playa-pdf 0.8.0__tar.gz → 0.9.0__tar.gz

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/.gitignore

@@ -29,3 +29,4 @@ poetry.lock
 /playa/_version.py
 .venv
 *.so
+.coverage

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/CHANGELOG.md

@@ -1,10 +1,21 @@
+## PLAYA 0.9.0: Unreleased
+
+- Refactor and add convenience methods to text objects
+- Insert blank pages for missing object references in page tree
+- Clean up type annotations (breaking change: PDFObject can no longer
+  be `str`, as the parser will never create this)
+
+## PLAYA 0.8.1: 2025-12-22
+
+- Correct subtle issues with mypyc-compiled pdfminer.six code
+
 ## PLAYA 0.8.0: 2025-12-17
 
 - Optionally accelerate image decoding with mypyc
 - Correct explicit string positioning in vertical text
 - Restore caching in text decoding under Python 3.8
 - Bring back pdfminer.six layout analysis algorithm
-- TODO: refactor and accelerate text extraction with mypyc
+- Optionally accelerate pdfminer.six compatibility with mypyc
 
 ## PLAYA 0.7.2: 2025-11-09
 

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: playa-pdf
-Version: 0.8.0
+Version: 0.9.0
 Summary: Parallel and LazY Analyzer for PDFs
 Project-URL: Homepage, https://dhdaines.github.io/playa
 Author-email: David Huggins-Daines <dhd@ecolingui.ca>
@@ -52,7 +52,7 @@ analysis](https://pdfminersix.readthedocs.io/en/latest/topic/converting_pdf_to_t
 algorithm from
 [pdfminer.six](https://github.com/pdfminer/pdfminer.six) anyways.  See
 [the
-documentation](https://dhdaines.github.io/playa/working-in-the-pdf-mine)
+documentation](https://dhdaines.github.io/playa/latest/working-in-the-pdf-mine)
 for more information on how to migrate your code.  You may be
 interested to know that PLAYA's implementation is also 15-50% faster,
 depending on how many CPUs you use.

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/README.md

@@ -21,7 +21,7 @@ analysis](https://pdfminersix.readthedocs.io/en/latest/topic/converting_pdf_to_t
 algorithm from
 [pdfminer.six](https://github.com/pdfminer/pdfminer.six) anyways.  See
 [the
-documentation](https://dhdaines.github.io/playa/working-in-the-pdf-mine)
+documentation](https://dhdaines.github.io/playa/latest/working-in-the-pdf-mine)
 for more information on how to migrate your code.  You may be
 interested to know that PLAYA's implementation is also 15-50% faster,
 depending on how many CPUs you use.

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.8.0'
-__version_tuple__ = version_tuple = (0, 8, 0)
+__version__ = version = '0.9.0'
+__version_tuple__ = version_tuple = (0, 9, 0)
 
 __commit_id__ = commit_id = None

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/ccitt.py

@@ -492,7 +492,8 @@ class CCITTG4Parser(BitParser):
         print(y, "".join(str(b) for b in bits))
 
     def _reset_line(self) -> None:
-        # FIXME: probably, we could just swap them, like in PNG prediction
+        # We could just swap them, like in PNG prediction, though it's
+        # not clear that would be much faster.
         self._refline = self._curline
         self._curline = [1] * self.width
         self._curpos = -1

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/cli.py

@@ -105,7 +105,6 @@ from playa.pdftypes import (
 from playa.structure import ContentItem
 from playa.structure import ContentObject as StructContentObject
 from playa.structure import Element
-from playa.utils import decode_text
 
 LOG = logging.getLogger(__name__)
 
@@ -486,7 +485,7 @@ def _extract_outline_item(item: Outline, indent: int, outfh: TextIO) -> bool:
 
     print(f"{ws}{{", file=outfh)
     if item.title is not None:
-        format_attr("title", decode_text(item.title))
+        format_attr("title", item.title)
     if item.destination is not None:
         format_attr("destination", asobj(item.destination))
     if s:

