feat(c++): API design for multi-label support

acezen commented 2 months ago

Describe the enhancement requested

TBF

Component(s)

C++

Elssky commented 2 months ago

Neo4j Cypher statements involving tags can be mainly divided into the following categories:

1. Creation

Create nodes with labels：
```
CREATE (n:Label {property: value})
```
Add a label to existing nodes：
```
MATCH (n) SET n:NewLabel
```

2. Deletion and Renaming

Remove all labels from nodes：
```
MATCH (n) REMOVE LABELS(n)
```
Rename a label (by adding a new label and removing the old one)：
```
MATCH (n:OldLabel) SET n:NewLabel REMOVE n:OldLabel
```

3. Querying and Filtering

Query nodes with a specific label：
```
MATCH (n:Label) RETURN n
```
Query nodes with multiple labels：
```
MATCH (n:Label1:Label2) RETURN n
```
Query nodes with a specific label and property：
```
MATCH (n:Label {property: value}) RETURN n
```

Check if a node has a specific label：

MATCH (n)
WHERE id(n) = <your_node_id> AND n:Label1
RETURN COUNT(n) > 0 AS hasLabel1

4. Statistics and Information

Count the number of nodes with a specific label：
```
MATCH (n:Label) RETURN COUNT(n)
```

Query all labels：

CALL db.labels() YIELD label RETURN label

acezen commented 2 months ago

LDBC SNB Cypher statements:

Querying and Filtering

bi-1：

/*
:params { datetime: datetime('2011-12-01T00:00:00.000') }
*/
MATCH (message:Message)
WHERE message.creationDate < $datetime
WITH count(message) AS totalMessageCountInt // this should be a subquery once Cypher supports it
WITH toFloat(totalMessageCountInt) AS totalMessageCount

bi-2：

MATCH (tag:Tag)-[:HAS_TYPE]->(:TagClass {name: $tagClass})
// window 1
OPTIONAL MATCH (message1:Message)-[:HAS_TAG]->(tag)
WHERE $date <= message1.creationDate
AND message1.creationDate < $date + duration({days: 100})
WITH tag, count(message1) AS countWindow1
// window 2
OPTIONAL MATCH (message2:Message)-[:HAS_TAG]->(tag)
WHERE $date + duration({days: 100}) <= message2.creationDate
AND message2.creationDate < $date + duration({days: 200})
WITH
tag,
countWindow1,
count(message2) AS countWindow2

bi-3：

/*
:params { tagClass: 'MusicalArtist', country: 'Burma' }
*/
MATCH
(:Country {name: $country})<-[:IS_PART_OF]-(:City)<-[:IS_LOCATED_IN]-
(person:Person)<-[:HAS_MODERATOR]-(forum:Forum)-[:CONTAINER_OF]->
(post:Post)<-[:REPLY_OF*0..]-(message:Message)-[:HAS_TAG]->(:Tag)-[:HAS_TYPE]->(:TagClass {name: $tagClass})

acezen commented 2 months ago

LDBC SNB Gremlin statements:

Querying and Filtering

bi-2：

g.V().has("TAGCLASS", "name", "$tagClass").in("HASTYPE").as("tag").in("HASTAG").hasLabel("POST", "COMMENT").union(    __.has("creationDate", P.gte($date).and(P.lt($ddate))).as("msg1"),    __.has("creationDate", P.gte($ddate).and(P.lt($dddate))).as("msg2")).group().by(select("tag")).by(    dedup("msg1").count().as("cnt1"),    dedup("msg2").count().as("cnt2")).select(expr("(@cnt2 - @cnt1)*(@cnt2 - @cnt1)")).as("diff").order().by(select("diff"), desc).by(select(keys).values("name"), asc).limit(100).select("cnt1", "cnt2", "diff").by().by().by()

bi-3

g.V().has("PLACE", "name", "$country").in("ISPARTOF").in("ISLOCATEDIN").as("person").in("HASMODERATOR").as("forum") .out("CONTAINEROF").in("0..5", "REPLYOF").endV().as("msg").out("HASTAG").out("HASTYPE").has("TAGCLASS", "name", "$tagClass").select("msg").group().by(select("forum").as("forum")).by(dedup().count().as("msgCnt")).order().by(select("msgCnt"), desc).by(select("forum").by("id"), asc).limit(20).select("forum", "msgCnt").by(valueMap("id", "title", "creationDate")).by()

bi-4

g.V().has("TAG", "name", "$tag").in("HASTAG").as("msg") .out("HASCREATOR").as("person") .union( select("msg").as("msg1"), select("msg").in("LIKES").as("like"), select("msg").in("REPLYOF").as("reply")) .group() .by(select("person")) .by( select("msg1").count().as("msgCnt"), select("like").count().as("likeCnt"), select("reply").count().as("replyCnt")) .select(keys).as("person") .select(expr("1*@msgCnt+2*@replyCnt+10*@likeCnt")).as("score") .order() .by(select("score"), desc) .by(select("person").by("id"), asc) .limit(100) .select("person", "replyCnt", "likeCnt", "msgCnt", "score") .by("id").by().by().by().by()

