前提条件介绍

GraphCompute 支持将MaxCompute 表批量导入，但是这个过程中涉及到读取用户MaxCompute表的操作，所以需要用户提前授权GraphCompute公共账号有读取用户MaxCompute表的权限，具体的操作步骤如下所示：

-- 1. 切换到用户MaxCompute表所在的project
use project ***;
-- 2. 给MaxGraph公共账号进行授权
add user ALIYUN$big_graph@aliyun-inner.com;
GRANT Describe,Select ON TABLE table_name TO USER ALIYUN$big_graph@aliyun-inner.com;

使用步骤及示例

使用SDK进行图数据的导入，需要先完成Client创建和Schema构建。

在此基础上， 本文介绍如何初始化mapping信息（maxCompute表的一行记录中的哪个字段写到graphCompute 点/边的哪个属性上）及bulk load的一些配置信息；

步骤1，准备好源数据

我们假设maxCompute上存在如下数据，可作为图模型的源数据。

maxCompute表person结构如下，可以作为GraphCompute中图模型的点person的源数据

maxCompute表knows结构如下，可以作为GraphCompute中图模型的点knows的源数据

步骤2，指定对应关系

用户需要在sdk中做好图属性和MaxCompute表项的对应关系。代码示例如下：

private String odpsEndpoint = "***";
private String yourOdpsAccessId = "***";
private String yourOdpsAccessKey = "**";
private String yourBizId = "**";
private String yourOdpsProject = "**";
// odps点表的table name
private String yourOdpsVertexTable = "**";
// odps边表的table name
private String yourOdpsEdgeTable = "**";
// odps中点表的colume
private String odpsPersonIdField = "**";
private String odpsPersonNameField = "**";
private String odpsPersonAgeField = "**";
// maxgraph 中点的属性
private String personIdProp = "**";
private String personNameProp = "**";
private String personAgeProp = "**";
// odps 中边表的colume
private String odpsKnowsIdField = "**";
private String odpsKnowsWeightField = "**";
// maxgaph 中边的属性
private String knowsIdProp = "**";
private String knowsWeightProp = "**";
// maxgraph 中边的起点主键和终点主键
private String knowsSrcPersonIdField = "**";
private String knowsDstPersonIdField = "**";

步骤3，从MaxCompute表批量导入点数据到person中

ResultSet resultSet = client.submit("graph.bulkloadVertexFromOdps('person')" +
        ".endpoint('" + odpsEndpoint + "')" +
        ".accessId('" + yourOdpsAccessId + "')" +
        ".accessKey('" + yourOdpsAccessKey + "')" +
        ".bizOwnerId('" + yourBizId + "')" +
        ".project('" + yourOdpsProject + "')" +
        ".table('" + yourOdpsVertexTable + "')" +
        //".partition('" + yourOdpsPartition + "')" +
        //".maxInvalidDataCount(10)" + // Default invalid data count is 0
        ".mappingColumn('" + odpsPersonIdField + "', '" + personIdProp + "')" +
        // mappingColumn 定义了将odps表的一行记录的colume与maxgraph中点的属性的映射关系
        ".mappingColumn('" + odpsPersonNameField + "', '" + personNameProp + "')" +
        ".mappingColumn('" + odpsPersonAgeField + "', '" + personAgeProp + "')" +
        ".signature('" + getCredentialsManager().getPassword() + "')");

步骤4，从MaxCompute表批量导入边数据knows中

ResultSet resultSet = client.submit("graph.bulkloadEdgeFromOdps('knows')" +
        ".endpoint('" + odpsEndpoint + "')" +
        ".accessId('" + yourOdpsAccessId + "')" +
        ".accessKey('" + yourOdpsAccessKey + "')" +
        ".bizOwnerId('" + yourBizId + "')" +
        ".project('" + yourOdpsProject + "')" +
        ".table('" + yourOdpsEdgeTable + "')" +
        //".partition('" + yourOdpsPartition + "')" +
        //".maxInvalidDataCount(10)" + // Default invalid data count is 0
        ".mappingColumn('" + odpsKnowsIdField + "', '" + knowsIdProp + "')" +
        ".mappingColumn('" + odpsKnowsWeightField + "', '" + knowsWeightProp + "')" +
        ".srcVertex('person')" +
        ".mappingSrcPrimaryKey('" + knowsSrcPersonIdField + "','" + personIdProp + "')" +
        ".dstVertex('person')" +
        ".mappingDstPrimaryKey('" + knowsDstPersonIdField + "','" + personIdProp + "')" +
        ".signature('" + getCredentialsManager().getPassword() + "')");