Speed up daily reindex: skip empty JSON rewrites and upsert only changed posts.

Keep channel mtime stable when merge finds no new posts, and rebuild FTS only for new/changed content so large daily channels stay cheap. Drop --reindex-all from renew.sh for incremental runs. Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-04 18:31:15 +07:00
parent 0b87a1b339
commit 4820fb884b
3 changed files with 175 additions and 58 deletions
@@ -624,6 +624,10 @@ def export_channel(d: Driver, channel: str, user_id_to_name: Dict[str, str], out
                merged_posts.append(post)
                seen_ids.add(post["id"])
                new_count += 1
+        if new_count == 0:
+            # Keep mtime stable so mm_index can skip this channel file.
+            print(f"  no new posts, leaving '{output_filepath}' unchanged")
+            return
        merged_posts.sort(key=lambda p: p["created"])
        for i, post in enumerate(merged_posts):
            post["idx"] = i
@@ -143,10 +143,102 @@ def load_posts(path: str) -> Optional[Tuple[Dict[str, Any], List[Dict[str, Any]]
    return channel_obj, posts


-def upsert_file(conn: sqlite3.Connection, path: str) -> Tuple[int, int]:
+def _delete_post_search_rows(conn: sqlite3.Connection, post_id: str) -> None:
+    conn.execute("DELETE FROM attachments_fts WHERE post_id = ?", (post_id,))
+    conn.execute("DELETE FROM attachments WHERE post_id = ?", (post_id,))
+    conn.execute("DELETE FROM posts_fts WHERE post_id = ?", (post_id,))
+
+
+def _delete_post(conn: sqlite3.Connection, post_id: str) -> None:
+    _delete_post_search_rows(conn, post_id)
+    conn.execute("DELETE FROM posts WHERE post_id = ?", (post_id,))
+
+
+def _index_attachments(
+    conn: sqlite3.Connection,
+    post_id: str,
+    files: Any,
+    idx: Optional[int],
+    base_dir: str,
+    max_text_size: int,
+) -> None:
+    if not isinstance(files, list) or not files:
+        return
+    for filename in files:
+        if not isinstance(filename, str) or not filename:
+            continue
+        disk_name = f"{idx:03d}_{filename}" if isinstance(idx, int) else filename
+        disk_path = os.path.join(base_dir, disk_name)
+        is_text = 0
+        content: Optional[str] = None
+        try:
+            st = os.stat(disk_path)
+            if st.st_size <= max_text_size:
+                with open(disk_path, "r", encoding="utf-8") as f:
+                    content = f.read()
+                is_text = 1
+        except (OSError, UnicodeDecodeError):
+            is_text = 0
+            content = None
+
+        cur = conn.execute(
+            """
+            INSERT INTO attachments (post_id, filename, filepath, is_text)
+            VALUES (?, ?, ?, ?)
+            """,
+            (post_id, filename, disk_path, is_text),
+        )
+        attachment_id = cur.lastrowid
+        conn.execute(
+            """
+            INSERT INTO attachments_fts (attachment_id, post_id, filename, content)
+            VALUES (?, ?, ?, ?)
+            """,
+            (attachment_id, post_id, filename, content if is_text else None),
+        )
+
+
+def _content_changed(prev: sqlite3.Row, message: str, username: str, created: str, files_json: Optional[str]) -> bool:
+    return (
+        prev["message"] != message
+        or prev["username"] != username
+        or prev["created"] != created
+        or prev["files_json"] != files_json
+    )
+
+
+def _meta_changed(
+    prev: sqlite3.Row,
+    idx: Optional[int],
+    channel_id: str,
+    channel_name: str,
+    channel_display_name: str,
+    team: str,
+    path: str,
+    exported_at: Optional[str],
+) -> bool:
+    return (
+        prev["idx"] != idx
+        or prev["channel_id"] != channel_id
+        or prev["channel_name"] != channel_name
+        or prev["channel_display_name"] != channel_display_name
+        or prev["team"] != team
+        or prev["source_path"] != path
+        or prev["exported_at"] != exported_at
+    )
+
+
+def upsert_file(conn: sqlite3.Connection, path: str, *, force: bool = False) -> Tuple[int, int, int]:
+    """Index posts from one export JSON.
+
+    Returns (upserted_or_deleted, invalid_skipped, unchanged).
+    When force=False, only new/changed posts rebuild FTS/attachments; idx-only
+    renumbers get a cheap posts-row UPDATE. force=True rebuilds every post
+    (--reindex-all).
+    """
    loaded = load_posts(path)
    if loaded is None:
-        return 0, 0
+        return 0, 0, 0

