使用pyld与rdflib将JSON转JSON-LD构建RDF图遇问题求助

云上老周

2026-8-1

问题描述

处理多源JSON转JSON-LD+RDF时遇到的问题

我在处理不同数据源的JSON数据时碰到了麻烦，计划用JSON-LD把这些数据转成RDF格式做分析，但不知道怎么正确把普通JSON转成JSON-LD，比如不知道怎么给JSON-LD对象配置正确的上下文。

项目里每个数据源都是基础设施配置信息，能提取成JSON，但每个数据源的结构都不一样。

下面是我用pyld和rdflib尝试转换JSON到RDF图的示例，但输出不符合预期：

注：在Stack Overflow提问时，即使是示例URL也会被判定为垃圾信息，所以运行示例时需要把<unique_iri>替换成真实URL。

示例代码

import json
from pyld import jsonld
from rdflib import Graph

# JSON data
nodes = [
    {
        "sysid": "vm_remote",
        "type": "vm",
        "name": "remote",
        "config": {
                "id": "worker_1",
                "cpu": "2",
        },
        "connect": [
            "db_users"
        ]
    },
    {
        "sysid": "db_users",
        "type": "db",
        "name": "users",
        "config": {
                "id": "database_1",
                "location": "eu_west",
        }
    }
]

# Define the context for the JSON-LD object
context = {
    "@version": 1.1,
    "@base": "<unique_iri>/team_name/",
    "@vocab": "<unique_iri>/resources/onprem/",
    "sysid": "@id",
    "type": "@type",
    "config": {
      "@id": "config",
      "@context": {
        "@base": "<unique_iri>/team_name/config/"
      }
    },
    "connect": {"@id": "relation#connect", "@type": "@id", "@container": "@set"}
}

doc = {
    "@context": context,
    "@graph": nodes,
    "@id": "graph",
    "@type": "graph"
}

print("\nInput JSON-LD:\n" + json.dumps(doc, indent=2))

expended_data = jsonld.expand(doc)
print("\n\Expanded JSON-LD:\n" + json.dumps(expended_data, indent=2))

graph = Graph().parse(data=json.dumps(expended_data), format='json-ld')
print("\nRDF Graph:\n" + graph.serialize(format='json-ld'))

# Find the type of each entry (a resource)
q = """
    PREFIX resources: <<unique_iri>/resources/onprem/>
    SELECT DISTINCT ?type
    WHERE
    {
        ?s resources:type ?type .
    }
    """
print()
for row in graph.query(q):
    print("Type: %s" % row)

输出结果

Input JSON-LD:
{
  "@context": {
    "@version": 1.1,
    "@base": "<unique_iri>/team_name/",
    "@vocab": "<unique_iri>/resources/onprem/",
    "sysid": "@id",
    "type": "@type",
    "config": {
      "@id": "config",
      "@context": {
        "@base": "<unique_iri>/team_name/config/"
      }
    },
    "connect": {
      "@id": "relation#connect",
      "@type": "@id",
      "@container": "@set"
    }
  },
  "@graph": [
    {
      "sysid": "vm_remote",
      "type": "vm",
      "name": "remote",
      "config": {
        "id": "worker_1",
        "cpu": "2"
      },
      "connect": [
        "db_users"
      ]
    },
    {
      "sysid": "db_users",
      "type": "db",
      "name": "users",
      "config": {
        "id": "database_1",
        "location": "eu_west"
      }
    }
  ],
  "@id": "graph",
  "@type": "graph"
}

\Expanded JSON-LD:
[
  {
    "@graph": [
      {
        "<unique_iri>/resources/onprem/config": [
          {
            "<unique_iri>/resources/onprem/cpu": [
              {
                "@value": "2"
              }
            ],
            "<unique_iri>/resources/onprem/id": [
              {
                "@value": "worker_1"
              }
            ]
          }
        ],
        "<unique_iri>/resources/onprem/relation#connect": [
          {
            "@id": "db_users"
          }
        ],
        "<unique_iri>/resources/onprem/name": [
          {
            "@value": "remote"
          }
        ],
        "@id": "vm_remote",
        "@type": [
          "<unique_iri>/resources/onprem/vm"
        ]
      },
      {
        "<unique_iri>/resources/onprem/config": [
          {
            "<unique_iri>/resources/onprem/id": [
              {
                "@value": "database_1"
              }
            ],
            "<unique_iri>/resources/onprem/location": [
              {
                "@value": "eu_west"
              }
            ]
          }
        ],
        "<unique_iri>/resources/onprem/name": [
          {
            "@value": "users"
          }
        ],
        "@id": "db_users",
        "@type": [
          "<unique_iri>/resources/onprem/db"
        ]
      }
    ],
    "@id": "graph",
    "@type": [
      "<unique_iri>/resources/onprem/graph"
    ]
  }
]