Elssky commented 2 months ago


class VerticesCollection {
public:
    /**
     * @brief Query vertices with a specific label
     *
     * @param filter_label The label to query vertices by
     * @param graph_info A smart pointer to GraphInfo that contains details about the graph
     * @param type The type of vertices to query 
     * @return A VerticesCollection containing all vertices that have the specified label
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithLabel(
        const std::string& filter_label, 
        const std::shared_ptr<GraphInfo>& graph_info, 
        const std::string& type);

    /**
     * @brief Query vertices with a specific label within a given collection
     *
     * @param filter_label The label to query vertices by
     * @param vertices_collection The collection of vertices to search within
     * @return A VerticesCollection containing all vertices from the specified collection that have the specified label
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithLabel(
        const std::string& filter_label, 
        const std::shared_ptr<VerticesCollection>& vertices_collection);

    /**
     * @brief Query vertices with multiple labels
     *
     * @param filter_labels A vector of labels to query vertices by
     * @param graph_info A smart pointer to GraphInfo that contains details about the graph
     * @param type The type of vertices to query 
     * @return A VerticesCollection containing all vertices that have all of the specified labels
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithMultipleLabels(
        const std::vector<std::string>& filter_labels, 
        const std::shared_ptr<GraphInfo>& graph_info, 
        const std::string& type);

    /**
     * @brief Query vertices with multiple labels within a given collection
     *
     * @param filter_labels A vector of labels to query vertices by
     * @param vertices_collection The collection of vertices to search within
     * @return A VerticesCollection containing all vertices from the specified collection that have all of the specified labels
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithMultipleLabels(
        const std::vector<std::string>& filter_labels, 
        const std::shared_ptr<VerticesCollection>& vertices_collection);

    /**
     * @brief Query vertices with a specific label and properties
     *
     * @param filter_label The label to query vertices by
     * @param propery_filter_expr A shared pointers to Property filter expression that the vertices should match
     * @param graph_info A smart pointer to GraphInfo that contains details about the graph
     * @param type The type of vertices to query 
     * @return A VerticesCollection containing all vertices that have the specified label and match the given properties
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithLabelAndProperty(
        const std::string& filter_label, 
        const std::shared_ptr<Expression> propery_filter_expr, 
        const std::shared_ptr<GraphInfo>& graph_info, 
        const std::string& type);

    /**
     * @brief Query vertices with a specific label and properties within a given collection
     *
     * @param filter_label The label to query vertices by
     * @param propery_filter_expr A shared pointers to Property filter expression that the vertices should match
     * @param vertices_collection The collection of vertices to search within
     * @return A VerticesCollection containing all vertices from the specified collection that have the specified label and match the given properties
     */
    static Result<std::shared_ptr<VerticesCollection>> verticesWithLabelAndProperty(
        const std::string& filter_label, 
        const std::shared_ptr<Expression> propery_filter_expr, 
        const std::shared_ptr<VerticesCollection>& vertices_collection);

};

class VertexIter {
    /**
     * @brief Check if the vertex has a specific label
     *
     * @param label The label to check for on the specified vertex
     * @return True if the vertex has the specified label, false otherwise
     */
    bool HasLabel(const std::string& label);
}

yecol commented 2 months ago

Please refers to the existing VerticesCollection, holding iterators would be better.

yecol commented 2 months ago

They should be static functions like Make I guess, rather than member functions of VerticesCollection

ref: https://github.com/apache/incubator-graphar/blob/a5c94a483f0373a619dcec18df06cbd48c10c103/cpp/src/graphar/high-level/graph_reader.h#L305

Elssky commented 2 months ago

MATCH (n:Label {property: value}) RETURN n
1. Query nodes with a specific label

Cypher:

MATCH (n:Label) RETURN n

Gremlin:

g.V().hasLabel('Label')

2. Query nodes with multiple labels

Cypher:

MATCH (n:Label1:Label2) RETURN n

Gremlin:

g.V().hasLabel('Label1').hasLabel('Label2')

3. Query nodes with a specific label and property

Cypher:

MATCH (n:Label {property: value}) RETURN n

Gremlin:

g.V().hasLabel('Label').has('property', value)

apache / incubator-graphar