Keep word spacing and drop empty link targets in HTMLToMarkdown (#1347)

src/python/txtai/pipeline/data/htmltomd.py

@@ -194,12 +194,17 @@ class HTMLToMarkdown(Pipeline):
             target, text = x if x.name else node, x.text
 
             if text.strip():
+                # Leading whitespace separates this text from the previous word, keep it outside of the formatting
+                space = " " if text[0].isspace() and texts and not texts[-1][-1:].isspace() else ""
+
                 if target.name in ("b", "strong"):
-                    text = f"**{text.strip()}** "
+                    text = f"{space}**{text.strip()}** "
                 elif target.name in ("i", "em"):
-                    text = f"*{text.strip()}* "
+                    text = f"{space}*{text.strip()}* "
                 elif target.name == "a":
-                    text = f"[{text.strip()}]({target.get('href')}) "
+                    # Links without a target are added as plain text
+                    href = target.get("href")
+                    text = f"{space}[{text.strip()}]({href}) " if href else f"{space}{text.strip()} "
 
             texts.append(text)
 

test/python/testpipeline/testdata/testtextractor.py

@@ -147,6 +147,18 @@ class TestTextractor(unittest.TestCase):
         self.assertMarkdown("<p>This is a <em>test</em></p>", "This is a *test*")
         self.assertMarkdown("<p>This is a <a href='link'>test</a>", "This is a [test](link)")
 
+        # Leading whitespace inside a formatting tag
+        self.assertMarkdown("<p>This is<b> a test</b></p>", "This is **a test**")
+        self.assertMarkdown("<p>This is<em> a test</em></p>", "This is *a test*")
+        self.assertMarkdown("<p>This is<a href='link'> a test</a></p>", "This is [a test](link)")
+        self.assertMarkdown("<p>This is a te<b>st</b></p>", "This is a te**st**")
+        self.assertMarkdown("<p>This is\n<b> a test</b></p>", "This is\n**a test**")
+
+        # Links without a target are plain text
+        self.assertMarkdown("<p>This is a <a name='anchor'>test</a></p>", "This is a test")
+        self.assertMarkdown("<p>This is<a name='anchor'> a test</a></p>", "This is a test")
+        self.assertMarkdown("<h2><a id='anchor'>This is a test</a></h2>", "## This is a test")
+
         # Collapse to outer tag
         self.assertMarkdown("<p>This is a <strong><em>test</em></strong></p>", "This is a **test**")
         self.assertMarkdown("<p>This is a <em><strong>test</strong></em></p>", "This is a *test*")