admin can split a sentence into 2 sentences

sentence_api/repository.py

@@ -331,6 +331,166 @@ class VideoRepository:
             reference_speech_duration_ms=row["reference_speech_duration_ms"],
         )
 
+    def get_sentence(
+        self,
+        video_hash: str,
+        sentence_index: int,
+    ) -> Optional[SentenceBoundary]:
+        normalized_hash = normalize_video_hash(video_hash)
+        with self._connect() as connection:
+            row = connection.execute(
+                """
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (normalized_hash, sentence_index),
+            ).fetchone()
+        if row is None:
+            return None
+        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"],
+        )
+
+    def replace_sentence(
+        self,
+        video_hash: str,
+        sentence_index: int,
+        replacements: List[SentenceBoundary],
+    ) -> Optional[List[SentenceBoundary]]:
+        """Replace one sentence row with one or more contiguous sentence rows."""
+        normalized_hash = normalize_video_hash(video_hash)
+        if not replacements:
+            raise ValueError("At least one replacement sentence is required.")
+
+        with self._connect() as connection:
+            source = connection.execute(
+                """
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (normalized_hash, sentence_index),
+            ).fetchone()
+            if source is None:
+                return None
+
+            rows = [
+                SentenceBoundary(
+                    index=index,
+                    start_ms=replacement.start_ms,
+                    end_ms=replacement.end_ms,
+                    text=replacement.text.strip(),
+                    language=replacement.language,
+                    reference_speech_duration_ms=max(
+                        1, replacement.reference_speech_duration_ms or 1
+                    ),
+                )
+                for index, replacement in enumerate(replacements)
+            ]
+            previous_end = source["start_ms"]
+            for replacement in rows:
+                if (
+                    replacement.start_ms < previous_end
+                    or replacement.end_ms <= replacement.start_ms
+                ):
+                    raise ValueError("Replacement sentence times must be contiguous.")
+                if replacement.end_ms > source["end_ms"]:
+                    raise ValueError(
+                        "Replacement sentences cannot exceed the source sentence."
+                    )
+                previous_end = replacement.end_ms
+            if rows[-1].end_ms != source["end_ms"]:
+                raise ValueError("Replacement sentences must cover the source sentence.")
+
+            # Shift following rows down first. Descending indexes avoid violating
+            # the (video_hash, sentence_index) primary key while making room.
+            added_count = len(rows) - 1
+            if added_count:
+                following = [
+                    row["sentence_index"]
+                    for row in connection.execute(
+                        """
+                        SELECT sentence_index FROM sentences
+                        WHERE video_hash = ? AND sentence_index > ?
+                        ORDER BY sentence_index DESC
+                        """,
+                        (normalized_hash, sentence_index),
+                    ).fetchall()
+                ]
+                for old_index in following:
+                    connection.execute(
+                        """
+                        UPDATE sentences SET sentence_index = ?
+                        WHERE video_hash = ? AND sentence_index = ?
+                        """,
+                        (old_index + added_count, normalized_hash, old_index),
+                    )
+
+            first = rows[0]
+            connection.execute(
+                """
+                UPDATE sentences
+                SET start_ms = ?, end_ms = ?, text = ?, language = ?,
+                    reference_speech_duration_ms = ?
+                WHERE video_hash = ? AND sentence_index = ?
+                """,
+                (
+                    first.start_ms,
+                    first.end_ms,
+                    first.text,
+                    first.language,
+                    first.reference_speech_duration_ms,
+                    normalized_hash,
+                    sentence_index,
+                ),
+            )
+            for replacement in rows[1:]:
+                connection.execute(
+                    """
+                    INSERT INTO sentences (
+                        video_hash, sentence_index, start_ms, end_ms, text,
+                        language, reference_speech_duration_ms
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?)
+                    """,
+                    (
+                        normalized_hash,
+                        replacement.index,
+                        replacement.start_ms,
+                        replacement.end_ms,
+                        replacement.text,
+                        replacement.language,
+                        replacement.reference_speech_duration_ms,
+                    ),
+                )
+            connection.execute(
+                "UPDATE videos SET updated_at = ? WHERE video_hash = ?",
+                (utc_now(), normalized_hash),
+            )
+            result_rows = connection.execute(
+                """
+                SELECT * FROM sentences
+                WHERE video_hash = ? AND sentence_index >= ? AND sentence_index < ?
+                ORDER BY sentence_index
+                """,
+                (normalized_hash, sentence_index, sentence_index + len(rows)),
+            ).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,
         *,