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/content.py

@@ -4,6 +4,7 @@ PDF content objects created by the interpreter.
 
 import itertools
 import logging
+from abc import abstractmethod
 from copy import copy
 from dataclasses import dataclass
 from typing import (
@@ -800,30 +801,95 @@ class PathObject(ContentObject):
         return transform_bbox(self.ctm, bbox)
 
 
-def _font_size(matrix: Matrix, vert: bool = False) -> float:
-    if vert:
-        # dx, dy = apply_matrix_norm(self.matrix, (1, 0))
-        dx, dy, _, _, _, _ = matrix
-    else:
-        # dx, dy = apply_matrix_norm(self.matrix, (0, 1))
-        _, _, dx, dy, _, _ = matrix
-    if dx == 0:  # Nearly always true
-        return abs(dy)
-    elif dy == 0:
-        return abs(dx)
-    else:
-        import math
+class TextBase(ContentObject):
+    """Common properties for text and glyph objects."""
 
-        return math.sqrt(dx * dx + dy * dy)
+    @property
+    @abstractmethod
+    def matrix(self) -> Matrix: ...
+
+    @property
+    def font(self) -> Font:
+        """Font for this text object."""
+        font = self.gstate.font
+        assert font is not None
+        return font
+
+    @property
+    def size(self) -> float:
+        """Font size for this text object.
+
+        This is the actual font size in device space, which is **not**
+        the same as `GraphicState.fontsize`.  That's the font size in
+        text space which is not a very useful number (it's usually 1).
+        """
+        vert = False if self.gstate.font is None else self.gstate.font.vertical
+        if vert:
+            # dx, dy = apply_matrix_norm(self.matrix, (1, 0))
+            dx, dy, _, _, _, _ = self.matrix
+        else:
+            # dx, dy = apply_matrix_norm(self.matrix, (0, 1))
+            _, _, dx, dy, _, _ = self.matrix
+        if dx == 0:  # Nearly always true
+            return abs(dy)
+        elif dy == 0:
+            return abs(dx)
+        else:
+            import math
+
+            return math.sqrt(dx * dx + dy * dy)
+
+    @property
+    def fontname(self) -> str:
+        """Font name for this text object"""
+        return self.font.fontname
+
+    @property
+    def fontbase(self) -> str:
+        """Original font name for this text object.
+
+        Fonts in PDF files are usually "subsetted", meaning only the
+        glyphs actually used in the document are included.  In this
+        case the font's `fontname` property usually consists of an
+        arbitrary "tag", plus (literally, a `+`) and the original
+        name.  This is a convenience property to get that original
+        name.
+
+        This is not the same as `GraphicState.font.basefont` which
+        usually also includes the subset tag.
+
+        """
+        fontname = self.fontname
+        subset, _, base = fontname.partition("+")
+        if base:
+            return base
+        return fontname
+
+    @property
+    def textfont(self) -> str:
+        """Convenient short form of the font name and size.
+
+        For example, "Helvetica 12".
+        """
+        return f"{self.fontbase} {round(self.size)}"
+
+    @property
+    def origin(self) -> Point:
+        """Origin of this text object in device space."""
+        _, _, _, _, dx, dy = self.matrix
+        return dx, dy
 
 
 @dataclass
-class GlyphObject(ContentObject):
+class GlyphObject(TextBase):
     """Individual glyph on the page.
 
     Attributes:
       font: Font for this glyph.
       size: Effective font size for this glyph.
+      fontname: Font name.
+      fontbase: Short (non-subset) font name.
+      textfont: Combined short name and size for the font.
       cid: Character ID for this glyph.
       text: Unicode mapping of this glyph, if any.
       matrix: Rendering matrix `T_rm` for this glyph, which transforms
@@ -837,7 +903,7 @@ class GlyphObject(ContentObject):
 
     cid: int
     text: Union[str, None]
-    matrix: Matrix
+    _matrix: Matrix
     _displacement: float
     _corners: bool
 
