Is it mandatory to use a transaction when retrieving data from a database?

Using transactions when retrieving data from a database is not always necessary, but it depends on the specific situation and data consistency requirements.

A transaction ensures atomicity and isolation of operations. For simple data reads where guaranteeing a consistent state during concurrent modifications is not required, it is possible to omit explicit transactions — many DBMSs perform SELECTs outside of transactions or in autocommit mode by default.

However, if you need to obtain a consistent snapshot of data, for example, during complex queries or when data may change concurrently, it is recommended to use a transaction with an appropriate isolation level (such as REPEATABLE READ or SERIALIZABLE) to avoid issues with non-repeatable reads or phantom records.

In Java, using JDBC, you can explicitly start a transaction by disabling autocommit:

connection.setAutoCommit(false);
try {
    // perform SELECT
    connection.commit();
} catch (SQLException e) {
    connection.rollback();
}

In summary, a transaction during reading is needed if data consistency is important under concurrent access conditions.