    channel_obj, posts = loaded
    channel_id = str(channel_obj.get("id") or "")
@@ -155,27 +247,28 @@ def upsert_file(conn: sqlite3.Connection, path: str) -> Tuple[int, int]:
    team = str(channel_obj.get("team") or "")
    exported_at = parse_exported_at(channel_obj)

-    # Remove any existing attachments for posts coming from this source file
-    conn.execute(
-        """
-        DELETE FROM attachments
-        WHERE post_id IN (SELECT post_id FROM posts WHERE source_path = ?)
-        """,
-        (path,),
-    )
-    conn.execute(
-        """
-        DELETE FROM attachments_fts
-        WHERE post_id IN (SELECT post_id FROM posts WHERE source_path = ?)
-        """,
-        (path,),
-    )
+    existing = {
+        row["post_id"]: row
+        for row in conn.execute(
+            """
+            SELECT post_id, idx, created, username, message, files_json,
+                   channel_id, channel_name, channel_display_name, team,
+                   source_path, exported_at
+            FROM posts
+            WHERE source_path = ?
+            """,
+            (path,),
+        )
+    }

    base_dir = os.path.dirname(path)
    max_text_size = 2 * 1024 * 1024  # 2 MiB

    inserted_or_updated = 0
+    unchanged = 0
    skipped = 0
+    seen_ids = set()
+
    for p in posts:
        post_id = p.get("id")
        created_raw = p.get("created")
@@ -198,6 +291,49 @@ def upsert_file(conn: sqlite3.Connection, path: str) -> Tuple[int, int]:
        files = p.get("files")
        files_json = json.dumps(files, ensure_ascii=False) if isinstance(files, list) else None
        idx = p.get("idx") if isinstance(p.get("idx"), int) else None
+        seen_ids.add(post_id)
+        prev = existing.get(post_id)
+
+        if (
+            not force
+            and prev is not None
+            and not _content_changed(prev, message, username, created, files_json)
+        ):
+            if _meta_changed(
+                prev,
+                idx,
+                channel_id,
+                channel_name,
+                channel_display_name,
+                team,
+                path,
+                exported_at,
+            ):
+                conn.execute(
+                    """
+                    UPDATE posts SET
+                        idx = ?,
+                        channel_id = ?,
+                        channel_name = ?,
+                        channel_display_name = ?,
+                        team = ?,
+                        source_path = ?,
+                        exported_at = ?
+                    WHERE post_id = ?
+                    """,
+                    (
+                        idx,
+                        channel_id,
+                        channel_name,
+                        channel_display_name,
+                        team,
+                        path,
+                        exported_at,
+                        post_id,
+                    ),
+                )
+            unchanged += 1
+            continue

