chg: improved HTML annotator.

AlbertWeichselbraun · AlbertWeichselbraun · commit d5e41919e6c7 · 2025-03-22T23:27:05.000+01:00
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "inscriptis"
-version = "2.6.0"
+version = "2.6.1"
 authors = ["Albert Weichselbraun <albert.weichselbraun@fhgr.ch>", "Fabian Odoni <fabian.odoni@fhgr.ch>"]
 description = "inscriptis - HTML to text converter."
 keywords = ["HTML", "converter", "text"]
diff --git a/src/inscriptis/annotation/__init__.py b/src/inscriptis/annotation/__init__.py
@@ -28,6 +28,12 @@ class Annotation(NamedTuple):
     metadata: str
     """the tag to be attached to the annotation."""
 
+    def __lt__(self, other):
+        return (self.start, -self.end) < (other.start, -other.end)
+
+    def __gt__(self, other):
+        return not self.__lt__(other)
+
 
 def horizontal_shift(
     annotations: List[Annotation],
diff --git a/src/inscriptis/annotation/output/html.py b/src/inscriptis/annotation/output/html.py
@@ -4,6 +4,7 @@
 from itertools import cycle
 from typing import Dict, Any, List
 
+from inscriptis.annotation import Annotation
 from inscriptis.annotation.output import AnnotationProcessor
 
 COLOR_SCHEMA = ("#D8115980", "#8F2D5680", "#21838080", "#FBB13C80", "#73D2DE80")
@@ -20,12 +21,11 @@ class HtmlExtractor(AnnotationProcessor):
 
     def __call__(self, annotated_text: Dict[str, Any]) -> str:
         tag_dict = defaultdict(list)
-
-        for start, end, label in reversed(annotated_text["label"]):
-            tag_dict[start].append(
-                f'<span class="{label}-label">{label}</span><span class="{label}">'
-            )
-            tag_dict[end].insert(0, "</span>")
+        for start, end, label in sorted(
+            Annotation(s, e, t) for s, e, t in reversed(annotated_text["label"])
+        ):
+            tag_dict[start].append(f'<span class="{label}" data-label="{label}">')
+            tag_dict[end].insert(0, f"</span>")
 
         tagged_content = [
             "<html><head><style>",
@@ -36,10 +36,10 @@ def __call__(self, annotated_text: Dict[str, Any]) -> str:
         text = annotated_text["text"]
         current_idx = 0
         for idx, tags in sorted(tag_dict.items()):
-            tagged_content.append(text[current_idx:idx].replace("\n", "</pre>\n<pre>"))
+            tagged_content.append(text[current_idx:idx])
             current_idx = idx
             tagged_content.extend(tags)
-        tagged_content.append(text[current_idx:].replace("\n", "</pre>\n</pre>"))
+        tagged_content.append(text[current_idx:])
         return "".join(tagged_content) + "</pre></body></html>"
 
     @staticmethod
@@ -72,17 +72,16 @@ def _get_css(self, labels: List[str]) -> str:
         for label, color in sorted(self._get_label_colors(labels).items()):
             css.append(
                 "pre{{"
-                "  position: relative;\n"
+                "position: relative; white-space: pre; line-height: 2.5; font-family: monospace;"
                 "}}\n"
-                ".{label} {{\n"
-                "  background-color: {color};\n"
-                "  border-radius: 0.4em;\n"
+                ".{label} {{ position: relative; display: inline-block; white-space: pre;"
+                "  background-color: {color}; \n"
+                "  border-radius: 0.4em; padding: 0 4px\n"
                 "}}\n"
-                ".{label}-label {{\n"
-                "  top: -1.0em;\n"
-                '  content: "{label}";\n'
-                "  position: absolute;\n"
-                "  background-color: {color};\n"
-                "  font-size: 75%; }}\n".format(label=label, color=color)
+                ".{label}::before {{"
+                "content: attr(data-label); position: absolute; top: -1.3em; left: 0;"
+                "background-color: {color}; "
+                "font-size: 65%; padding: 0px 2px; border-radius: 1px; white-space: nowrap; font-weight: bold;"
+                " }}\n".format(label=label, color=color)
             )
         return "\n".join(css)
diff --git a/src/inscriptis/annotation/output/xml.py b/src/inscriptis/annotation/output/xml.py
@@ -3,6 +3,7 @@
 from collections import defaultdict
 from typing import Dict, Any
 
