Upload 11 files

subtitle_parser.py

@@ -1,12 +1,6 @@
 """
-subtitle_parser.py — Subtitle Parsing & Writing  (Step 3)
+subtitle_parser.py — Subtitle Parsing & Writing
 Handles .srt / .vtt / .ass via pysubs2.
-
-ASS-specific handling:
-  - Inline override tags {\\pos(...)\\be5} are preserved on replace
-  - Style-based skip (EDR/romaji) — lines returned as "" → kept original
-  - Style: Default, Alt, Note, EDE → translated
-  - Style: EDR (romaji) → skipped (returned as-is)
 """
 
 import copy
@@ -20,17 +14,32 @@ import pysubs2
 
 logger = logging.getLogger(__name__)
 
-# ── Workspace ──────────────────────────────────────────────────────────────
 WORKSPACE_DIR = os.path.join(tempfile.gettempdir(), "subsync_workspace")
 os.makedirs(WORKSPACE_DIR, exist_ok=True)
 
-# ── Styles to skip (do not send to AI, keep original text) ────────────────
-# EDR = Japanese romaji in fansub ASS files — should never be translated
-SKIP_STYLES = {"EDR", "Romaji", "JP", "Karaoke", "KAR"}
+# ASS styles that should NOT be translated (romaji, karaoke, etc.)
+SKIP_STYLES = {
+    "EDR", "Romaji", "JP", "Karaoke", "KAR",
+    "Signs", "Lyrics_JP", "Romaji2", "OP_Romaji", "ED_Romaji",
+}
 
-# ── Format map ─────────────────────────────────────────────────────────────
 _EXT_MAP = {".srt": "srt", ".vtt": "vtt", ".ass": "ass", ".ssa": "ass"}
 
+_ASS_TAG_RE      = re.compile(r"\{[^}]*\}")
+_HTML_TAG_RE     = re.compile(r"<[^>]+>")
+_LEADING_TAGS_RE = re.compile(r"^(\{[^}]*\})+")
+
+
+def is_skip_style(style_name: str) -> bool:
+    """Return True if this ASS style should be skipped (not translated)."""
+    if not style_name:
+        return False
+    s = style_name.strip()
+    if s in SKIP_STYLES:
+        return True
+    sl = s.lower()
+    return any(k in sl for k in ["romaji", "karaoke", "kar_", "_kar", "furigana"])
+
 
 def _detect_format(filename: str) -> str:
     ext = os.path.splitext(filename.lower())[1]
@@ -42,10 +51,14 @@ def _detect_format(filename: str) -> str:
     return fmt
 
 
-# ── Parse ──────────────────────────────────────────────────────────────────
+def _strip_tags(text: str) -> str:
+    text = _ASS_TAG_RE.sub("", text)
+    text = _HTML_TAG_RE.sub("", text)
+    return text.strip()
+
 
 def parse_subtitle(file_bytes: bytes, filename: str):
-    """Parse subtitle bytes → (SSAFile, fmt_string). Raises ValueError on failure."""
+    """Parse subtitle bytes → (SSAFile, fmt_string)."""
     fmt = _detect_format(filename)
 
     text = None
@@ -78,47 +91,27 @@ def parse_subtitle(file_bytes: bytes, filename: str):
         os.unlink(tmp)
         return subs, fmt
     except Exception as e2:
-        raise ValueError(f"Parse မရပါ:\nAttempt 1: {e1}\nAttempt 2: {e2}") from e2
-
-
-# ── Tag stripping ──────────────────────────────────────────────────────────
-_ASS_TAG_RE  = re.compile(r"\{[^}]*\}")   # {\\an8} {\\pos(x,y)} etc.
-_HTML_TAG_RE = re.compile(r"<[^>]+>")     # <i> <b> etc.
-# Leading tag block: one or more {...} at the start of the string
-_LEADING_TAGS_RE = re.compile(r"^(\{[^}]*\})+")
-
-
-def _strip_tags(text: str) -> str:
-    text = _ASS_TAG_RE.sub("", text)
-    text = _HTML_TAG_RE.sub("", text)
-    return text.strip()
-
+        raise ValueError(
+            f"Parse မရပါ:\nAttempt 1: {e1}\nAttempt 2: {e2}"
+        ) from e2
 