        conn.execute(
            """
@@ -234,45 +370,16 @@ def upsert_file(conn: sqlite3.Connection, path: str) -> Tuple[int, int]:
                exported_at,
            ),
        )
-        conn.execute("DELETE FROM posts_fts WHERE post_id = ?", (post_id,))
+        _delete_post_search_rows(conn, post_id)
        conn.execute("INSERT INTO posts_fts(post_id, message) VALUES (?, ?)", (post_id, message))
-
-        # Index attachments (file names and, for text files, contents)
-        if isinstance(files, list) and files:
-            for filename in files:
-                if not isinstance(filename, str) or not filename:
-                    continue
-                disk_name = f"{idx:03d}_{filename}" if isinstance(idx, int) else filename
-                disk_path = os.path.join(base_dir, disk_name)
-                is_text = 0
-                content: Optional[str] = None
-                try:
-                    st = os.stat(disk_path)
-                    if st.st_size <= max_text_size:
-                        with open(disk_path, "r", encoding="utf-8") as f:
-                            content = f.read()
-                        is_text = 1
-                except (OSError, UnicodeDecodeError):
-                    is_text = 0
-                    content = None
-
-                cur = conn.execute(
-                    """
-                    INSERT INTO attachments (post_id, filename, filepath, is_text)
-                    VALUES (?, ?, ?, ?)
-                    """,
-                    (post_id, filename, disk_path, is_text),
-                )
-                attachment_id = cur.lastrowid
-                conn.execute(
-                    """
-                    INSERT INTO attachments_fts (attachment_id, post_id, filename, content)
-                    VALUES (?, ?, ?, ?)
-                    """,
-                    (attachment_id, post_id, filename, content if is_text else None),
-                )
+        _index_attachments(conn, post_id, files, idx, base_dir, max_text_size)
        inserted_or_updated += 1

+    for post_id in existing:
+        if post_id not in seen_ids:
+            _delete_post(conn, post_id)
+            inserted_or_updated += 1
+
    st = os.stat(path)
    conn.execute(
        """
@@ -285,7 +392,7 @@ def upsert_file(conn: sqlite3.Connection, path: str) -> Tuple[int, int]:
        """,
        (path, st.st_mtime_ns, st.st_size, dt.datetime.now(dt.timezone.utc).isoformat()),
    )
-    return inserted_or_updated, skipped
+    return inserted_or_updated, skipped, unchanged


 def main() -> int:
@@ -302,6 +409,7 @@ def main() -> int:
    skipped_files = 0
    rows_upserted = 0
    rows_skipped = 0
+    rows_unchanged = 0

    with conn:
        for path in files:
@@ -310,9 +418,10 @@ def main() -> int:
                skipped_files += 1
                continue
            processed_files += 1
-            upserted, skipped = upsert_file(conn, path)
+            upserted, skipped, unchanged = upsert_file(conn, path, force=args.reindex_all)
            rows_upserted += upserted
            rows_skipped += skipped
+            rows_unchanged += unchanged

        users_json = mm_users.find_users_json(args.root)
        if users_json:
@@ -324,11 +433,13 @@ def main() -> int:

    print(f"DB: {args.db}")
    print(f"Files scanned: {len(files)} (processed={processed_files}, unchanged_skipped={skipped_files})")
-    print(f"Rows upserted: {rows_upserted}, invalid_posts_skipped: {rows_skipped}")
+    print(
+        f"Rows upserted: {rows_upserted}, unchanged_posts: {rows_unchanged}, "
+        f"invalid_posts_skipped: {rows_skipped}"
+    )
    print(f"Totals: posts={total_posts}, sources={total_sources}")
    return 0


 if __name__ == "__main__":
    raise SystemExit(main())
-
@@ -1,2 +1,4 @@
 cd /home/grayhook/Documents/mattermost-dl && .venv/bin/python mattermost-dl.py --team esr-daily --channel all
-cd /home/grayhook/Documents/mattermost-dl && .venv/bin/python mm_index.py --reindex-all
+cd /home/grayhook/Documents/mattermost-dl && .venv/bin/python mattermost-dl.py --team esr-dev-and-qa --channel QoSd
+cd /home/grayhook/Documents/mattermost-dl && .venv/bin/python mattermost-dl.py --team esr-dev-and-qa --channel VSAT
+cd /home/grayhook/Documents/mattermost-dl && .venv/bin/python mm_index.py