Cypher

介绍图数据库中属性图的查询语言——Cypher，这是一篇CCF A类文章，发表于数据库顶会SIGMOD，讲得很详细很不错！原文链接

---

@Authors：Nadime Francis，Alastair Green，Paolo Guagliardo
@Published in: SIGMOD Conference 2018: 1433-1445
@Presented by: Yina Lv , Time: Qct 10, 2018
@Action：October 11, 2018 9:21 AM

---

Schema

• 介绍了属性图和Cypher语言的语法结构
• 以事例的形式讲述Cypher查询过程
• Cypher中的核心元素：数据模型和查询语言
• pattern matching
• 比较Cypher语言和其他数据库查询语言的特征，描述其扩展性、先进性，并将发展成Cypher10

1.Introduce

1.1 Property graph database

• Property graph databasesuch as Neo4j，JanusGraph and Sparksee have become more widespread in industry and academia.
• used in multiple domains, such as master data and knowledge management, recommendation engines,fraud detection, IT operations and network management, authorization and access control , bioinformatics , social networks, software system analysis , and in investigative journalism.
• Benefits:
  • explicit support for modeling graph data
  • native indexing
  • storage for fast graph traversal operations
  • built-in support for graph algorithms
  • the provision of graph language

1.2 Cypher

• Cypher: a well-established language for querying and updating property graph databases

  • began life in the Neo4j product, but has now been implemented commercially in other products such as SAP HANA Graph, Redis Graph, Agens Graph (over PostgreSQL) and Memgraph.
  • Cypher is also used in several research projects (e.g., Ingraph ,Gradoop , and Cytosm) as well as in recent or incubating open-source projects, such as Cypher for Apache Spark and Cypher over Gremlin.
  • Cypher是线性查询，输入property graph，输出table
  • 数据更新：
    • CREATE: creating new nodes and relationships
    • DELETE: moving entities
    • SET: updating properties
    • MERGE: try to match the given patten,and creates the pattern if no match was found

2.Example

Graph Schema：

• PS:图中有三个label，分别是researcher、Student、Publication，三种关系:SUPERVISES、AUTHORS、CITES
• 研究员有指导的学生，研究员发表刊物，那么就是某刊物的作者，刊物之间可以互相引用。
• CITES我画成了自循环，但是大家这样理解比较合适：一个刊物可以被另外一个刊物引用。

Big graph:(整个数据库的属性图如下)

2.1 查询示例语句如下：

Q: 返回图G中每个researcher的name，他们所supervises指向的学生数，以及他们撰写的出版物被其他出版物直接或间接引用的次数。

我们分析每一个语句的查询结果。

1. MATCH (r:researcher)

• 在这句语句中，r是instance，researcher是label

2. OPTIONAL MATCH (r)-[:SUPERVISES]->(s:Student)

• OPTIONAL MATCH与MATCH的区别在于，OPTIONAL MATCH会返回查询结果为NULL的记录。
• SUPERVISES为关系的label，(r)-[:SUPERVISES]->(s:Student)表示找出researcher对应指导的学生实例
• 1、2两句语句执行结果如下：
  

3. WITH r,count(s) AS studentsSupervised

• AS：取别名
• count(s):计算上面找出的学生数
• 在with语句前面部分找出的子图作为with语句后面查询的驱动表
  

4. MATCH (r)-[:AUTHORS]->(p1:Publication)

• n10不写是因为MATCH不返回NULL的记录
  

5. OPTIONAL MATCH (p1)-[:CITES*]->(p2:Publication)

• CITES后面的*表示任意步长，当然也可以指定步长，比如CITES*1..3表示跳一步到三步
• 结果中的n9有两个是因为n9可以有两条路径达到n2
  

6、7. RETURN r.name,studentsSupervised,count(DISTINCT p2) AS citedCount

• 返回最后的结果
  

2.2 Examples from industry

Q1: A real-world query from the network management domain

• The query returns the component that is depended upon – both directly and indirectly – by the largest number of entities.
• order by: 排序，可指定增序还是降序
• LIMIT：取几行记录
• MATCH中的箭头可以向左也可以向右

Q2: A query in the domain of fraud detection

• pInfo后面三个为label，可以将pInfo理解为节点信息，那么语句的意思就是信息可以是SSN，或者PhoneNumber，或者Address
• WHERE后面跟着的是一些过滤条件，这些过滤条件也可以放在MATCH后面，比如第一个where后面跟着实例的标签，那么也可以写成MATCH (accHolder:AccountHolder)-[:HAS]->(pInfo:SSN:PhoneNumber:Address)
• collect(): 返回表达式表示的值，放入list并返回
• labels(): 返回一个节点的所有标签，以list的形式返回
• count(*): 用于计算匹配的行数

3.FORMAL SPECIFICATION

• The key elements of Cypher are as follows:

  • data model, that includes values, graphs, and tables;
  • query language, that includes expressions, patterns, clauses, and queries.

• Property graphs
  

  • We now refer to the property graph in Figure 1 and show how, for a sample of its nodes and relationships, it is formally represented in this model as a graph G = (N, R, src, tgt, ι, λ, τ ).
  • N: 实例
  • R: 关系
  • src：从关系到源节点这样的存储结构
  • tgt：从关系到目标节点这样的存储结构
  • ι(n1,name)=Nils: 节点1的名字叫Nils，存储(key,value)的形式
  • λ(n1)={Student}: 存储节点label
  • τ®=AUTHORS: 存储关系的label

---

• ← Previous Post
• Next Post →