fix(interactive): Update example movie graph and fix some bug (#3304)

Update the example movie graph's schema and raw data for interactive. Add some demo cypher queries. --------- Co-authored-by: BingqingLyu <[email protected]>
alibaba · Oct 23, 2023 · 7a7bd4e · 7a7bd4e
1 parent 43f5176
commit 7a7bd4e
Show file tree

Hide file tree

Showing 19 changed files with 214 additions and 77 deletions.
diff --git a/flex/engines/hqps_db/core/operator/limit.h b/flex/engines/hqps_db/core/operator/limit.h
@@ -41,6 +41,7 @@ class LimitOp {
         cur_offset += 1;
       }
       new_offsets.push_back(cur_offset);
+      cur_ind += 1;
     }
 
     std::vector<size_t> selected_indices;

diff --git a/flex/engines/hqps_db/core/operator/sink.h b/flex/engines/hqps_db/core/operator/sink.h
@@ -378,7 +378,7 @@ class SinkOp {
     }
   }
 
-  // sink collection of pod, expect for LabelKey type
+  // sink collection of pod type
   template <size_t Ind, size_t act_tag_id, typename T,
             typename std::enable_if<
                 (!gs::is_vector<T>::value) && (!gs::is_tuple<T>::value) &&

diff --git a/flex/engines/hqps_db/structures/multi_edge_set/flat_edge_set.h b/flex/engines/hqps_db/structures/multi_edge_set/flat_edge_set.h
@@ -416,6 +416,8 @@ class SingleLabelEdgeSetBuilder {
     vec_.push_back(std::get<1>(tuple));
   }
 
+  void Insert(const ele_tuple_t& tuple) { vec_.push_back(tuple); }
+
   result_t Build() {
     return result_t(std::move(vec_), std::move(label_triplet_), prop_names_,
                     direction_);

