优化 SQLite 内连接查询 - Android

Question

我正在与 GTFS 合作，并试图找出一种方法来加快我的内部连接。表格“Trips”和“StopTimes”有一个公共列“TripID”，表格“StopTimes”和“Stops”有一个公共列“StopID”。内连接检索与某个行程对应的所有停靠点，但是它需要大约 20 秒。此外，“StopTimes”表包含 900 万行，而其他表包含几千行。以下代码是我的查询。

String joinQuery = "SELECT s.stop_id, s.stop_code, s.stop_name "
+ "FROM Trips as t INNER JOIN StopTimes as st ON     st.trip_id = t.trip_id "
+ "INNER JOIN Stops as s ON s.stop_id = st.stop_id "
+ "WHERE t.trip_id = " + TripID + " AND t.shape_id = " + ShapeID
+ " ORDER BY st.stop_sequence";

这是创建数据库时执行的代码。

String CREATE_TRIPS_TABLE = "CREATE VIRTUAL TABLE " + TABLE_TRIPS + " USING FTS3("
            + KEY_T_ROUTE_ID + " INTEGER,"
            + KEY_T_NAME + " TEXT,"
            + KEY_TRIP_ID + " INTEGER,"
            + KEY_DIRECTION_ID + " INTEGER,"
            + KEY_SHAPE_ID + " INTEGER);";

    String CREATE_STOPS_TABLE = "CREATE VIRTUAL TABLE " + TABLE_STOPS + " USING FTS3("
            + KEY_STOP_ID + " INTEGER,"
            + KEY_STOP_CODE + " TEXT,"
            + KEY_STOP_NAME + " TEXT,"
            + KEY_STOP_LAT + " DOUBLE,"
            + KEY_STOP_LON + " DOUBLE,"
            + KEY_STOP_WHEELCHAIR + " INTEGER);";

    String CREATE_STOPTIMES_TABLE = "CREATE VIRTUAL TABLE " + TABLE_STOP_TIMES + " USING FTS3("
            + KEY_S_TRIP_ID + " INTEGER,"
            + KEY_S_STOP_ID + " INTEGER,"
            + KEY_ARRIVAL_TIME + " TEXT,"
            + KEY_STOP_SEQUENCE + " INTEGER,"
            + KEY_SHAPE_DIST_TRAVELLED + " DOUBLE);";

我遵循了我在这里找到的一些建议，但仍然无法加快我的查询速度，感谢任何反馈。

更新 我更改了创建表的方式，并为 INNER JOIN 查询中涉及的两个索引添加了索引，现在 JOINS 是即时的。

String CREATE_TRIPS_TABLE = "CREATE TABLE " + TABLE_TRIPS + "("
            + KEY_T_ID + " integer PRIMARY KEY AUTOINCREMENT,"
            + KEY_T_ROUTE_ID + " INTEGER,"
            + KEY_T_NAME + " TEXT,"
            + KEY_TRIP_ID + " INTEGER,"
            + KEY_DIRECTION_ID + " INTEGER,"
            + KEY_SHAPE_ID + " INTEGER);";

    String CREATE_STOPS_TABLE = "CREATE TABLE " + TABLE_STOPS + "("
            + KEY_S_ID + " integer PRIMARY KEY AUTOINCREMENT,"
            + KEY_STOP_ID + " INTEGER,"
            + KEY_STOP_CODE + " TEXT,"
            + KEY_STOP_NAME + " TEXT,"
            + KEY_STOP_LAT + " DOUBLE,"
            + KEY_STOP_LON + " DOUBLE,"
            + KEY_STOP_WHEELCHAIR + " INTEGER);";

    String CREATE_STOPTIMES_TABLE = "CREATE TABLE " + TABLE_STOP_TIMES + "("
            + KEY_ST_ID + " integer PRIMARY KEY AUTOINCREMENT,"
            + KEY_S_TRIP_ID + " INTEGER,"
            + KEY_S_STOP_ID + " INTEGER,"
            + KEY_ARRIVAL_TIME + " TEXT,"
            + KEY_STOP_SEQUENCE + " INTEGER,"
            + KEY_SHAPE_DIST_TRAVELLED + " DOUBLE);";

    db.execSQL(CREATE_ROUTES_TABLE);
    db.execSQL(CREATE_TRIPS_TABLE);
    db.execSQL(CREATE_STOPS_TABLE);
    db.execSQL(CREATE_STOPTIMES_TABLE);
    db.execSQL("CREATE INDEX T_id ON Trips(trip_id)");
    db.execSQL("CREATE INDEX S_t_id ON StopTimes(trip_id)");

Answer 1

我看到您正在使用SQLite's FTS3 extension，它旨在提高全文搜索的性能。我怀疑这实际上不利于您发出的那种查询（以及通常针对 GTFS 数据运行的那种查询）。

我建议您首先创建一个传统的关系数据库，在适当的列上使用索引，然后测试查询的性能。在这种类型的应用程序中使用时，SQLite 无疑具有良好的性能，所以我想你会感到惊喜。一旦您证明使用传统技术取得了成功，您就可以在必要时寻找其他方法使查询运行得更快。

最后，请注意SQLite does not allow indices on virtual tables，我怀疑这是您的查询现在需要这么长时间才能完成的具体原因。