Stable char-to-phoneme mapping with alignment across phonetic transformations

src/quran_transcript/phonetics/phonetizer.py

@@ -1,30 +1,95 @@
 import re
-from dataclasses import dataclass
+from dataclasses import dataclass, field
+from difflib import SequenceMatcher
 
 from .operations import OPERATION_ORDER
 from .moshaf_attributes import MoshafAttributes
 from .sifa import process_sifat, SifaOutput
 from .. import alphabet as alph
 
 
+def _distribute_mapping(segment: list[int | None], new_len: int) -> list[int | None]:
+    """Spread original indices over a new segment length."""
+    if new_len <= 0:
+        return []
+    if not segment:
+        return [None] * new_len
+    if len(segment) == 1:
+        return [segment[0]] * new_len
+
+    distributed: list[int | None] = []
+    seg_len = len(segment)
+    for k in range(new_len):
+        idx = int(k * seg_len / new_len)
+        if idx >= seg_len:
+            idx = seg_len - 1
+        distributed.append(segment[idx])
+    return distributed
+
+
+def _align_mapping(
+    before_text: str, after_text: str, before_mapping: list[int | None]
+) -> list[int | None]:
+    """Align mapping list after a text transformation using sequence matching."""
+
+    matcher = SequenceMatcher(a=before_text, b=after_text)
+    new_mapping: list[int | None] = []
+
+    for tag, i1, i2, j1, j2 in matcher.get_opcodes():
+        if tag == "equal":
+            new_mapping.extend(before_mapping[i1:i2])
+        elif tag == "replace":
+            segment = before_mapping[i1:i2]
+            new_mapping.extend(_distribute_mapping(segment, j2 - j1))
+        elif tag == "insert":
+            new_mapping.extend([None] * (j2 - j1))
+        elif tag == "delete":
+            continue
+
+    return new_mapping
+
+
+def _substitute_with_mapping(
+    text: str, mapping: list[int | None], pattern: str, replacement: str
+) -> tuple[str, list[int | None]]:
+    """Apply regex substitution while updating the origin mapping via alignment."""
+
+    new_text = re.sub(pattern, replacement, text)
+    if new_text == text:
+        return text, mapping
+
+    new_mapping = _align_mapping(text, new_text, mapping)
+    return new_text, new_mapping
+
+
 @dataclass
 class QuranPhoneticScriptOutput:
     phonemes: str
     sifat: list[SifaOutput]
+    char_map: list[int | None] = field(default_factory=list)
 
 
 def quran_phonetizer(
     uhtmani_text: str, moshaf: MoshafAttributes, remove_spaces=False
 ) -> QuranPhoneticScriptOutput:
     """الرسم الصوتي للقآن الكريم على طبقتين: طبقة الأحرف وطبقة الصفات"""
     text = uhtmani_text
+    mapping: list[int | None] = list(range(len(text)))
 
     # cleaning extra scpace
-    text = re.sub(r"\s+", rf"{alph.uthmani.space}", text)
-    text = re.sub(r"(\s$|^\s)", r"", text)
+    text, mapping = _substitute_with_mapping(
+        text, mapping, r"\s+", rf"{alph.uthmani.space}"
+    )
+    text, mapping = _substitute_with_mapping(text, mapping, r"(\s$|^\s)", r"")
 
     for op in OPERATION_ORDER:
+        prev_text = text
         text = op.apply(text, moshaf)
+        if text != prev_text:
+            # print(f"Applied: {op.arabic_name}")
+            # print(f"  Before: {prev_text}")
+            # print(f"  After : {text}")
+            mapping = _align_mapping(prev_text, text, mapping)
 
     sifat = process_sifat(
         uthmani_script=uhtmani_text,
@@ -33,6 +98,13 @@ def quran_phonetizer(
     )
 
     if remove_spaces:
-        text = re.sub(alph.uthmani.space, r"", text)
+        filtered_chars: list[str] = []
+        filtered_map: list[int | None] = []
+        for ch, idx in zip(text, mapping):
+            if ch != alph.uthmani.space:
+                filtered_chars.append(ch)
+                filtered_map.append(idx)
+        text = "".join(filtered_chars)
+        mapping = filtered_map
 
-    return QuranPhoneticScriptOutput(phonemes=text, sifat=sifat)
+    return QuranPhoneticScriptOutput(phonemes=text, sifat=sifat, char_map=mapping)