+from inscriptis.annotation import Annotation
 from inscriptis.annotation.output import AnnotationProcessor
 
 
@@ -13,7 +14,9 @@ class XmlExtractor(AnnotationProcessor):
 
     def __call__(self, annotated_text: Dict[str, Any], root_element="content"):
         tag_dict = defaultdict(list)
-        for start, end, tag in reversed(annotated_text["label"]):
+        for start, end, tag in sorted(
+            Annotation(s, e, t) for s, e, t in reversed(annotated_text["label"])
+        ):  # noqa: C414
             tag_dict[start].append(f"<{tag}>")
             tag_dict[end].insert(0, f"</{tag}>")
 
diff --git a/tests/test_annotation.py b/tests/test_annotation.py
@@ -69,3 +69,47 @@ def test_horizontal_shift():
         align=HorizontalAlignment.center,
         shift=7,
     ).pop() == Annotation(10, 14, "test")
+
+
+def test_comparison():
+    a = Annotation(24, 62, "tableheading")
+    b = Annotation(24, 600, "table")
+    c = Annotation(51, 56, "tableheading")
+
+    assert b > b
+    assert a < c
+    assert b < c
+
+
+def test_sorting():
+    annotations = [
+        Annotation(24, 62, "tableheading"),
+        Annotation(24, 600, "table"),
+        Annotation(51, 56, "tableheading"),
+        Annotation(59, 115, "tableheading"),
+        Annotation(79, 104, "emphasis"),
+        Annotation(125, 139, "tableheading"),
+        Annotation(140, 160, "emphasis"),
+        Annotation(254, 263, "link"),
+        Annotation(254, 271, "bold"),
+        Annotation(266, 268, "link"),
+        Annotation(271, 280, "link"),
+        Annotation(369, 385, "link"),
+        Annotation(484, 498, "link"),
+    ]
+
+    assert sorted(annotations) == [
+        Annotation(24, 600, "table"),
+        Annotation(24, 62, "tableheading"),
+        Annotation(51, 56, "tableheading"),
+        Annotation(59, 115, "tableheading"),
+        Annotation(79, 104, "emphasis"),
+        Annotation(125, 139, "tableheading"),
+        Annotation(140, 160, "emphasis"),
+        Annotation(254, 271, "bold"),
+        Annotation(254, 263, "link"),
+        Annotation(266, 268, "link"),
+        Annotation(271, 280, "link"),
+        Annotation(369, 385, "link"),
+        Annotation(484, 498, "link"),
+    ]
diff --git a/tests/test_annotation_output_processor.py b/tests/test_annotation_output_processor.py
@@ -63,13 +63,11 @@ def test_html_annotator():
     assert result.startswith("<html><head><style>")
     assert result.split("</style>")[1] == (
         "</head>"
-        '<body><pre><span class="heading-label">heading'
-        '</span><span class="heading">'
-        '<span class="h1-label">h1</span><span class="h1">'
-        "Chur</span></span></pre>\n"
-        "<pre></pre>\n"
-        '<pre><span class="emphasis-label">emphasis</span>'
-        '<span class="emphasis">Chur</span> is the capital '
+        '<body><pre><span class="heading" data-label="heading">'
+        '<span class="h1" data-label="h1">'
+        "Chur</span></span>\n\n"
+        '<span class="emphasis" data-label="emphasis">'
+        "Chur</span> is the capital "
         "and largest town of the Swiss canton of the "
         "Grisons and lies in the Grisonian Rhine Valley."
         "</pre></body></html>"
