Color and font information for chars, words and boxes (#39)

The information originates in chars, but is pushed to words, lines and boxes in case the values don't differ.

Author: kreuzberger

libpdf/models/horizontal_box.py

@@ -19,6 +19,8 @@ class Char:  # pylint: disable=too-few-public-methods # simplicity is good.
     :ivar y1: distance from the bottom of the page to the upper edge of the character
         (greater than y0)
     :vartype y1: float
+    :ivar ncolor: non-stroking-color as rgb value
+    :vartype ncolor: Tuple[float, float, float]
     """
 
     def __init__(
@@ -28,13 +30,17 @@ def __init__(
         y0: float | None = None,
         x1: float | None = None,
         y1: float | None = None,
+        ncolor: tuple | None = None,
+        fontname: str | None = None,
     ):
         """Init with plain char of a character and its rectangular coordinates."""
         self.x0 = x0
         self.y0 = y0
         self.x1 = x1
         self.y1 = y1
         self.text = text
+        self.ncolor = ncolor
+        self.fontname = fontname
 
     def __repr__(self) -> str:
         """Make the text part of the repr for better debugging."""
@@ -65,13 +71,24 @@ def __init__(
         self.x1 = x1
         self.y1 = y1
         self.chars = chars
+        self.ncolor = None
+        self.fontname = None
+
         if self.chars:
             # Obtain the rectangle coordinates from a list of libpdf text objects
             self.x0 = min(text_obj.x0 for text_obj in self.chars)
             self.y0 = min(text_obj.y0 for text_obj in self.chars)
             self.x1 = max(text_obj.x1 for text_obj in self.chars)
             self.y1 = max(text_obj.y1 for text_obj in self.chars)
 
+            for n in ["ncolor", "fontname"]:
+                if all(
+                    getattr(x, n) == getattr(self.chars[0], n)
+                    and getattr(x, n) is not None
+                    for x in self.chars
+                ):
+                    setattr(self, n, getattr(self.chars[0], n))
+
     @property
     def text(self) -> str:
         """Return plain text."""
@@ -106,13 +123,24 @@ def __init__(
         self.x1 = x1
         self.y1 = y1
         self.words = words
+        self.ncolor = None
+        self.fontname = None
+
         if self.words:
             # Obtain the rectangle coordinates from a list of libpdf text objects
             self.x0 = min(text_obj.x0 for text_obj in self.words)
             self.y0 = min(text_obj.y0 for text_obj in self.words)
             self.x1 = max(text_obj.x1 for text_obj in self.words)
             self.y1 = max(text_obj.y1 for text_obj in self.words)
 
+            for n in ["ncolor", "fontname"]:
+                if all(
+                    getattr(x, n) == getattr(self.words[0], n)
+                    and getattr(x, n) is not None
+                    for x in self.words
+                ):
+                    setattr(self, n, getattr(self.words[0], n))
+
     @property
     def text(self) -> str:
         """Return plain text."""
@@ -147,18 +175,35 @@ def __init__(
         self.x1 = x1
         self.y1 = y1
         self.lines = lines
+        self.ncolor = None
+        self.fontname = None
+
         if self.lines:
             # Obtain the rectangle coordinates from a list of libpdf text objects.
             self.x0 = min(text_obj.x0 for text_obj in self.lines)
             self.y0 = min(text_obj.y0 for text_obj in self.lines)
             self.x1 = max(text_obj.x1 for text_obj in self.lines)
             self.y1 = max(text_obj.y1 for text_obj in self.lines)
 
+            _words = [word for line in self.lines for word in line.words]
+
+            for n in ["ncolor", "fontname"]:
+                if all(
+                    getattr(x, n) == getattr(_words[0], n) and getattr(x, n) is not None
+                    for x in _words
+                ):
+                    setattr(self, n, getattr(_words[0], n))
+
     @property
     def text(self) -> str:
         """Return plain text."""
         return "\n".join([x.text for x in self.lines])
 
