集合运算符¶

合并多个请求时，可以使用集合运算符，包括UNION、UNION ALL、INTERSECT和MINUS。

所有集合运算符的优先级相同，如果一个 nGQL 语句中有多个集合运算符，NebulaGraph 会从左到右进行计算，除非用括号指定顺序。

Caution

集合运算符前后的查询语句中定义的变量名及顺序必需保持一致，例如RETURN a,b,c UNION RETURN a,b,c中的a,b,c的名称及顺序需要保持一致。

UNION、UNION DISTINCT、UNION ALL¶

<left> UNION [DISTINCT | ALL] <right> [ UNION [DISTINCT | ALL] <right> ...]

运算符UNION DISTINCT（或使用缩写UNION）返回两个集合 A 和 B 的并集，不包含重复的元素。

运算符UNION ALL返回两个集合 A 和 B 的并集，包含重复的元素。

left和right必须有相同数量的列和数据类型。如果需要转换数据类型，请参见类型转换。

示例¶

# 返回两个查询结果的并集，不包含重复的元素。
nebula> GO FROM "player102" OVER follow YIELD dst(edge) \
        UNION \
        GO FROM "player100" OVER follow YIELD dst(edge);
+-------------+
| dst(EDGE)   |
+-------------+
| "player100" |
| "player101" |
| "player125" |
+-------------+

# 查询 Tag 为 player 的点，根据名称排序后获取前 3 条数据，并与数组合并返回，不包含重复的元素。
nebula> MATCH (v:player) \
        WITH v.player.name AS n \
        RETURN n ORDER BY n LIMIT 3 \
        UNION \
        UNWIND ["Tony Parker", "Ben Simmons"] AS n \
        RETURN n;
+---------------------+
| n                   |
+---------------------+
| "Amar'e Stoudemire" |
| "Aron Baynes"       |
| "Ben Simmons"       |
| "Tony Parker"       |
+---------------------+

# 返回两个查询结果的并集，包含重复的元素。
nebula> GO FROM "player102" OVER follow YIELD dst(edge) \
        UNION ALL \
        GO FROM "player100" OVER follow YIELD dst(edge);
+-------------+
| dst(EDGE)   |
+-------------+
| "player100" |
| "player101" |
| "player101" |
| "player125" |
+-------------+

# 查询 Tag 为 player 的点，根据名称排序后获取前 3 条数据，并与数组合并返回，包含重复的元素。
nebula> MATCH (v:player) \
        WITH v.player.name AS n \
        RETURN n ORDER BY n LIMIT 3 \
        UNION ALL \
        UNWIND ["Tony Parker", "Ben Simmons"] AS n \
        RETURN n;
+---------------------+
| n                   |
+---------------------+
| "Amar'e Stoudemire" |
| "Aron Baynes"       |
| "Ben Simmons"       |
| "Tony Parker"       |
| "Ben Simmons"       |
+---------------------+

# UNION 也可以和 YIELD 语句一起使用，去重时会检查每一行的所有列，每列都相同时才会去重。
nebula> GO FROM "player102" OVER follow \
        YIELD dst(edge) AS id, properties(edge).degree AS Degree, properties($$).age AS Age \
        UNION /* DISTINCT */ \
        GO FROM "player100" OVER follow \
        YIELD dst(edge) AS id, properties(edge).degree AS Degree, properties($$).age AS Age;
+-------------+--------+-----+
| id          | Degree | Age |
+-------------+--------+-----+
| "player100" | 75     | 42  |
| "player101" | 75     | 36  |
| "player101" | 95     | 36  |
| "player125" | 95     | 41  |
+-------------+--------+-----+

INTERSECT¶

<left> INTERSECT <right>

运算符INTERSECT返回两个集合 A 和 B 的交集。

left和right必须有相同数量的列和数据类型。如果需要转换数据类型，请参见类型转换。

示例¶

# 返回两个查询结果的交集。
nebula> GO FROM "player102" OVER follow \
        YIELD dst(edge) AS id, properties(edge).degree AS Degree, properties($$).age AS Age \
        INTERSECT \
        GO FROM "player100" OVER follow \
        YIELD dst(edge) AS id, properties(edge).degree AS Degree, properties($$).age AS Age;
+----+--------+-----+
| id | Degree | Age |
+----+--------+-----+
+----+--------+-----+