diff --git a/flex/interactive/examples/movies/0_get_user.cypher b/flex/interactive/examples/movies/0_get_user.cypher
@@ -0,0 +1 @@
+MATCH (u: User) Return u.id, u.name LIMIT 5;
diff --git a/flex/interactive/examples/movies/1_get_user_rated_movies.cypher b/flex/interactive/examples/movies/1_get_user_rated_movies.cypher
@@ -0,0 +1 @@
+MATCH (u: User) -[r:REVIEW]->(movie: Movie) WHERE r.rating > $rateThresh RETURN COUNT(u);
diff --git a/flex/interactive/examples/movies/2_user_colfilter.cypher b/flex/interactive/examples/movies/2_user_colfilter.cypher
@@ -0,0 +1,5 @@
+MATCH (u1: User) -[r1:REVIEW]->(movie: Movie)<-[r2: REVIEW]- (u2: User)
+WHERE u1.id > u2.id
+             AND r1.rating > $rateThresh
+             AND r2.rating > $rateThresh
+Return u1.name, u2.name, movie.title LIMIT 10;
diff --git a/flex/interactive/examples/movies/3_high_rate_ratio.cypher b/flex/interactive/examples/movies/3_high_rate_ratio.cypher
@@ -0,0 +1,6 @@
+MATCH (u: User)-[:REVIEW]->(m: Movie)
+WITH u, COUNT(m) as cnt1
+MATCH (u)-[r:REVIEW]->(likeM: Movie)
+WHERE r.rating > $rateThresh
+WITH u, cnt1, COUNT(likeM) as cnt2
+RETURN u.name, cnt2 / cnt1;
diff --git a/flex/interactive/examples/movies/4_high_rate_ratio_for_followed_actor.cypher b/flex/interactive/examples/movies/4_high_rate_ratio_for_followed_actor.cypher
@@ -0,0 +1,8 @@
+MATCH (u: User)-[:REVIEW]->(m: Movie)<-[:ACTED_IN]-(actor: Person),
+              (u)-[:FOLLOWS]->(actor)
+WITH DISTINCT u, COUNT(m) as cnt1
+MATCH (u: User)-[r:REVIEW]->(likeM: Movie)<-[:ACTED_IN]-(actor: Person)
+MATCH (u:User)-[:FOLLOWS]->(actor)
+WHERE r.rating > $rateThresh
+WITH DISTINCT u, cnt1, COUNT(likeM) as cnt2
+RETURN u.name, cnt2 / cnt1;
diff --git a/flex/interactive/examples/movies/5_recommend_rule.cypher b/flex/interactive/examples/movies/5_recommend_rule.cypher
@@ -0,0 +1,3 @@
+MATCH (u: User)-[:FOLLOWS]->(a: Person)-[:ACTED_IN]->(m: Movie)
+WHERE NOT (u : User)-[:REVIEW]->(m)
+RETURN u.name, m.title LIMIT 5;
diff --git a/flex/interactive/examples/movies/FOLLOWS.csv b/flex/interactive/examples/movies/FOLLOWS.csv
@@ -1,4 +1,9 @@
 start|end
-170|169
-168|169
-167|168
+169|71
+171|71
+172|71
+173|71
+174|71
+169|16
+172|16
+174|16
diff --git a/flex/interactive/examples/movies/Person.csv b/flex/interactive/examples/movies/Person.csv
@@ -129,6 +129,3 @@ id|born|name
 165|1963|Lori Petty
 166|1943|Penny Marshall
 167|2003|Paul Blythe
-168|2003|Angela Scope
-169|2003|Jessica Thompson
-170|2003|James Thompson
diff --git a/flex/interactive/examples/movies/README.md b/flex/interactive/examples/movies/README.md
@@ -0,0 +1,73 @@
+## Step 1
+(1) 用户探索
+```cypher
+MATCH (u: User)
+Return u.id, u.name;
+```
+
+
+## Step 2
+(2) 探索用户好评的电影，评价可能给一个阈值，通过一些探索，找到一个打好评的较好的阈值
+
+```cypher
+MATCH (u: User) -[r:REVIEWED]->(movie: Movie)
+WHERE r.rating > $rateThresh
+RETURN COUNT(u);
+```
+
+high_rate_number(0) 其实就是用户所有的评价数量
+high_rate_number(0.5) 好评threshold的占比
+high_rate_number(0.7)
+
+
+## Step 3
+(3) 探索两名用户同时给一个电影打好评
+```cypher
+MATCH (u1: User) -[r1:REVIEWED]->(movie: Movie)<-[r2: REVIEWED]- (u2: User)
+Return u1.name, u2.name, movie.name
+WHERE u1.id > u2.id
+             AND r1.rate > $rateThresh
+             AND r2.rate > $rateThresh;
+```
+
+但可能不是很有代表性，而且我们的数据集里并没有维护这两名用户直接的关系。
+我们继续探索可能的关联。
+
+## Step 4
+(4)探索是否用户会比较倾向于给自己关注的演员参演的电影打好评
+
+我们得到用户给一个电影打高分的可能性有大
+```cypher
+MATCH (u: User) -[:REVIEWED]->(m: Movie)
+WITH u, COUNT(m) as cnt1
+MATCH (u) - [r:REVIEWED]->(likeM: Movie)
+WHERE r.rate > $rateThresh
+WITH u, cnt1, COUNT(likeM) as cnt2
+RETURN u.name, cnt2 / cnt1
+```
+
+
+在这个电影是该用户关注的演员参演的情况下，我们再次计算这个可能性
+```cypher
+MATCH (u: User) -[:REVIEWED]->(m: Movie)<-[:ACTED_IN]-(actor: Person),
+              (u) -[:FOLLOWS]-(actor)
+WITH DISTINCT u, COUNT(m) as cnt1
+MATCH (u) - [r:REVIEWED]->(likeM: Movie)<-[:ACTED_IN]-(actor: Person),
+              (u) -[:FOLLOWS]-(actor)
+WHERE r.rate > $rateThresh
+WITH DISTINCT u, cnt1, COUNT(likeM) as cnt2
+RETURN u.name, cnt2 / cnt1
+```
+
+
+## Step 5
+(5)那么我们搜索这样的推荐规则：
+         给用户u推荐电影m，如果
+         （1）u关注演员a
+         （2）a参演了电影m
+         （3）u还没有看过电影m （还未做评价）
+```cypher
+MATCH (u: User) -[:FOLLOWS] -> (a: Person) -[: ACTED_IN] -> (m: Movie),
+WHERE NOT((u) -[:REVIEW] -(m))
+RETURN u1.name, m.name
+```
diff --git a/flex/interactive/examples/movies/REVIEWED.csv b/flex/interactive/examples/movies/REVIEWED.csv
@@ -1,10 +1,18 @@
-start|end|review|rating
-169|105|yes|95
-169|87|bravo|65
-170|87|good|100
-168|87|bad|62
-169|97|cool|85
-169|95|poor|45
-169|111|bad|68
-170|111|bad|65
-169|37|bravo|92
+start|end|rating|summary
+
+169|105|95|An amazing journey
+171|105|100|A perfect movie
+173|105|90|Unforgettable
+169|87|65|Silly, but fun
+170|87|100|The coolest football movie ever
+168|87|62|Pretty funny at times
+169|97|85|Dark, but compelling
+169|95|45|Slapstick redeemed only by the Robin Williams and Gene Hackman's stellar performances
+169|111|68|A solid romp
+170|111|65|Fun, but a little far fetched
+172|111|60|Just too popcorn
+169|37|92|You had me at Jerry
+174|37|90|Gonna watch again
+170|116|60|The borderline
+172|116|61|A little confused
+174|116|68|Good, but can do better
diff --git a/flex/interactive/examples/movies/User.csv b/flex/interactive/examples/movies/User.csv
@@ -0,0 +1,8 @@
+id|born|name
+168|2003|Angela Scope
+169|2003|Jessica Thompson
+170|2003|James Thompson
+171|2005|Lei Li
+172|2006|Mei Wang
+173|2006|Thomas Christ
+174|2007|Edward Zhang
diff --git a/flex/interactive/examples/movies/graph.yaml b/flex/interactive/examples/movies/graph.yaml
@@ -30,6 +30,19 @@ schema:
             primitive_type: DT_STRING
       primary_keys:
         - id
+    - type_name: User
+      properties:
+        - property_name: id
+          property_type:
+            primitive_type: DT_SIGNED_INT64
+        - property_name: born
+          property_type:
+            primitive_type: DT_SIGNED_INT32
+        - property_name: name
+          property_type:
+            primitive_type: DT_STRING
+      primary_keys:
+        - id
   edge_types:
     - type_name: ACTED_IN
       vertex_type_pair_relations:
@@ -43,7 +56,7 @@ schema:
           relation: MANY_TO_MANY
     - type_name: REVIEW
       vertex_type_pair_relations:
-        - source_vertex: Person
+        - source_vertex: User
           destination_vertex: Movie
           relation: MANY_TO_MANY
       properties:
@@ -52,7 +65,7 @@ schema:
             primitive_type: DT_SIGNED_INT32
     - type_name: FOLLOWS
       vertex_type_pair_relations:
-        - source_vertex: Person
+        - source_vertex: User
           destination_vertex: Person
           relation: MANY_TO_MANY
     - type_name: WROTE

