java - 多次高效地使用 Prepared Statement

Question

下面是我使用准备语句在 Oracle 数据库中插入多条记录(大约 5000-7000) 的代码。

我现在的做法好吗？还是可以使用一些批处理工具进一步改进？

pstatement = db_connection.prepareStatement(PDSLnPConstants.UPSERT_SQL);

for (Entry<Integer, LinkedHashMap<Integer, String>> entry : MAPPING.entrySet()) {

    pstatement.setInt(1, entry.getKey());
    pstatement.setString(2, entry.getValue().get(LnPConstants.CGUID_ID));
    pstatement.setString(3, entry.getValue().get(LnPConstants.PGUID_ID));
    pstatement.setString(4, entry.getValue().get(LnPConstants.SGUID_ID));
    pstatement.setString(5, entry.getValue().get(LnPConstants.UID_ID));
    pstatement.setString(6, entry.getValue().get(LnPConstants.ULOC_ID));
    pstatement.setString(7, entry.getValue().get(LnPConstants.SLOC_ID));
    pstatement.setString(8, entry.getValue().get(LnPConstants.PLOC_ID));
    pstatement.setString(9, entry.getValue().get(LnPConstants.ALOC_ID));
    pstatement.setString(10, entry.getValue().get(LnPConstants.SITE_ID));
    pstatement.executeUpdate();

    pstatement.clearParameters();
}

我正在使用的更新代码:-

public void runNextCommand() {

    Connection db_connection = null;
    PreparedStatement pstatement = null;
    int batchLimit = 1000;
    boolean autoCommit = false;

    try {
        db_connection = getDBConnection();

        autoCommit = db_connection.getAutoCommit();
        db_connection.setAutoCommit(false); //Turn off autoCommit
        pstatement = db_connection.prepareStatement(LnPConstants.UPSERT_SQL); // create a statement

        for (Entry<Integer, LinkedHashMap<Integer, String>> entry : GUID_ID_MAPPING.entrySet()) {
            pstatement.setInt(1, entry.getKey());
            pstatement.setString(2, entry.getValue().get(LnPConstants.CGUID_ID));
            pstatement.setString(3, entry.getValue().get(LnPConstants.PGUID_ID));
            pstatement.setString(4, entry.getValue().get(LnPConstants.SGUID_ID));
            pstatement.setString(5, entry.getValue().get(LnPConstants.UID_ID));
            pstatement.setString(6, entry.getValue().get(LnPConstants.ULOC_ID));
            pstatement.setString(7, entry.getValue().get(LnPConstants.SLOC_ID));
            pstatement.setString(8, entry.getValue().get(LnPConstants.PLOC_ID));
            pstatement.setString(9, entry.getValue().get(LnPConstants.ALOC_ID));
            pstatement.setString(10, entry.getValue().get(LnPConstants.SITE_ID));
            pstatement.addBatch();

            batchLimit--;

            if(batchLimit == 0){
                pstatement.executeBatch();
                pstatement.clearBatch();
                batchLimit = 1000;
            }
            pstatement.clearParameters();
        }

    } catch (SQLException e) {
        getLogger().log(LogLevel.ERROR, e);
    } finally {
        try {
            pstatement.executeBatch();
            db_connection.commit();
            db_connection.setAutoCommit(autoCommit);
        } catch (SQLException e1) {
            getLogger().log(LogLevel.ERROR, e1.getMessage(), e1.fillInStackTrace());
        }

        if (pstatement  != null) {
            try {
                pstatement.close();
                pstatement = null;
            } catch (SQLException e) {
                getLogger().log(LogLevel.ERROR, e.getMessage(), e.fillInStackTrace());
            }
        }
        if (db_connection!= null) {
            try {
                db_connection.close();
                db_connection = null;
            } catch (SQLException e) {
                getLogger().log(LogLevel.ERROR, e.getMessage(), e.fillInStackTrace());
            }
        }
    }
}

Answer 1

可以考虑使用addBatch()并一次执行后面的语句。此外，正如@pst 在您的问题中评论的那样，请考虑使用 trasaction。

你会做的方式如下:

boolean autoCommit = connection.getAutoCommit();
try{
    connection.setAutoCommit(false //Turn off autoCommit
    pstatement = db_connection.prepareStatement(PDSLnPConstants.UPSERT_SQL);

    int batchLimit = 1000;

    try{
        for (Entry<Integer, LinkedHashMap<Integer, String>> entry : MAPPING.entrySet()) {
            pstatement.setInt(1, entry.getKey());
            pstatement.setString(2, entry.getValue().get(LnPConstants.CGUID_ID));
            pstatement.setString(3, entry.getValue().get(LnPConstants.PGUID_ID));
            pstatement.setString(4, entry.getValue().get(LnPConstants.SGUID_ID));
            pstatement.setString(5, entry.getValue().get(LnPConstants.UID_ID));
            pstatement.setString(6, entry.getValue().get(LnPConstants.ULOC_ID));
            pstatement.setString(7, entry.getValue().get(LnPConstants.SLOC_ID));
            pstatement.setString(8, entry.getValue().get(LnPConstants.PLOC_ID));
            pstatement.setString(9, entry.getValue().get(LnPConstants.ALOC_ID));
            pstatement.setString(10, entry.getValue().get(LnPConstants.SITE_ID));
            
            pstatement.addBatch();
            batchLimit--;
            
            if(batchLimit == 0){
                pstatement.executeBatch();
                pstatement.clearBatch();
                batchLimit = 1000;
            }
             pstatement.clearParameters();
        }
    }finally{
        //for the remaining ones
        pstatement.executeBatch();
        
        //commit your updates
        connection.commit();
    }
}finally{
    connection.setAutoCommit(autoCommit);
}

这个想法是为批量更新设置一个限制，只有当你达到一个特定的限制时才执行数据库更新。通过这种方式，您可以将数据库调用限制为您定义的每个 batchLimit 一次。这样会更快。

另请注意事务，我刚刚展示了如何以及何时提交。这可能并不总是正确的 commit 点，因为这个决定将基于您的要求。您可能还想在出现异常时执行 rollback。因此，由您决定。

看看"Using Transaction"教程以更好地了解如何使用 transaction。