# 返回 player102 的邻居和边数据与 player100 的邻居和边数据之间的交集。
nebula> MATCH (v:player)-[e:follow]->(v2) \
        WHERE id(v) == "player102" \
        RETURN id(v2) As id, e.degree As Degree, v2.player.age AS Age \
        INTERSECT \
        MATCH (v:player)-[e:follow]->(v2) \
        WHERE id(v) == "player100" \
        RETURN id(v2) As id, e.degree As Degree, v2.player.age AS Age;
+----+--------+-----+
| id | Degree | Age |
+----+--------+-----+
+----+--------+-----+

# 返回 [1,2] 与 [1,2,3,4] 的交集。
nebula> UNWIND [1,2] AS a RETURN a \
        INTERSECT \
        UNWIND [1,2,3,4] AS a \
        RETURN a;
+---+
| a |
+---+
| 1 |
| 2 |
+---+

MINUS¶

<left> MINUS <right>

运算符MINUS返回两个集合 A 和 B 的差异，即A-B。请注意left和right的顺序，A-B表示在集合 A 中，但是不在集合 B 中的元素。

示例¶

# 返回在第一个查询结果中，但是不在第二个查询结果中的元素。
nebula> GO FROM "player100" OVER follow YIELD dst(edge) \
        MINUS \
        GO FROM "player102" OVER follow YIELD dst(edge);
+-------------+
| dst(EDGE)   |
+-------------+
| "player125" |
+-------------+

# 返回在 player102 的邻居中，但不在 player100 的邻居中的元素。
nebula> GO FROM "player102" OVER follow YIELD dst(edge) AS id \
        MINUS \
        GO FROM "player100" OVER follow YIELD dst(edge) AS id;
+-------------+
| id          |
+-------------+
| "player100" |
+-------------+

# 返回在 player102 的邻居中，但不在 player100 的邻居中的元素。
nebula> MATCH (v:player)-[e:follow]->(v2) \
        WHERE id(v) =="player102" \
        RETURN id(v2) AS id\
        MINUS \
        MATCH (v:player)-[e:follow]->(v2) \
        WHERE id(v) =="player100" \
        RETURN id(v2) AS id;
+-------------+
| id          |
+-------------+
| "player100" |
+-------------+

# 返回 [1,2,3] 中不与 4 相同的元素。
nebula> UNWIND [1,2,3] AS a RETURN a \
        MINUS \
        WITH 4 AS a \
        RETURN a;
+---+
| a |
+---+
| 1 |
| 2 |
| 3 |
+---+

集合运算符和管道符的优先级¶

当查询包含集合运算符和管道符（|）时，管道符的优先级高。例如GO FROM 1 UNION GO FROM 2 | GO FROM 3相当于GO FROM 1 UNION (GO FROM 2 | GO FROM 3)。

示例¶

nebula> GO FROM "player102" OVER follow \
        YIELD dst(edge) AS play_dst  \
        UNION \
        GO FROM "team200" OVER serve REVERSELY \
        YIELD src(edge) AS play_src \
        | GO FROM $-.play_src OVER follow YIELD dst(edge) AS play_dst;

+-------------+
| play_dst    |
+-------------+
| "player100" |
| "player101" |
| "player117" |
| "player105" |
+-------------+

该查询会先执行红框内的语句，然后执行绿框的UNION操作。

圆括号可以修改执行的优先级，例如：

nebula> (GO FROM "player102" OVER follow \
        YIELD dst(edge) AS play_dst  \
        UNION \
        GO FROM "team200" OVER serve REVERSELY \
        YIELD src(edge) AS play_dst) \
        | GO FROM $-.play_dst OVER follow YIELD dst(edge) AS play_dst;

该查询中，圆括号包裹的部分先执行，即先执行UNION操作，再将结果结合管道符进行下一步操作。

最后更新: April 15, 2024