Nebula Graph介绍和SpringBoot环境连接和查询
Nebula Graph介绍和SpringBoot环境连接和查询
转载请注明来源 https://www.cnblogs.com/milton/p/16784098.html
说明
当前Nebula Graph的最新版本是3.2.1, 根据官方的文档进行配置
https://docs.nebula-graph.io/3.2.1/14.client/4.nebula-java-client/
Nebula Graph 的一些特点
- 支持分布式. 相对于Neo4j, TigerGraph这些图数据库, Nebula 是面向分布式设计的, 因此对集群的支持比较完备, 在规模上上限要高很多. 在实际项目中存储了180亿的点边, 这个数量对于Neo4j和TigerGraph是比较困难的.
- 支持图空间. 各个图空间的ID是互不干扰的, 但是在同一个图空间里ID的类型和长度必须一致. 注意这个一致约束的是所有的点和边. Nebula 可以使用int64作为ID, 也可以用字符串, 但是字符串需要指定一个长度, 例如64个字节. 相对于只能用长整数的Neo4j, ID设计上更自由灵活.
- 点对应的类型叫TAG, 边对应的类型叫EDGE
- TAG和EDGE都会对应一组的属性(map, 或者说dict)
- 一个点可以对多个TAG, 每个TAG一组属性, 多组属性. 项目中建议一开始不要用多TAG, 在整个图结构稳定后, 再做合并
- 一个边只对应一个EDGE, 一组属性
- Nebula 用的是自定义的查询语法 GQL, 和 cypher 语法基本一样
- 除了点边的ID和关联关系外, 只有带索引的属性可以查询. 这点和其它图数据库不一样, 其它数据库即使没有索引, 慢是慢点但是不报错, Nebula直接给你返回错误.
- 对于返回数量较大的查询, Nebula会强制查询必须带limit
- Nebula 单节点稳定性是有问题的, 在3.2.1版本中观察到偶尔会出现服务自行退出, 如果在生产环境使用, 需要有后台监控进行心跳检测和自动启动
GQL 常用查询
下面列出一些常用的查询
-- 列出图空间
SHOW SPACES;
-- 列出tag(点类型)和edge(边类型), 需要先 USE 一个图空间
SHOW TAGS;
SHOW EDGES;
列出某一类型的点和边
MATCH ()-[e:follow]-() RETURN e
MATCH (v:player) RETURN v
带条件的查询, 在结果数量较多时必须带limit, 否则Nebula会报错
match (v:ADDRESS)-[e]-() where id(v)==\"ADD:82388116\" return v,e limit 100
基础配置和使用
在上面的链接中, 提供了最小的配置和测试代码
pom.xml 增加包依赖
对于Nebula Graph 3.2.1, 需要使用3.0.0的版本. client的每个版本只能对应特定的一两个服务端版本
<dependency>
<groupId>com.vesoft</groupId>
<artifactId>client</artifactId>
<version>3.0.0</version>
</dependency>
Java调用
Java调用主要是三部分, 创建连接池, 创建会话, 执行查询
创建 NebulaPool 连接池
连接到地址127.0.0.1, 端口9669, 连接池大小100. 注意地址和端口是一个列表, Nebula是支持集群的. 连接时不需要用户和密码
NebulaPool pool = new NebulaPool();
try {
NebulaPoolConfig nebulaPoolConfig = new NebulaPoolConfig();
nebulaPoolConfig.setMaxConnSize(100);
List<HostAddress> addresses = Arrays.asList(new HostAddress("127.0.0.1", 9669));
Boolean initResult = pool.init(addresses, nebulaPoolConfig);
if (!initResult) {
log.error("pool init failed.");
return;
}
} catch ()
//...
