allow ajust sentences

sentence_api/repository.py

@@ -491,6 +491,121 @@ class VideoRepository:
             for row in result_rows
         ]
 
+    def adjust_sentence_boundary(
+        self,
+        video_hash: str,
+        sentence_index: int,
+        delta_ms: int,
+    ) -> Optional[List[SentenceBoundary]]:
+        """Move one sentence end and link the next sentence's start to it."""
+        normalized_hash = normalize_video_hash(video_hash)
+        with self._connect() as connection:
+            video = connection.execute(
+                "SELECT duration_ms FROM videos WHERE video_hash = ?",
+                (normalized_hash,),
+            ).fetchone()
+            if video is None or video["duration_ms"] is None:
+                return None
+
+            current = connection.execute(
+                """
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (normalized_hash, sentence_index),
+            ).fetchone()
+            if current is None:
+                return None
+
+            following = connection.execute(
+                """
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (normalized_hash, sentence_index + 1),
+            ).fetchone()
+
+            duration_ms = int(video["duration_ms"])
+            minimum_end = current["start_ms"] + 1
+            maximum_end = duration_ms
+            if following is not None:
+                maximum_end = min(maximum_end, following["end_ms"] - 1)
+            new_end_ms = current["end_ms"] + delta_ms
+            if new_end_ms < minimum_end or new_end_ms > maximum_end:
+                raise ValueError(
+                    "The adjusted boundary must remain inside both sentences."
+                )
+
+            current_duration_ms = new_end_ms - current["start_ms"]
+            current_reference_ms = current["reference_speech_duration_ms"] or current_duration_ms
+            current_reference_ms = min(
+                current_duration_ms,
+                max(1, current_reference_ms + delta_ms),
+            )
+            connection.execute(
+                """
+                UPDATE sentences
+                SET end_ms = ?, reference_speech_duration_ms = ?
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (
+                    new_end_ms,
+                    current_reference_ms,
+                    normalized_hash,
+                    sentence_index,
+                ),
+            )
+
+            updated_indexes = [sentence_index]
+            if following is not None:
+                old_start_ms = following["start_ms"]
+                next_duration_ms = following["end_ms"] - new_end_ms
+                next_reference_ms = following["reference_speech_duration_ms"] or next_duration_ms
+                next_reference_ms = min(
+                    next_duration_ms,
+                    max(1, next_reference_ms - (new_end_ms - old_start_ms)),
+                )
+                connection.execute(
+                    """
+                    UPDATE sentences
+                    SET start_ms = ?, reference_speech_duration_ms = ?
+                    WHERE video_hash = ? AND sentence_index = ?
+                    """,
+                    (
+                        new_end_ms,
+                        next_reference_ms,
+                        normalized_hash,
+                        sentence_index + 1,
+                    ),
+                )
+                updated_indexes.append(sentence_index + 1)
+
+            connection.execute(
+                "UPDATE videos SET updated_at = ? WHERE video_hash = ?",
+                (utc_now(), normalized_hash),
+            )
+            placeholders = ", ".join("?" for _ in updated_indexes)
+            result_rows = connection.execute(
+                f"""
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index IN ({placeholders})
+                ORDER BY sentence_index
+                """,
+                (normalized_hash, *updated_indexes),
+            ).fetchall()
+
+        return [
+            SentenceBoundary(
+                index=row["sentence_index"],
+                start_ms=row["start_ms"],
+                end_ms=row["end_ms"],
+                text=row["text"],
+                language=row["language"],
+                reference_speech_duration_ms=row["reference_speech_duration_ms"],
+            )
+            for row in result_rows
+        ]
+
     def create_dub_share(
         self,
         *,