@@ -881,20 +947,12 @@ class GlyphObject(ContentObject):
         return itor
 
     @property
-    def font(self) -> Font:
-        font = self.gstate.font
-        assert font is not None
-        return font
-
-    @property
-    def size(self) -> float:
-        vert = False if self.gstate.font is None else self.gstate.font.vertical
-        return _font_size(self.matrix, vert)
+    def matrix(self) -> Matrix:
+        return self._matrix
 
     @property
-    def origin(self) -> Point:
-        _, _, _, _, dx, dy = self.matrix
-        return dx, dy
+    def chars(self) -> str:
+        return self.text or ""
 
     @property
     def displacement(self) -> Point:
@@ -932,7 +990,7 @@ class GlyphObject(ContentObject):
 
 
 @dataclass
-class TextObject(ContentObject):
+class TextObject(TextBase):
     """Text object (contains one or more glyphs).
 
     Attributes:
@@ -943,7 +1001,11 @@ class TextObject(ContentObject):
       origin: Origin of this text object in device space.
       displacement: Vector to the origin of the next text object in
                     device space.
+      font: Font for this text object.
       size: Effective font size for this text object.
+      fontname: Font name.
+      fontbase: Short (non-subset) font name.
+      textfont: Combined short name and size for the font.
       text_matrix: Text matrix `T_m` for this text object, which
                    transforms text space coordinates to user space.
       line_matrix: Text line matrix `T_lm` for this text object, which
@@ -966,6 +1028,7 @@ class TextObject(ContentObject):
     _chars: Union[List[str], None] = None
     _bbox: Union[Rect, None] = None
     _next_glyph_offset: Union[Point, None] = None
+    _displacement: Union[Point, None] = None
 
     def __iter__(self) -> Iterator[GlyphObject]:
         """Generate glyphs for this text object"""
@@ -1048,7 +1111,7 @@ class TextObject(ContentObject):
                         mcstack=self.mcstack,
                         cid=cid,
                         text=text,
-                        matrix=matrix,
+                        _matrix=matrix,
                         _displacement=disp,
                         _corners=corners,
                     )
@@ -1091,11 +1154,6 @@ class TextObject(ContentObject):
         )
         return self._matrix
 
-    @property
-    def size(self) -> float:
-        vert = False if self.gstate.font is None else self.gstate.font.vertical
-        return _font_size(self.matrix, vert)
-
     @property
     def scaling_matrix(self) -> Matrix:
         horizontal_scaling = self.gstate.scaling * 0.01
@@ -1113,15 +1171,11 @@ class TextObject(ContentObject):
     def text_matrix(self) -> Matrix:
         return translate_matrix(self.line_matrix, self._glyph_offset)
 
-    @property
-    def origin(self) -> Point:
-        _, _, _, _, dx, dy = self.matrix
-        return dx, dy
-
     @property
     def displacement(self) -> Point:
+        if self._displacement is not None:
+            return self._displacement
         matrix = self.matrix
-        # FIXME: This should be either cached or optimized
         next_matrix = mult_matrix(
             self.scaling_matrix,
             mult_matrix(
@@ -1129,7 +1183,8 @@ class TextObject(ContentObject):
                 self.ctm,
             ),
         )
-        return next_matrix[-2] - matrix[-2], next_matrix[-1] - matrix[-1]
+        self._displacement = next_matrix[-2] - matrix[-2], next_matrix[-1] - matrix[-1]
+        return self._displacement
 
     @property
     def bbox(self) -> Rect:

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/data/metadata.py

@@ -39,7 +39,7 @@ from playa.structure import ContentItem as _StructContentItem
 from playa.structure import ContentObject as _StructContentObject
 from playa.structure import Element as _Element
 from playa.structure import Tree as _Tree
-from playa.utils import Matrix, Rect, decode_text
+from playa.utils import Matrix, Rect
 
 log = logging.getLogger(__name__)
 