+    @property
+    def words(self) -> list[str]:
+        """Return list of words."""
+        return [word for line in self.lines for word in line.words]
+
     def __repr__(self) -> str | None:
         """Make the text part of the repr for better debugging."""
         if self.lines:

libpdf/utils.py

@@ -488,7 +488,15 @@ def assemble_to_textlines(
     for lt_obj in flatten_lt_objs:
         if lt_obj.get_text() != " " and lt_obj.get_text() != "\n":
             # instantiate Char
-            char = Char(lt_obj.get_text(), lt_obj.x0, lt_obj.y0, lt_obj.x1, lt_obj.y1)
+            char = Char(
+                lt_obj.get_text(),
+                lt_obj.x0,
+                lt_obj.y0,
+                lt_obj.x1,
+                lt_obj.y1,
+                lt_obj.graphicstate.ncolor if hasattr(lt_obj, "graphicstate") else None,
+                lt_obj.fontname,
+            )
             chars.append(char)
 
             if lt_obj is flatten_lt_objs[-1]:

tests/conftest.py

@@ -34,6 +34,9 @@
 # test PDF for rect extraction generateby by sphinx-simplepdf
 PDF_RECTS_EXTRACTION = Path(__file__).parent / "pdf" / "test_rects_extraction.pdf"
 
+# test PDF for color style info
+PDF_COLOR_STYLE = Path(__file__).parent / "pdf" / "test_words_color_style.pdf"
+
 
 @pytest.fixture(scope="session")
 def load_full_features_pdf(

tests/pdf/test_words_color_style.odt

@@ -0,0 +1,143 @@
+"""Test catalog extraction."""
+
+import libpdf
+from tests.conftest import PDF_COLOR_STYLE
+
+
+def test_colors_0() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if chapter.title == "Color in Text and Heading":
+            assert chapter.textbox.ncolor == (1, 0, 0)
+
+
+def test_colors_1() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if chapter.title == "HorizontalLine":
+            for content in chapter.content:
+                if (
+                    content.type == "paragraph"
+                    and "Paragraph text is blue" in content.textbox.text
+                ):
+                    assert content.textbox.ncolor == (0, 0, 1)
+                if (
+                    content.type == "paragraph"
+                    and "This chapter is for" in content.textbox.text
+                ):
+                    assert content.textbox.ncolor == (0, 0, 0)
+
+
+def test_colors_2() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if chapter.title == "HorizontalBox":
+            for content in chapter.content:
+                if content.type == "paragraph":
+                    assert content.textbox.ncolor == (0, 1, 0)
+        elif chapter.title == "UncoloredHorizontalbox":
+            for content in chapter.content:
+                if content.type == "paragraph":
+                    assert content.textbox.ncolor is None
+                    for line in content.textbox.lines:
+                        assert line.ncolor is not None
+
+
+def test_colors_3() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if "Words" in chapter.title:
+            for content in chapter.content:
+                if (
+                    content.type == "paragraph"
+                    and "This line has no color" in content.textbox.text
+                ):
+                    assert content.textbox.ncolor is None
+
+                    for word in content.textbox.words:
+                        if word.text == "has":
+                            assert word.ncolor == (0, 0, 1)
+                        elif word.text == "color":
+                            assert word.ncolor in [(0, 1, 0), (0, 0, 0)]
+                        elif word.text == "changes":
+                            assert word.ncolor == (1, 0, 0)
+                        elif word.text == "words":
+                            assert word.ncolor == (0, 0, 1)
+
+
+def test_colors_4() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if "Words" in chapter.title:
+            for content in chapter.content:
+                if "This words have no color" in content.textbox.text:
+                    assert content.textbox.ncolor is None
+
+                    for word in content.textbox.words:
+                        assert word.ncolor is None or word.ncolor == (0, 0, 0)
+
+
+def test_colors_5() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if "Words" in chapter.title:
+            for content in chapter.content:
+                if "These words are printed" in content.textbox.text:
+                    assert content.textbox.ncolor is None
+
+                    for word in content.textbox.words:
+                        if word.text in ["words", "but"]:
+                            assert word.ncolor == (0, 1, 0)
+                        elif word.text == "printed":
+                            assert word.ncolor == (0, 0, 1)
+                        elif word.text == "background":
+                            assert word.ncolor == (1, 0, 0)
+
+
+def test_colors_6() -> None:
+    """Test word colors in given chapter paragraph."""
+    objects = libpdf.load(PDF_COLOR_STYLE)
+    assert objects is not None
+    assert objects.flattened.chapters
+
+    for chapter in objects.flattened.chapters:
+        if "Styled Text" in chapter.title:
+            for content in chapter.content:
+                if "bold text format" in content.textbox.text:
+                    for word in content.textbox.words:
+                        if word.text == "bold":
+                            assert "Bold" in word.fontname
+                        else:
+                            assert "Bold" not in word.fontname
+                elif "italic text format" in content.textbox.text:
+                    if word.text == "italic":
+                        assert "Italic" in word.fontname
+                    else:
+                        assert "Italic" not in word.fontname
+                elif "underline text format" in content.textbox.text:
+                    # this seems to be exracted as rect
+                    pass