防止状态持久化

DeepSeek V3 中英对照 Preventing State Persistence

默认情况下，所有 ItemReader 和 ItemWriter 实现都会在提交前将其当前状态存储在 ExecutionContext 中。然而，这并非总是期望的行为。例如，许多开发者选择通过使用处理指示器使其数据库读取器具备"可重新运行"的能力。他们会在输入数据中添加一个额外列，用于指示数据是否已被处理。当读取（或写入）特定记录时，处理标志会从 false 翻转为 true。随后可以在 SQL 语句的 where 子句中添加额外条件，例如 where PROCESSED_IND = false，从而确保在重启时仅返回未处理的记录。在此场景中，最好不存储任何状态（例如当前行号），因为重启时这些状态无关紧要。因此，所有读取器和写入器都包含 saveState 属性。

Java

以下 bean 定义展示了如何在 Java 中防止状态持久化：

@Bean
public JdbcCursorItemReader playerSummarizationSource(DataSource dataSource) {
	return new JdbcCursorItemReaderBuilder<PlayerSummary>()
				.dataSource(dataSource)
				.rowMapper(new PlayerSummaryMapper())
				.saveState(false)
				.sql("SELECT games.player_id, games.year_no, SUM(COMPLETES),"
				  + "SUM(ATTEMPTS), SUM(PASSING_YARDS), SUM(PASSING_TD),"
				  + "SUM(INTERCEPTIONS), SUM(RUSHES), SUM(RUSH_YARDS),"
				  + "SUM(RECEPTIONS), SUM(RECEPTIONS_YARDS), SUM(TOTAL_TD)"
				  + "from games, players where players.player_id ="
				  + "games.player_id group by games.player_id, games.year_no")
				.build();

}

XML

以下 bean 定义展示了如何在 XML 中防止状态持久化：

<bean id="playerSummarizationSource" class="org.spr...JdbcCursorItemReader">
    <property name="dataSource" ref="dataSource" />
    <property name="rowMapper">
        <bean class="org.springframework.batch.samples.PlayerSummaryMapper" />
    </property>
    <property name="saveState" value="false" />
    <property name="sql">
        <value>
            SELECT games.player_id, games.year_no, SUM(COMPLETES),
            SUM(ATTEMPTS), SUM(PASSING_YARDS), SUM(PASSING_TD),
            SUM(INTERCEPTIONS), SUM(RUSHES), SUM(RUSH_YARDS),
            SUM(RECEPTIONS), SUM(RECEPTIONS_YARDS), SUM(TOTAL_TD)
            from games, players where players.player_id =
            games.player_id group by games.player_id, games.year_no
        </value>
    </property>
</bean>

上面配置的 ItemReader 不会在其参与的任何执行过程中向 ExecutionContext 写入任何条目。