@@ -556,7 +556,7 @@ def asobj_stream(obj: _ContentStream) -> Dict:
 def asobj_outline(obj: _Outline, recurse: bool = True) -> Outline:
     out = Outline()
     if obj.title is not None:
-        out["title"] = decode_text(obj.title)
+        out["title"] = obj.title
     if obj.destination is not None:
         out["destination"] = asobj(obj.destination)
     if recurse:

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/document.py

@@ -447,7 +447,7 @@ class Document:
             if m is None:
                 raise PDFSyntaxError(
                     f"Not an indirect object at position {pos}: "
-                    f"{self.buffer[pos:pos+8]!r}"
+                    f"{self.buffer[pos : pos + 8]!r}"
                 )
             _, obj = next(self.parser)
             if obj.objid != objid:
@@ -695,8 +695,9 @@ class Document:
             try:
                 page_object = dict_value(self[object_id])
             except IndexError as e:
-                log.warning("Skipping missing page object: %s", e)
-                continue
+                log.warning("Missing page object: %s", e)
+                # Create an empty page to match what pdfium does
+                page_object = {"Type": LIT("Page")}
 
             # Avoid recursion errors by keeping track of visited nodes
             # (again, this should never actually happen in a valid PDF)

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/font.py

@@ -88,7 +88,7 @@ class Font:
         fontname = resolve1(descriptor.get("FontName"))
         if isinstance(fontname, PSLiteral):
             self.fontname = literal_name(fontname)
-        elif isinstance(fontname, (bytes, str)):
+        elif isinstance(fontname, bytes):
             self.fontname = decode_text(fontname)
         else:
             self.fontname = "unknown"
@@ -532,16 +532,16 @@ class CIDFont(Font):
         # These are *supposed* to be ASCII (PDF 1.7 section 9.7.3),
         # but for whatever reason they are sometimes UTF-16BE
         cid_registry = resolve1(self.cidsysteminfo.get("Registry"))
-        if isinstance(cid_registry, (str, bytes)):
-            cid_registry = decode_text(cid_registry)
+        if isinstance(cid_registry, bytes):
+            regstr = decode_text(cid_registry).strip()
         else:
-            cid_registry = "unknown"
+            regstr = "unknown"
         cid_ordering = resolve1(self.cidsysteminfo.get("Ordering"))
-        if isinstance(cid_ordering, (str, bytes)):
-            cid_ordering = decode_text(cid_ordering)
+        if isinstance(cid_ordering, bytes):
+            ordstr = decode_text(cid_ordering).strip()
         else:
-            cid_ordering = "unknown"
-        self.cidcoding = f"{cid_registry.strip()}-{cid_ordering.strip()}"
+            ordstr = "unknown"
+        self.cidcoding = f"{regstr}-{ordstr}"
         self.cmap: CMapBase = self.get_cmap_from_spec(spec)
 
         try:

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/image.py

@@ -337,7 +337,7 @@ def write_cmyk_tiff(
 
     # 6. --- Write the Actual Pixel Data ---
     # The current file position should now match `offset_image_data`
-    assert (
-        outfh.tell() == offset_image_data
-    ), f"File position mismatch: at {outfh.tell()}, expected {offset_image_data}"
+    assert outfh.tell() == offset_image_data, (
+        f"File position mismatch: at {outfh.tell()}, expected {offset_image_data}"
+    )
     outfh.write(data)

{playa_pdf-0.8.0 → playa_pdf-0.9.0}/playa/interp.py

@@ -19,7 +19,6 @@ from typing import (
     Tuple,
     Union,
     Sequence,
-    cast,
 )
 
 from playa.color import PREDEFINED_COLORSPACE, ColorSpace, get_colorspace
@@ -58,7 +57,7 @@ from playa.pdftypes import (
     resolve1,
     stream_value,
 )
-from playa.utils import decode_text, mult_matrix
+from playa.utils import mult_matrix
 from playa.worker import _deref_document
 
