
         return cls(gettempdir())
 
+    @classmethod
+    def data(cls, name='data') -> 'Path':
+        """
+
+        Fetch canonical "data"/"artifacts" path, whether calling package is regular or namespace package.
+
+        """
+        from fmtr.tools.inspection_tools import get_call_path
+        path = get_call_path()
+        path = path.absolute().parent.parent
+
+        path /= name
+
+        if path.exists():
+            return path
+
+        path = path.parent.parent / name
+
+        if path.exists():
+            return path
+
+        raise FileNotFoundError(f'No "{name}" directory found at "{path}"')
+
     def write_json(self, obj) -> int:
         """
 

fmtr.tools-1.0.33/fmtr/tools/pdf_tools.py

@@ -0,0 +1,172 @@
+import pymupdf as pm
+import pymupdf4llm
+from typing import List, Tuple, Dict, Any, Self
+
+from fmtr.tools import data_modelling_tools
+
+
+class BoundingBox(data_modelling_tools.Base):
+    left: float
+    top: float
+    right: float
+    bottom: float
+
+    @property
+    def order(self):
+        """
+
+        Approximate natural reading order
+
+        """
+        return (self.top, self.left), (self.bottom, self.right)
+
+    @property
+    def rect(self) -> pm.Rect:
+        """
+
+        Position as a PyMuPDF Rect
+
+        """
+        return pm.Rect(self.left, self.top, self.right, self.bottom)
+
+    @classmethod
+    def from_dict(cls, data: Tuple[float]) -> Self:
+        """
+
+        Instantiate from PyMuPDF dictionary data
+
+        """
+        data = {key: value for key, value in zip(cls.model_fields.keys(), data)}
+        return cls(**data)
+
+
+class Span(data_modelling_tools.Base):
+    size: float
+    flags: int
+    font: str
+    color: int
+    ascender: float
+    descender: float
+    text: str
+    origin: Tuple[float, float]
+    bbox: BoundingBox
+
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> Self:
+        data['bbox'] = BoundingBox.from_dict(data['bbox'])
+        return cls(**data)
+
+
+class Line(data_modelling_tools.Base):
+    spans: List[Span]
+    wmode: int
+    dir: Tuple[float, float]
+    bbox: BoundingBox
+
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> 'Line':
+        """
+
+        Instantiate from PyMuPDF dictionary data
+
+        """
+        data['spans'] = [Span.from_dict(span) for span in data['spans']]
+        data['bbox'] = BoundingBox.from_dict(data['bbox'])
+        return cls(**data)
+
+    @property
+    def text(self) -> str:
+        """
+
+        Simple text representation
+
+        """
+        return ' '.join([span.text for span in self.spans])
+
+
+class Block(data_modelling_tools.Base):
+    number: int
+    type: int
+    bbox: BoundingBox
+    lines: List[Line]
+
+    @property
+    def text(self) -> str:
+        """
+
+        Simple text representation
+
+        """
+        return ' '.join([span.text for span in self.spans])
+
+    @classmethod
+    def from_dict(cls, data: Dict) -> Self:
+        """
+
+        Instantiate from PyMuPDF dictionary data
+
+        """
+        data['lines'] = [Line.from_dict(line) for line in data['lines']]
+        data['bbox'] = BoundingBox.from_dict(data['bbox'])
+        return cls(**data)
+
+    @property
+    def rect(self) -> pm.Rect:
+        """
+
+        Position as a PyMuPDF Rect
+
+        """
+        return self.bbox.rect
+
+
+class Document(pm.Document):
+    """
+
+    Subclassed Document object with data-modelled elements property and markdown conversion.
+
+    """
+
+    @property
+    def data(self) -> List[Block]:
+        """
+
+        Get representation of Document elements as Python objects.
+
+        """
+
+        blocks = []
+
+        for page in self:
+            for block in page.get_text("dict")["blocks"]:
+                obj = Block.from_dict(block)
+                blocks.append(obj)
+
+        return blocks
+
+    def to_markdown(self, **kwargs) -> str:
+        """
+
+        Markdown output via `pymupdf4llm`
+
+        """
+        return pymupdf4llm.to_markdown(self, **kwargs)
+
+
+if __name__ == '__main__':
+    from fmtr.tools.path_tools import Path
+
+    PATH_DATA = Path.data()
+    # PATH_PDF=PATH_DATA/'chib.pdf'
+    PATH_PDF = PATH_DATA / 'kvm.pdf'
+    assert PATH_PDF.exists()
+
+    doc = Document(PATH_PDF)
+    doc.data
+
+    for page in doc:
+        print(page.get_text('dict'))
+        print(page.get_text('html'))
+
+    md = doc.to_markdown()
+    md

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr/tools/string_tools.py

@@ -145,3 +145,10 @@ class Mask:
             return text
         except (KeyError, IndexError):
             return self
+
+
+if __name__ == '__main__':
+    import numpy as np
+
+    st = join([1, None, 'test', np.nan, 0, '', 'yeah'])
+    st

fmtr.tools-1.0.33/fmtr/tools/version

@@ -0,0 +1 @@
+1.0.33

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr.tools.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: fmtr.tools
-Version: 1.0.31
+Version: 1.0.33
 Summary: Collection of high-level tools to simplify everyday development tasks, with a focus on AI/ML
 Home-page: https://github.com/fmtr/fmtr.tools
 Author: Frontmatter
@@ -25,6 +25,7 @@ Provides-Extra: metric
 Provides-Extra: netrc
 Provides-Extra: openai.api
 Provides-Extra: parallel
+Provides-Extra: pdf
 Provides-Extra: process
 Provides-Extra: profiling
 Provides-Extra: semantic

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr.tools.egg-info/SOURCES.txt

@@ -34,6 +34,7 @@ setup.py
 ./fmtr/tools/openai_tools.py
 ./fmtr/tools/parallel_tools.py
 ./fmtr/tools/path_tools.py
+./fmtr/tools/pdf_tools.py
 ./fmtr/tools/platform_tools.py
 ./fmtr/tools/process_tools.py
 ./fmtr/tools/profiling_tools.py

{fmtr.tools-1.0.31 → fmtr.tools-1.0.33}/fmtr.tools.egg-info/requires.txt

@@ -76,6 +76,11 @@ bokeh
 dask[bag]
 distributed
 
+[pdf]
+pydantic
+pymupdf
+pymupdf4llm
+
 [process]
 logfire
 semver
@@ -118,6 +123,8 @@ pandas
 peft
 pydantic
 pydantic-ai[logfire,openai]
+pymupdf
+pymupdf4llm
 pytest-cov
 pyyaml
 semver

fmtr.tools-1.0.31/fmtr/tools/version

@@ -1 +0,0 @@
-1.0.31