TODOLIST

[CorvusYe]

TODOLIST

• [x] Change nebula-client dependence from 3.0.0 to 3.3.0

• [x] Use nebula-client SessionPool

• [ ] Move unit test from ngbatis-demo to ngbatis

• [ ] Add batch interface in NebulaDaoBasic :

  • [ ] insertTripletBatch
  • [x] insertEdgeBatch
  • [ ] show metas

• [ ] ResultSetUtil more column types support

  • [ ] Geography
  • [x] Duration

• [x] Vertex multi labels support.

• [ ] i18n: docs / comments / examples

• [ ] Performance optimization of nGQL in NebulaDaoBasic

  • [ ] selectByIds

    MATCH(n) WHERE id(n) == <id> RETURN n
    

  • [ ] selectByIds

    MATCH(n: <tag_name>)
    WHERE id(n) in <id_list>
    RETURN n
    

  • [ ] selectBySelective

    MATCH(n: <tag_name>)
    WHERE 1 == 1 <and id(n) == <id>>
    [ and n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN n
    LIMIT 4000
    

  • [ ] selectBySelectiveStringLike

    MATCH(n: <tag_name>)
    WHERE 1 == 1 <and id(n) == <id>>
    [ and n.`<tag_name>`.<prop_name> =~ '.*<prop_value>.*' ]
    RETURN n
    LIMIT 4000
    

  • [ ] selectIdBySelective

    MATCH(n: <tag_name>)
    WHERE 1 == 1 <and id(n) == <id>>
    [ and n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN id(n)
    LIMIT 4000
    

  • [ ] selectIdBySelectiveStringLike

    MATCH(n: <tag_name>)
    WHERE 1 == 1 <and id(n) == <id>>
    [ and n.`<tag_name>`.<prop_name> =~ '.*<prop_value>.*' ]
    RETURN id(n)
    LIMIT 4000
    

  • [ ] selectByMap

    MATCH (n:`${ tagName }`)
    WHERE 1 == 1
    [ and n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN n
    LIMIT 4000
    

  • [ ] countByMap

    MATCH (n:`${ tagName }`)
    WHERE 1 == 1
    [ and n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN count(n)
    

  • [ ] selectPage

    MATCH (n:${ tag })
    WHERE
      [ [ and ] n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN n skip <start_row>limit <page_size>
    

  • [ ] countPage

    MATCH (n:`${ tagName }`)
    WHERE
      [ [ and ] n.`<tag_name>`.<prop_name> == <prop_value> ]
    RETURN n skip <start_row>limit <page_size>
    

  • [ ] insert

    INSERT VERTEX IF NOT EXISTS `<tag_name>` (
        <prop_name_group>
    )
    VALUES 
    <id> : ( <prop_value_group> )
    

  • [ ] insertBatch

    INSERT VERTEX IF NOT EXISTS `<tag_name>` (
        <prop_name_group>
    )
    VALUES 
    [<id> : ( <prop_value_group> ) [,] ]
    

  • [ ] updateById 、updateByIdSelective

    UPDATE  VERTEX ON `<tag_name>` <id>
    [
        SET
        [ <prop_group> ]
    ]
    YIELD <id> as <primary_key_name>, <prop_name_group>
    

  • [ ] updateByIdBatchSelective

    [
        UPDATE  VERTEX ON `<tag_name>` <id>
        SET
        [ <prop_name> = <prop_value> [,] ]
    ]
    

  • [ ] deleteWithEdgeById

    DELETE VERTEX <v_id> WITH EDGE
    

  • [ ] deleteById

    DELETE VERTEX <v_id>
    

  • [ ] insertEdge insertEdgeSelective

    INSERT EDGE `<edge_name>` (
         <prop_name_group>
    )
    VALUES <v_id_1>-> <v_id_2> @<ranking> :(
         <prop_value_group>
    );
    

  • [ ] upsertEdgeSelective

    UPSERT EDGE ON `<tag_name>`
    <v_id_1> -> <v_id_2> @[ranking]
    [ 
       SET
           [ `<prop_name>` = <prop_value> [, ] ]
    ]
    

  • [ ] existsEdge

    MATCH (n)-[r: <edge_name>]-(n2)
    WHERE id(n) == $p0 AND id(n2) == <v_id_2>
    RETURN count(*) > 0
    

  • [ ] listStartNodes

    MATCH (n: `<tag_name>`)-[r: `<edge_name>`]->(n2)
    WHERE id(n2) == <v_id_2>
    RETURN n
    

  • [ ] startNode

    MATCH (n: `<tag_name>`)-[r: `<edge_name>`]->(n2)
    WHERE id(n2) == <v_id_2>
    RETURN n
    LIMIT 1
    

[soul-gin]

insertEdgeBatch 如果扩展 insertEdge(Object v1, Object e, Object v2) 似乎不太好处理, 是否考虑使用类似 NgEdge<I> 的对象结构(类似离线批处理, vid srcVid dstVid rank为一类, 其他column为一类), 并在 KvFn.KV 中添加 vid srcVid dstVid rank 属性? 在定义java bean中, 是否需要使用自定义注解来区分, 根据@Key 中name值(指定常量)对KvFn.KV进行赋值?

[CorvusYe]

@soul-gin 连同上面提的方案，大概有三种方案可以参考：

// 方案1
insertEdgeBatch( List<E> edges ); // 使用注解并拓展KvFn

// 方案2
insertEdgeBatch( List<?> srcIdList, List<E> edges, List<?> dstIdList );

// 方案3，考虑与  insertTripletBatch 统一入参
class NgTriplet<ID> {
    ID srcId;
    ID dstId;
    Object start;
    Object edge;
    Object end;
}
// 插入 start edge end 三元组数据
insertTripletBatch( List<NgTriplet> triples ); 
// 仅插入 edge，节点数据使用 srcId 或 start.getId，与 insertEdge 参数的逻辑保持一致
insertEdgeBatch( List<NgTriplet> triples ); 

以上几个方案还是让人比较纠结的。

[soul-gin]

方案3对目前代码侵入性小,能复用insertTripletBatch确实是个不错选择,NgTriplet是否还需要rank属性,还是通过注解获取? 方案1相比较侵入性大些,在某些场景下更适用(src, dst数据来源与edge不同, 只是根据一份数据插入多条边).

[CorvusYe]

方案三以ngbatis-demo里的实为例：

    NgTriplet triplet = new NgTriplet();
    triplet.setStart( new Person("Tom") );  // setStart 或 setSrcId 二者选其一，与 insertEdge 方法的入参逻辑一致
    triplet.setEdge( new Like() );
    triplet.setEnd( new Person("Jerry") ); // setEnd 或 setDstId 二选一

这样的元组构成集合，实体不需要重新再定义，之前的注解也不需要有所变动，直接可以复用。 同时构成的入参，批增点跟批增边 形式一致，api 的熟悉成本更低。