 if TYPE_CHECKING:
@@ -440,7 +439,7 @@ class LazyInterpreter:
             # Inline images are not XObjects, have no xobjid
             return self.render_image(None, obj)
         else:
-            # FIXME: Do... something?
+            log.warning("EI has unknown argument type: %r", obj)
             return None
 
     def do_Do(self, xobjid_arg: PDFObject) -> Union[ContentObject, None]:
@@ -452,8 +451,7 @@ class LazyInterpreter:
             log.debug("Undefined xobject id: %r", xobjid)
             return None
         except TypeError as e:
-            log.debug("Empty or invalid xobject with id %r: %s", xobjid, e)
-            return None
+            raise TypeError(f"Empty or invalid xobject with id {xobjid!r}") from e
         subtype = xobj.get("Subtype")
         if subtype is LITERAL_FORM:
             # PDF Ref 1.7, # 4.9
@@ -530,7 +528,15 @@ class LazyInterpreter:
         f1: PDFObject,
     ) -> None:
         """Concatenate matrix to current transformation matrix"""
-        self.ctm = mult_matrix(cast(Matrix, (a1, b1, c1, d1, e1, f1)), self.ctm)
+        cm = (
+            num_value(a1),
+            num_value(b1),
+            num_value(c1),
+            num_value(d1),
+            num_value(e1),
+            num_value(f1),
+        )
+        self.ctm = mult_matrix(cm, self.ctm)
 
     def do_w(self, linewidth: PDFObject) -> None:
         """Set line width"""
@@ -557,8 +563,11 @@ class LazyInterpreter:
         """Set color rendering intent"""
         if self.ignore_colours:
             return
-        # FIXME: Should actually be a (runtime checked) enum
-        self.graphicstate.intent = cast(PSLiteral, intent)
+        if isinstance(intent, PSLiteral):
+            # Should possibly check that it is a valid intent
+            self.graphicstate.intent = intent
+        else:
+            raise TypeError(f"Not a name: {intent!r}")
 
     def do_i(self, flatness: PDFObject) -> None:
         """Set flatness tolerance"""
@@ -600,7 +609,12 @@ class LazyInterpreter:
             if isinstance(bm, PSLiteral):
                 self.graphicstate.blend_mode = bm
             else:
-                self.graphicstate.blend_mode = cast(List[PSLiteral], list_value(bm))
+                bml: List[PSLiteral] = []
+                for x in list_value(bm):
+                    if isinstance(PSLiteral, x):
+                        raise TypeError(f"Not a name: {x!r}")
+                    bml.append(x)
+                self.graphicstate.blend_mode = bml
         if "SMask" in extgstate:
             smask = extgstate["SMask"]
             if isinstance(smask, PSLiteral):
@@ -883,8 +897,8 @@ class LazyInterpreter:
             e_new = tx * a + ty * c + e
             f_new = tx * b + ty * d + f
             self.textstate.line_matrix = (a, b, c, d, e_new, f_new)
-        except TypeError:
-            log.warning("Invalid offset (%r, %r) for Td", tx, ty)
+        except TypeError as e:
+            raise TypeError(f"Invalid offset ({tx!r}, {ty!r})") from e
         self.textstate.glyph_offset = (0, 0)
 
     def do_TD(self, tx: PDFObject, ty: PDFObject) -> None:
@@ -969,12 +983,14 @@ class LazyInterpreter:
     def begin_tag(self, tag: PDFObject, props: Dict[str, PDFObject]) -> None:
         """Handle beginning of tag, setting current MCID if any."""
         assert isinstance(tag, PSLiteral)
-        tag = decode_text(tag.name)
         if "MCID" in props:
             mcid = int_value(props["MCID"])
         else:
             mcid = None
-        self.mcstack = (*self.mcstack, MarkedContent(mcid=mcid, tag=tag, props=props))
+        self.mcstack = (
+            *self.mcstack,
+            MarkedContent(mcid=mcid, tag=tag.name, props=props),
+        )
 
     def do_BMC(self, tag: PDFObject) -> None:
         """Begin marked-content sequence"""