创建 Session 会话
创建会话时需要用户名和密码
Session session = pool.getSession("root", "nebula", false);
执行查询
创建一个SPACE, 然后使用这个SPACE, 创建一个TAG person, 创建一个EDGE like
String createSchema = "CREATE SPACE IF NOT EXISTS test(vid_type=fixed_string(20)); "
+ "USE test;"
+ "CREATE TAG IF NOT EXISTS person(name string, age int);"
+ "CREATE EDGE IF NOT EXISTS like(likeness double)";
ResultSet resp = session.execute(createSchema);
if (!resp.isSucceeded()) {
log.error(String.format("Execute: `%s', failed: %s",
createSchema, resp.getErrorMessage()));
System.exit(1);
}
添加一个点记录
String insertVertexes = "INSERT VERTEX person(name, age) VALUES "
+ "'Bob':('Bob', 10), "
+ "'Lily':('Lily', 9), "
+ "'Tom':('Tom', 10), "
+ "'Jerry':('Jerry', 13), "
+ "'John':('John', 11);";
ResultSet resp = session.execute(insertVertexes);
if (!resp.isSucceeded()) {
log.error(String.format("Execute: `%s', failed: %s",
insertVertexes, resp.getErrorMessage()));
System.exit(1);
}
查询
String query = "GO FROM \"Bob\" OVER like "
+ "YIELD $^.person.name, $^.person.age, like.likeness";
ResultSet resp = session.execute(query);
if (!resp.isSucceeded()) {
log.error(String.format("Execute: `%s', failed: %s",
query, resp.getErrorMessage()));
System.exit(1);
}
printResult(resp);
在 SpringBoot 项目中使用 Nebula Graph
pom.xml 增加包依赖
<dependency>
<groupId>com.vesoft</groupId>
<artifactId>client</artifactId>
<version>3.0.0</version>
</dependency>
Session工厂: NebulaSessionFactory.java
配合@Bean(destroyMethod = "close"), 创建一个工厂类, 接收pool并实现close()方法
public class NebulaSessionFactory {
private final NebulaPool pool;
private final String username;
private final String password;
public NebulaSessionFactory(NebulaPool pool, String username, String password) {
this.pool = pool;
this.username = username;
this.password = password;
}
public Session getSession() {
try {
return pool.getSession(username, password, false);
} catch (NotValidConnectionException|IOErrorException|AuthFailedException|ClientServerIncompatibleException e) {
throw new RuntimeException("Nebula session exception", e);
}
}
public void close() {
pool.close();
}
}
为什么不直接将 NebulaPool 配置为Bean? 因为 Session 每次创建时需要带用户名密码, 将密码作为config注入到每个Service中肯定是大家都不愿意看到的.
配置修改: application.yml
- 这里的值如果不打算使用profile配置, 可以直接写入
- hosts是逗号分隔的地址端口列表, 例如
10.22.33.33:9669,10.22.33.34:9669
myapp:
nebula:
hosts: @nebula.hosts@
username: @nebula.username@
password: @nebula.password@
max-conn: @nebula.max-conn@
Spring启动配置: NebulaGraphConfig.java
应用启动时读取配置, 创建 NebulaPool, 并实例化 NebulaSessionFactory, destroyMethod = "close", 这个表示在项目shutdown时会调用Bean的close方法释放资源.
@Configuration
public class NebulaGraphConfig {
@Value("${myapp.nebula.hosts}")
private String hosts;
@Value("${myapp.nebula.max-conn}")
private int maxConn;
@Value("${myapp.nebula.username}")
private String username;
@Value("${myapp.nebula.password}")
private String password;
@Bean(destroyMethod = "close")
public NebulaSessionFactory nebulaSessionFactory() {
List<HostAddress> hostAddresses = new ArrayList<>();
String[] hostList = hosts.split(",[ ]*");
for (String host : hostList) {
String[] hostParts = host.split(":");
if (hostParts.length != 2 || !hostParts[1].matches("\\d+")) {
throw new RuntimeException("Invalid host name set for Nebula: " + host);
}
hostAddresses.add(new HostAddress(hostParts[0], Integer.parseInt(hostParts[1])));
}
NebulaPoolConfig poolConfig = new NebulaPoolConfig();
poolConfig.setMaxConnSize(maxConn);
NebulaPool pool = new NebulaPool();
try {
pool.init(hostAddresses, poolConfig);
} catch (UnknownHostException e) {
throw new RuntimeException("Unknown Nebula hosts");
}
return new NebulaSessionFactory(pool, username, password);
}
}
Service调用
在 Service 中进行调用
@Service
@Slf4j
public class GraphServiceImpl implements GraphService {
@Autowired
private NebulaSessionFactory sessionFactory;
@Override
public <T> NebulaResult<T> query(String graphSpace, String gql) {
Session session = null;
try {
log.info("GQL: {}", gql);
session = sessionFactory.getSession();
NebulaResult<Void> res = query(session, "USE " + graphSpace);
if (!res.isSuccess() || res.getResults() == null || res.getResults().size() == 0) {
log.error("Failed to use space:{}", graphSpace);
return null;
}
if (!graphSpace.equals(res.getResults().get(0).getSpaceName())) {
log.error("Failed to use space:{}, result:{}", graphSpace, res.getResults().get(0).getSpaceName());
return null;
}
return query(session, gql);
} catch (IOErrorException e) {
log.error(e.getMessage(), e);
return null;
} finally {
if (session != null) {
session.release();
}
}
}
private <T> NebulaResult<T> query(Session session, String gql) throws IOErrorException {
String json = session.executeJson(gql);