diff --git a/flex/interactive/examples/movies/import.yaml b/flex/interactive/examples/movies/import.yaml
@@ -20,6 +20,9 @@ vertex_mappings:
   - type_name: Person  # must align with the schema
     inputs:
       - Person.csv
+  - type_name: User
+    inputs:
+      - User.csv
   - type_name: Movie
     inputs:
       - Movie.csv
@@ -28,70 +31,30 @@ edge_mappings:
       edge: ACTED_IN
       source_vertex:  Person
       destination_vertex:  Movie
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     inputs:
       - ACTED_IN.csv
   - type_triplet:
       edge: DIRECTED
       source_vertex:  Person
       destination_vertex:  Movie
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     inputs:
       - DIRECTED.csv
   - type_triplet:
       edge: FOLLOWS
-      source_vertex:  Person
+      source_vertex:  User
       destination_vertex:  Person
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     inputs:
       - FOLLOWS.csv
   - type_triplet:
       edge: PRODUCED
       source_vertex:  Person
       destination_vertex:  Movie
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     inputs:
       - PRODUCED.csv
   - type_triplet:
       edge: REVIEW
-      source_vertex:  Person
+      source_vertex:  User
       destination_vertex:  Movie
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     column_mappings:
       - column:
           index: 3
@@ -103,13 +66,5 @@ edge_mappings:
       edge: WROTE
       source_vertex:  Person
       destination_vertex:  Movie
-    source_vertex_mappings:
-      - column:
-          index: 0
-          name: id
-    destination_vertex_mappings:
-      - column:
-          index: 1
-          name: id
     inputs:
       - WROTE.csv
diff --git a/...ler/src/main/java/com/alibaba/graphscope/cypher/integration/suite/movie/MovieQueries.java b/...ler/src/main/java/com/alibaba/graphscope/cypher/integration/suite/movie/MovieQueries.java
@@ -50,14 +50,14 @@ public static QueryContext get_movie_query3_test() {
                 Arrays.asList(
                         "Record<{personName: \"Aaron Sorkin\"}>",
                         "Record<{personName: \"Al Pacino\"}>",
-                        "Record<{personName: \"Angela Scope\"}>",
                         "Record<{personName: \"Annabella Sciorra\"}>",
                         "Record<{personName: \"Anthony Edwards\"}>",
                         "Record<{personName: \"Audrey Tautou\"}>",
                         "Record<{personName: \"Ben Miles\"}>",
                         "Record<{personName: \"Bill Paxton\"}>",
                         "Record<{personName: \"Bill Pullman\"}>",
-                        "Record<{personName: \"Billy Crystal\"}>");
+                        "Record<{personName: \"Billy Crystal\"}>",
+                        "Record<{personName: \"Bonnie Hunt\"}>");
         return new QueryContext(query, expected);
     }
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		MATCH (u: User) -[r:REVIEW]->(movie: Movie) WHERE r.rating > $rateThresh RETURN COUNT(u);