-# ── Plain line extraction ──────────────────────────────────────────────────
 
 def get_plain_lines(ssafile) -> list:
     """
     Return list[str] same length as ssafile.events.
-
-    Translate decisions per event:
-    ✅ Dialogue events with translatable style → return clean text
-    ✗ Comment events (editor notes, not shown to viewer) → return ""
-    ✗ Style in SKIP_STYLES (EDR/Romaji/JP/Karaoke) → return ""
-      Empty string → replace_lines keeps original event.text unchanged.
-
-    ASS inline tags {\\an8}{\\pos(...)} stripped before returning.
-    \\N / \\n normalised to Python newline.
+    - Comment events  → "" (keep original, not displayed)
+    - Skip styles     → "" (EDR/romaji/karaoke etc.)
+    - Normal events   → clean text (tags stripped)
+    Empty string → replace_lines keeps original text unchanged.
     """
     lines = []
     for event in ssafile.events:
-        # Skip Comment lines (editor notes — not displayed to viewer)
-        # pysubs2 exposes both Dialogue and Comment in events list
-        is_comment = getattr(event, "is_comment", False)
-        if is_comment:
+        # Skip ASS Comment lines (editor notes, not shown to viewer)
+        if getattr(event, "is_comment", False):
             lines.append("")
             continue
 
-        # Skip non-English / romaji / karaoke style tracks (exact + pattern match)
+        # Skip romaji / karaoke / non-English style tracks
         if is_skip_style(event.style):
             lines.append("")
             continue
@@ -129,24 +122,11 @@ def get_plain_lines(ssafile) -> list:
     return lines
 
 
-# ── Replace lines — tag-safe ───────────────────────────────────────────────
-
 def replace_lines(ssafile, new_lines: list):
     """
-    Return new SSAFile with translated text injected index-for-index.
-
-    TIMESTAMP SAFETY:
-    - event.start / event.end never touched (deep-copy preserves them)
-    - Only event.text is changed
-
-    ASS TAG PRESERVATION:
-    - Leading override tags {\\pos(...)\\be5} etc. are extracted from the
-      original event.text and re-prepended to the translated text.
-    - This keeps positioning, blur, fade effects intact after translation.
-
-    SKIP STYLES:
-    - If translated line is "" (empty) → original event.text kept unchanged.
-      This handles EDR (romaji) and any other skip-style events.
+    Return new SSAFile with translated text. Timestamps never touched.
+    Leading ASS override tags {\\pos(...)\\be5} preserved from original.
+    Empty translation → keep original event.text.
     """
     n_events = len(ssafile.events)
     n_lines  = len(new_lines)
@@ -157,44 +137,26 @@ def replace_lines(ssafile, new_lines: list):
             f"!= subtitle events ({n_events})"
         )
 
-    result       = copy.deepcopy(ssafile)
-    kept_count   = 0
-    tagged_count = 0
+    result = copy.deepcopy(ssafile)
 
     for i, event in enumerate(result.events):
         translated = str(new_lines[i]).strip()
 
-        # ── Empty translation → keep original (skip-style or parse fail) ──
+        # Empty → keep original (skip-style or fallback)
         if not translated:
-            kept_count += 1
-            continue   # event.text unchanged
+            continue
 
-        # ── Preserve leading inline override tags ──────────────────────────
-        # e.g. {\\pos(960,540)\\fad(200,0)} Hello → {\\pos(960,540)\\fad(200,0)}မင်္ဂလာပါ
-        original_text = ssafile.events[i].text   # from original (pre-deepcopy)
+        # Re-prepend leading inline tags from original
+        original_text = ssafile.events[i].text
         leading_m     = _LEADING_TAGS_RE.match(original_text)
         tag_prefix    = leading_m.group(0) if leading_m else ""
 
-        if tag_prefix:
-            tagged_count += 1
+        event.text = tag_prefix + translated.replace("\n", "\\N")
 
-        # Convert Python newlines back to ASS \\N
-        body       = translated.replace("\n", "\\N")
-        event.text = tag_prefix + body
-
-    if kept_count:
-        logger.info("replace_lines: %d/%d events kept original (skip/empty)",
-                    kept_count, n_events)
-    if tagged_count:
-        logger.info("replace_lines: %d events had leading tags preserved",
-                    tagged_count)
     return result
 
 
-# ── Write ──────────────────────────────────────────────────────────────────
-
 def write_subtitle(ssafile, fmt: str) -> bytes:
-    """Serialise SSAFile to UTF-8 bytes in original format."""
     try:
         return ssafile.to_string(fmt).encode("utf-8")
     except Exception as exc:
@@ -202,13 +164,10 @@ def write_subtitle(ssafile, fmt: str) -> bytes:
 
 
 def output_filename(original_name: str) -> str:
-    """'movie.en.ass' → 'movie.en.my.ass'"""
     base, ext = os.path.splitext(original_name)
     return f"{base}.my{ext}"
 
 
-# ── Workspace management ───────────────────────────────────────────────────
-
 def clean_workspace() -> None:
     try:
         if os.path.exists(WORKSPACE_DIR):