feat: sb verses extracted

2026-04-14 15:58:06 +05:30
parent 82e720b10a
commit c5d4194d00
6502 changed files with 16345 additions and 7851 deletions
@@ -86,18 +86,49 @@ def normalize_devanagari_anusvara(text: str) -> str:


 def join_virama_words(text: str) -> str:
-    """Join Devanagari words where the previous word ends with virāma (्)
-    and the next word starts with a consonant.
+    """Join Devanagari words where the previous word ends with virāma (्).

-    In standard Devanagari verse rendering, a word ending in a halant
-    (virāma) connects to the following consonant-initial word without a space.
-    e.g., श्रीर् विजयो → श्रीर्विजयो
+    Handles both virāma+consonant and virāma+vowel cases.
+    In Devanagari script, a halant (virāma) consonant always absorbs the
+    following letter — whether consonant or vowel — without a space.

-    But when followed by a vowel, the space is preserved:
-    e.g., किम् अकुर्वत → किम् अकुर्वत (stays separate)
+    virāma + consonant: श्रीर् विजयो → श्रीर्विजयो
+    virāma + vowel:     किम् अकुर्वत → किमकुर्वत
+                        तैर् अलातायुधैः → तैरलातायुधैः
+                        माम् एकं → मामेकं
+
+    Uses [ \\t]+ instead of \\s+ to avoid joining across newlines (line breaks
+    separate paired pādas and must be preserved).
    """
-    # Devanagari consonants: क (U+0915) through ह (U+0939)
-    return re.sub(r"्\s+([क-ह])", r"्\1", text)
+    # Join virāma + space + consonant (क U+0915 through ह U+0939)
+    text = re.sub(r"्[ \t]+([क-ह])", r"्\1", text)
+
+    # Join virāma + space + independent vowel (अ U+0905 through औ U+0914, plus ॠ ॡ)
+    # Drop the virāma and replace the independent vowel with its dependent mātrā form.
+    _VOWEL_TO_MATRA = {
+        "अ": "",  # inherent vowel — just drop virāma and the अ
+        "आ": "ा",
+        "इ": "ि",
+        "ई": "ी",
+        "उ": "ु",
+        "ऊ": "ू",
+        "ऋ": "ृ",
+        "ॠ": "ॄ",
+        "ऌ": "ॢ",
+        "ॡ": "ॣ",
+        "ए": "े",
+        "ऐ": "ै",
+        "ओ": "ो",
+        "औ": "ौ",
+    }
+
+    def _virama_vowel_replace(m):
+        vowel = m.group(1)
+        matra = _VOWEL_TO_MATRA.get(vowel, vowel)
+        return matra
+
+    text = re.sub(r"्[ \t]+([अ-औॠॡ])", _virama_vowel_replace, text)
+    return text


 def iast_to_devanagari(iast_text: str) -> str:
@@ -85,8 +85,52 @@ def normalize_devanagari_anusvara(text: str) -> str:


 def join_virama_words(text: str) -> str:
-    """Join Devanagari words where the previous word ends with virāma."""
-    return re.sub(r"्\s+([क-ह])", r"्\1", text)
+    """Join Devanagari words where the previous word ends with virāma.
+
+    Handles both virāma+consonant and virāma+vowel cases.
+    In Devanagari script, a halant (virāma) consonant always absorbs the
+    following letter — whether consonant or vowel — without a space.
+
+    virāma + consonant: श्रीर् विजयो → श्रीर्विजयो
+    virāma + vowel:     तैर् अलातायुधैः → तैरलातायुधैः
+    """
+    # Join virāma + space + consonant (क U+0915 through ह U+0939)
+    # Use [ \t]+ instead of \s+ to avoid joining across newlines (line breaks
+    # separate paired pādas and must be preserved).
+    text = re.sub(r"्[ \t]+([क-ह])", r"्\1", text)
+    # Join virāma + space + independent vowel (अ U+0905 through औ U+0914, plus ॠ ॡ)
+    # The transliterator will have produced virāma on the consonant and a separate
+    # independent vowel glyph. We need to:
+    #   1. Drop the virāma (the consonant takes the vowel)
+    #   2. Replace the independent vowel with its dependent mātrā form
+    # However, a simpler approach: just remove the space and let the
+    # virāma+independent-vowel stay — this is visually wrong.
+    # Instead, we do the proper mapping:
+    _VOWEL_TO_MATRA = {
+        "अ": "",  # inherent vowel — just drop virāma and the अ
+        "आ": "ा",
+        "इ": "ि",
+        "ई": "ी",
+        "उ": "ु",
+        "ऊ": "ू",
+        "ऋ": "ृ",
+        "ॠ": "ॄ",
+        "ऌ": "ॢ",
+        "ॡ": "ॣ",
+        "ए": "े",
+        "ऐ": "ै",
+        "ओ": "ो",
+        "औ": "ौ",
+    }
+
+    def _virama_vowel_replace(m):
+        vowel = m.group(1)
+        matra = _VOWEL_TO_MATRA.get(vowel, vowel)
+        # Remove the virāma (्) and replace independent vowel with mātrā
+        return matra
+
+    text = re.sub(r"्[ \t]+([अ-औॠॡ])", _virama_vowel_replace, text)
+    return text


 def iast_to_devanagari(iast_text: str) -> str: