nebula 版本：1.0
背景： 我们这边需要做1.0升级到3.6版本，发现相同的数据1.0和3.6相同的nGQL不一致，经过排查是1.0的执行语句返回结果有问题，也就是线上依赖的数据是错误的，但是貌似升级成3.6数据是对了，但是可能依赖错误数据的程序就有问题
发帖目的： 搞清楚1.0是如何执行的，然后通过3.6曲线实现1.0的结果
bug说明： 在特定的场景下，
go from a,b over u yield u._dst as id | go from $-.dst_id over xxx
与
go from a over u yield u._dst as id | go from $-.dst_id over xxx
union all
go from b over u yield u._dst as id | go from $-.dst_id over xxx
不一致
bug复现
insert vertex user(uname,old) values 11:("T_A",0);
insert vertex user(uname,old) values 12:("T_B",0);
insert vertex user(uname,old) values 13:("T_C",0);
insert vertex user(uname,old) values 14:("T_D",0);
insert vertex user(uname,old) values 15:("T_E",0);
insert vertex user(uname,old) values 16:("T_F",0);
insert vertex ball(bname) values 21:("basktball"),22:("football");
insert edge like(pname) values 11->21:("a"),12->21:("b"),13->21:("c"),14->21:("d"),15->22("e");
insert edge like(pname) values 16->22:("h"),13->22:("g"),14->22:("f");
找出与11有共同爱好的人的列表：
GO FROM 11 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id
找出与12有共同爱好的人的列表：
GO FROM 12 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id
找出与11、12有共同爱好的人的列表：
GO FROM 11,12 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id | order by $-.dst_id ASC,$-.start_id ASC;
** 
结果错了，只包含了12的共同爱好的人列表，从结果上看感觉12检索了两次 **
将11和12调换顺序：
GO FROM 12,11 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id | order by $-.dst_id ASC,$-.start_id ASC;
结果显示，只包含了11的共同爱好的人列表
居然调换from的顺序都会影响结果
使用union all的方式重新查一次：
(GO FROM 11 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id union all GO FROM 12 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id) | order by $-.dst_id ASC,$-.start_id ASC;
结果正确，既包含了11，又包含了12
猜想： 这里因为11和12有共同的喜好，可能会导致问题，那么取不同的喜好的人一起查看是否会有问题？
11喜欢basktball，15喜欢football，这里就取11和15作为共同搜索
先查询15共同爱好的人列表：
GO FROM 15 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id
11和15有共同爱好的人列表：
GO FROM 15,11 OVER like YIELD like._dst AS dst_id, like._src as src_id | GO FROM $-.dst_id over like REVERSELY where $-.src_id != like._dst YIELD $-.src_id as start_id,like._dst as dst_id | order by $-.dst_id ASC,$-.start_id ASC;
** 结果正确**
结论： nebula 1.0在合并查询时如果from的点有共同喜好就会出现覆盖问题，并且后面的点覆盖之前的点的数据