return JacksonUtil.extractByType(json, new TypeReference<>() {});
}
}
辅助类 NebulaResult.java 等
外层结构
这里定义了 json 格式响应的外层结构
@Data
public class NebulaResult<T> implements Serializable {
private List<Error> errors;
private List<Result<T>> results;
@JsonIgnore
public boolean isSuccess() {
return (errors != null && errors.size() == 1 && errors.get(0).getCode() == 0);
}
@Data
public static class Error implements Serializable {
private int code;
}
@Data
@JsonIgnoreProperties(ignoreUnknown = true)
@JsonInclude(JsonInclude.Include.NON_NULL)
public static class Result<T> implements Serializable {
private String spaceName;
private List<Element<T>> data;
private List<String> columns;
private Error errors;
private long latencyInUs;
}
@Data
public static class Element<T> implements Serializable {
private List<Meta<T>> meta;
private List<Serializable> row;
}
@Data
public static class Meta<T> implements Serializable {
private String type;
private T id;
}
}
内层因为区分Edge和Vertex, 结构不一样. 如果是混合返回的结果, 可以用 Serializable
String gql = "match (v:ADDR)-[e]-() where id(v)==\"ADD:123123\" return v,e limit 100";
NebulaResult<Serializable> res = graphService.query("insurance", gql);
log.info(JacksonUtil.compress(res));
Assertions.assertThat(res).isNotNull();
对于边, 需要使用结构化的ID
@Data
@JsonIgnoreProperties(ignoreUnknown = true)
@JsonInclude(JsonInclude.Include.NON_NULL)
public class EdgeId implements Serializable {
private int ranking;
private int type;
private String dst;
private String src;
private String name;
}
用这个结构进行查询
NebulaResult<EdgeId> res3 = graphService.query("t_test1", "MATCH ()-[e:follow]-() RETURN e");
对于点, ID就是String
NebulaResult<String> res2 = graphService.query("t_test1", "MATCH (v:player) RETURN v");
工具类 JacksonUtil.java
public class JacksonUtil {
private static final Logger log = LoggerFactory.getLogger(JacksonUtil.class);
private static final ObjectMapper MAPPER = createObjectMapper();
private JacksonUtil() {}
private static ObjectMapper createObjectMapper() {
ObjectMapper objectMapper = new ObjectMapper();
objectMapper.configure(JsonGenerator.Feature.AUTO_CLOSE_TARGET, false);
objectMapper.disable(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES);
SimpleDateFormat df = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
objectMapper.setDateFormat(df);
return objectMapper;
}
public static ObjectMapper getObjectMapper() {
return MAPPER;
}
public static String compressByView(Object o, Class<?> clazz) {
if (o == null) {
return null;
}
try {
return MAPPER.writerWithView(clazz).writeValueAsString(o);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static JsonNode readTree(String json) {
if (json == null || json.length() == 0) {return null;}
try {
return MAPPER.readTree(json);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static JsonNode readTree(Object obj) {
if (obj == null) {return null;}
return MAPPER.valueToTree(obj);
}
public static <T> T extractByType(JsonNode jsonNode, TypeReference<T> tp) {
if (jsonNode == null) {return null;}
try {
return MAPPER.readerFor(tp).readValue(jsonNode);
} catch (IOException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByType(String json, TypeReference<T> tp) {
if (json == null || json.length() == 0) {return null;}
try {
return MAPPER.readValue(json, tp);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByType(JsonNode jsonNode, Type type) {
if (jsonNode == null) {return null;}
try {
JavaType javaType = MAPPER.getTypeFactory().constructType(type);
return MAPPER.readerFor(javaType).readValue(jsonNode);
} catch (IOException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByType(String json, Type type) {
JavaType javaType = MAPPER.getTypeFactory().constructType(type);
return extractByType(json, javaType);
}
public static <T> T extractByType(String json, JavaType type) {
if (json == null || json.length() == 0) {return null;}
try {
return MAPPER.readValue(json, type);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByClass(JsonNode jsonNode, Class<T> clazz) {
if (jsonNode == null) {return null;}
try {
return MAPPER.treeToValue(jsonNode, clazz);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByClass(String json, Class<T> clazz) {
if (json == null || json.length() == 0) {return null;}
try {
return MAPPER.readValue(json, clazz);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
public static <T> T extractByClass(Map<String, Object> map, Class<T> clazz) {
if (map == null) {return null;}
return MAPPER.convertValue(map, clazz);
}
public static <T> String compress(T object) {
if (object == null) {return null;}
try {
return MAPPER.writer().writeValueAsString(object);
} catch (JsonProcessingException e) {
log.error(e.getMessage(), e);
}
return null;
}
}