Fix chord alignment: snap to word boundaries

Improve merge_chord_lyric() to snap chord positions to the start
of the word they fall within, instead of splitting words mid-way.
Fixes artifacts like "Liebespaar \chord{C}e" → "\chord{C}Liebespaare".

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

import-songs.py

@@ -273,7 +273,11 @@ def parse_chords_with_positions(chord_line: str) -> list:
 
 
 def merge_chord_lyric(chord_line: str, lyric_line: str) -> str:
-    """Merge a chord line and lyric line by inserting \\chord{X} at positions."""
+    """Merge a chord line and lyric line by inserting \\chord{X} at positions.
+
+    Snaps chord positions to word boundaries: if a chord falls inside a word,
+    it is moved to the start of that word.
+    """
     chords = parse_chords_with_positions(chord_line)
 
     if not chords:
@@ -281,27 +285,45 @@ def merge_chord_lyric(chord_line: str, lyric_line: str) -> str:
 
     original_len = len(lyric_line)
 
-    # Separate chords into: within lyric range vs beyond
+    # Only keep chords within lyric range
     in_range = [(pos, ch) for pos, ch in chords if pos < original_len]
-    beyond = [(pos, ch) for pos, ch in chords if pos >= original_len]
 
-    # Insert in-range chords from right to left to maintain positions
+    # Snap each chord position to the start of the word it falls in
+    snapped = []
+    for pos, chord in in_range:
+        # If position is inside a word (not at a space or start of word),
+        # scan backwards to find the word start
+        snap_pos = pos
+        if pos > 0 and pos < len(lyric_line) and lyric_line[pos] != ' ':
+            # Check if previous char is also non-space (mid-word)
+            if lyric_line[pos - 1] != ' ':
+                # Scan backwards to find word start
+                while snap_pos > 0 and lyric_line[snap_pos - 1] != ' ':
+                    snap_pos -= 1
+        # If we're at a space, move forward to the next non-space
+        while snap_pos < len(lyric_line) and lyric_line[snap_pos] == ' ':
+            snap_pos += 1
+        snapped.append((snap_pos, chord))
+
+    # Deduplicate: if multiple chords snap to the same position, keep all
+    # but ensure we don't create \chord{A}\chord{B} without space
+    # Insert from right to left to maintain positions
     result = lyric_line
-    for pos, chord in reversed(in_range):
+    prev_insert_pos = len(result) + 1
+    for pos, chord in reversed(snapped):
         chord_macro = f"\\chord{{{chord}}}"
-        result = result[:pos] + chord_macro + result[pos:]
-
-    # Append beyond-range chords (they have no corresponding lyric text)
-    # These are typically overflow chords from the PDF layout
-    # Skip them as they don't have corresponding lyrics on this line
-    # (they may belong to the next lyric line or be layout artifacts)
+        # Avoid inserting at same position as previous (from right)
+        if pos < prev_insert_pos:
+            result = result[:pos] + chord_macro + result[pos:]
+            prev_insert_pos = pos
 
     result = result.strip()
+    # Collapse multiple spaces but preserve single spaces
     result = re.sub(r'  +', ' ', result)
 
-    # Ensure there's a space before \chord{} when preceded by non-space text
-    # This prevents leadsheets parsing issues with adjacent chord macros
-    result = re.sub(r'([^\s\\])\\chord\{', r'\1 \\chord{', result)
+    # For adjacent chords without text between them (e.g. \chord{A}\chord{B}),
+    # add a space to prevent leadsheets tabular nesting issues
+    result = re.sub(r'\}(\\chord\{)', r'} \1', result)
 
     return result