在Postgres中收集递归JSON密钥

Che*_*yne 6 postgresql recursion json

我在JSON数据类型(Postgres 9.3)下存储了Postgres中的JSON文档,我需要以递归方式从树中收集密钥名称.

例如,给定这个JSON树

{
 "files": {
  "folder": {
   "file1": {
    "property": "blah"
   },
   "file2": {
    "property": "blah"
   },
   "file3": {
    "property": "blah"
   },
   "file4": {
    "property": "blah"
   }
 }
},
"software": {
  "apt": {
    "package1": {
        "version": 1.2
    },
    "package2": {
        "version": 1.2
    },
    "package3": {
        "version": 1.2
    },
    "package4": {
        "version": 1.2
    }
  }
 }
}
Run Code Online (Sandbox Code Playgroud)

我想提取类似[file1,file2,file3,file3,package1,package2,package3,package4]的东西

基本上只是我可以用于文本搜索索引的键列表.

我知道我可以使用类似的东西获得最外层对象的键列表

SELECT DISTINCT(json_object_keys(data))
Run Code Online (Sandbox Code Playgroud)

而且我知道可以用类似的东西递归地爬过树

WITH RECURSIVE data()
Run Code Online (Sandbox Code Playgroud)

但我把两者放在一起很麻烦.

有人可以帮忙吗?

Clé*_*ost 9

诀窍是json_typeof在正确的位置添加一些最终条件测试.

jsonb如果您不关心对象键顺序,您也应该使用.

这是我的工作环境:

CREATE TABLE test (
  id  SERIAL PRIMARY KEY,
  doc JSON
);

INSERT INTO test (doc) VALUES ('{
 "files": {
  "folder": {
   "file1": {
    "property": "blah"
   },
   "file2": {
    "property": "blah"
   },
   "file3": {
    "property": "blah"
   },
   "file4": {
    "property": "blah",
    "prop" : {
      "clap": "clap"
    }
   }
 }
},
"software": {
  "apt": {
    "package1": {
        "version": 1.2
    },
    "package2": {
        "version": 1.2
    },
    "package3": {
        "version": 1.2
    },
    "package4": {
        "version": 1.2
    }
  }
 }
}');
Run Code Online (Sandbox Code Playgroud)

当第二个查询未返回任何行时,将停止递归.这是通过传递一个空对象来完成的json_each.

 WITH RECURSIVE doc_key_and_value_recursive(key, value) AS (
  SELECT
    t.key,
    t.value
  FROM test, json_each(test.doc) AS t

  UNION ALL

  SELECT
    t.key,
    t.value
  FROM doc_key_and_value_recursive,
    json_each(CASE 
      WHEN json_typeof(doc_key_and_value_recursive.value) <> 'object' THEN '{}' :: JSON
      ELSE doc_key_and_value_recursive.value
    END) AS t
)
SELECT *
FROM doc_key_and_value_recursive
WHERE json_typeof(doc_key_and_value_recursive.value) <> 'object';
Run Code Online (Sandbox Code Playgroud)


Sim*_*ane 7

我写了一个函数来做到这一点:

CREATE OR REPLACE FUNCTION public.jsonb_keys_recursive(_value jsonb)
 RETURNS TABLE(key text)
 LANGUAGE sql
AS $function$
WITH RECURSIVE _tree (key, value) AS (
  SELECT
    NULL   AS key,
    _value AS value
  UNION ALL
  (WITH typed_values AS (SELECT jsonb_typeof(value) as typeof, value FROM _tree)
   SELECT v.*
     FROM typed_values, LATERAL jsonb_each(value) v
     WHERE typeof = 'object'
   UNION ALL
   SELECT NULL, element
     FROM typed_values, LATERAL jsonb_array_elements(value) element
     WHERE typeof = 'array'
  )
)
SELECT DISTINCT key
  FROM _tree
  WHERE key IS NOT NULL
$function$;
Run Code Online (Sandbox Code Playgroud)

例如,尝试:

SELECT jsonb_keys_recursive('{"A":[[[{"C":"B"}]]],"X":"Y"}');
Run Code Online (Sandbox Code Playgroud)

请注意,其他两个答案在数组内的对象中找不到键,我的解决方案确实如此.(问题根本没有给出任何数组的例子,所以在数组中查找键可能不是原始提问者所需要的,但这正是我所需要的.)