RDF Graph:
[
  {
    "@id": "file:///C:...",
    "@type": [
      "<unique_iri>/resources/onprem/graph"
    ]
  }
]

Type: <unique_iri>/resources/onprem/graph

遇到的具体问题

生成的RDF图丢失了节点，找不到原因
不确定怎么处理config节点，这些节点需要有唯一标识符，因为其他数据源可能会引用它们
Python库的转换结果和JSON-LD在线工具的结果不一致

解决方案

1. 修复RDF图丢失节点的问题

节点丢失的核心原因是手动扩展后的JSON-LD嵌套了多层@graph，rdflib解析时无法正确处理这种嵌套结构。

解决方法：移除手动调用jsonld.expand()的步骤，直接让rdflib解析原始JSON-LD文档。rdflib本身支持JSON-LD格式解析，不需要提前手动扩展。修改代码如下：

# 移除手动扩展步骤，直接解析原始doc
graph = Graph().parse(data=json.dumps(doc), format='json-ld')
print("\nRDF Graph:\n" + graph.serialize(format='json-ld', indent=2))

2. 给config节点添加唯一标识符

当前config是匿名节点，要让它拥有唯一ID，需在上下文配置中给config指定@type，同时将config内的id映射为@id，结合@base生成唯一IRI：

# 修改上下文里的config部分
"config": {
  "@id": "config",
  "@type": "@id",  # 标记config是IRI引用
  "@context": {
    "@base": "<unique_iri>/team_name/config/",
    "id": "@id"  # 把config里的id映射为@id
  }
}

修改后，config里的id会被解析为完整IRI，比如worker_1会变成<unique_iri>/team_name/config/worker_1，每个config都有了唯一标识，可被其他数据源引用。

3. 解决Python库与在线工具结果不一致的问题

在线工具严格遵循JSON-LD 1.1规范，而pyld/rdflib的行为受版本和配置影响，要对齐结果：

升级到最新版本的库：pip install --upgrade pyld rdflib
解析时明确指定JSON-LD版本：

graph = Graph().parse(
    data=json.dumps(doc),
    format='json-ld',
    context={"@version": 1.1}
)

避免手动扩展JSON-LD，让库自行处理转换逻辑，减少人为嵌套问题。

修正后的完整代码

import json
from rdflib import Graph

# JSON data
nodes = [
    {
        "sysid": "vm_remote",
        "type": "vm",
        "name": "remote",
        "config": {
                "id": "worker_1",
                "cpu": "2",
        },
        "connect": [
            "db_users"
        ]
    },
    {
        "sysid": "db_users",
        "type": "db",
        "name": "users",
        "config": {
                "id": "database_1",
                "location": "eu_west",
        }
    }
]

# 修正后的上下文
context = {
    "@version": 1.1,
    "@base": "<unique_iri>/team_name/",
    "@vocab": "<unique_iri>/resources/onprem/",
    "sysid": "@id",
    "type": "@type",
    "config": {
      "@id": "config",
      "@type": "@id",
      "@context": {
        "@base": "<unique_iri>/team_name/config/",
        "id": "@id"
      }
    },
    "connect": {"@id": "relation#connect", "@type": "@id", "@container": "@set"}
}

doc = {
    "@context": context,
    "@graph": nodes,
    "@id": "<unique_iri>/team_name/graph",  # 给graph指定完整IRI，避免默认生成file:///...
    "@type": "Graph"
}

print("\nInput JSON-LD:\n" + json.dumps(doc, indent=2))

# 直接用rdflib解析原始JSON-LD
graph = Graph().parse(data=json.dumps(doc), format='json-ld')
print("\nRDF Graph:\n" + graph.serialize(format='json-ld', indent=2))

# 查询所有资源类型
q = """
    PREFIX resources: <{unique_iri}/resources/onprem/>
    SELECT DISTINCT ?s ?type
    WHERE
    {
        ?s a ?type .
    }
    """.format(unique_iri="<unique_iri>")
print()
for row in graph.query(q):
    print(f"资源: {row[0]}, 类型: {row[1]}")

修正后的效果

RDF图会包含所有节点（vm_remote、db_users、worker_1、database_1）
config节点拥有唯一IRI，可被外部数据源引用
查询结果能正确返回所有资源的类型

内容的提问来源于stack exchange，提问作者Derk

使用pyld与rdflib将JSON转JSON-LD构建RDF图遇问题求助

示例代码 #

输出结果 #

遇到的具体问题 #

1. 修复RDF图丢失节点的问题 #

2. 给config节点添加唯一标识符 #

3. 解决Python库与在线工具结果不一致的问题 #

修正后的完整代码 #

修正后的效果 #