使用PostgreSQL(v9.5),JSONB格式提供了绝佳的机会。但现在我陷入了一个看似简单的操作;
比较两个jsonb对象;看一个文档中哪个与另一个文档相比有什么不同或缺失。
到目前为止我有什么
WITH reports(id,DATA) AS (
VALUES (1,'{"a":"aaa", "b":"bbb", "c":"ccc"}'::jsonb),
(2,'{"a":"aaa", "b":"jjj", "d":"ddd"}'::jsonb) )
SELECT jsonb_object_agg(anon_1.key, anon_1.value)
FROM
(SELECT anon_2.key AS KEY,
reports.data -> anon_2.KEY AS value
FROM reports,
(SELECT DISTINCT jsonb_object_keys(reports.data) AS KEY
FROM reports) AS anon_2
ORDER BY reports.id DESC) AS anon_1
应该返回第1行与第2行的差异:
'{"b":"bbb", "c":"ccc", "d":null}'
相反,它也返回重复({"a": "aaa"}
)。也;一般来说可能会有更优雅的方法!
CREATE OR REPLACE FUNCTION jsonb_diff_val(val1 JSONB,val2 JSONB)
RETURNS JSONB AS $$
DECLARE
result JSONB;
v RECORD;
BEGIN
result = val1;
FOR v IN SELECT * FROM jsonb_each(val2) LOOP
IF result @> jsonb_build_object(v.key,v.value)
THEN result = result - v.key;
ELSIF result ? v.key THEN CONTINUE;
ELSE
result = result || jsonb_build_object(v.key,'null');
END IF;
END LOOP;
RETURN result;
END;
$$ LANGUAGE plpgsql;
查询:
SELECT jsonb_diff_val(
'{"a":"aaa", "b":"bbb", "c":"ccc"}'::jsonb,
'{"a":"aaa", "b":"jjj", "d":"ddd"}'::jsonb
);
jsonb_diff_val
---------------------------------------
{"b": "bbb", "c": "ccc", "d": "null"}
(1 row)
我创建了类似的函数,可以递归地扫描对象,并返回新对象和旧对象之间的差异。我无法找到'更好'的方法来确定jsonb对象'是否为空' - 所以如果有任何建议如何简化它将会很感激。我计划使用它来跟踪对jsonb对象所做的更新,因此我只存储已更改的内容。
这是功能:
CREATE OR REPLACE FUNCTION jsonb_diff_val(val1 JSONB,val2 JSONB)
RETURNS JSONB AS $$
DECLARE
result JSONB;
object_result JSONB;
i int;
v RECORD;
BEGIN
IF jsonb_typeof(val1) = 'null'
THEN
RETURN val2;
END IF;
result = val1;
FOR v IN SELECT * FROM jsonb_each(val1) LOOP
result = result || jsonb_build_object(v.key, null);
END LOOP;
FOR v IN SELECT * FROM jsonb_each(val2) LOOP
IF jsonb_typeof(val1->v.key) = 'object' AND jsonb_typeof(val2->v.key) = 'object'
THEN
object_result = jsonb_diff_val(val1->v.key, val2->v.key);
-- check if result is not empty
i := (SELECT count(*) FROM jsonb_each(object_result));
IF i = 0
THEN
result = result - v.key; --if empty remove
ELSE
result = result || jsonb_build_object(v.key,object_result);
END IF;
ELSIF val1->v.key = val2->v.key THEN
result = result - v.key;
ELSE
result = result || jsonb_build_object(v.key,v.value);
END IF;
END LOOP;
RETURN result;
END;
$$ LANGUAGE plpgsql;
那么简单的查询看起来像这样:
SELECT jsonb_diff_val(
'{"a":"aaa", "b":{"b1":"b","b2":"bb","b3":{"b3a":"aaa","b3c":"ccc"}}, "c":"ccc"}'::jsonb,
'{"a":"aaa", "b":{"b1":"b1","b3":{"b3a":"aaa","b3c":"cccc"}}, "d":"ddd"}'::jsonb
);
jsonb_diff_val
-------------------------------------------------------------------------------
{"b": {"b1": "b1", "b2": null, "b3": {"b3c": "cccc"}}, "c": null, "d": "ddd"}
(1 row)
我的解决方案不是递归的,但您可以使用它来检测公共键/值:
-- Diff two jsonb objects
CREATE TYPE jsonb_object_diff_result AS (
old jsonb,
new jsonb,
same jsonb
);
CREATE OR REPLACE FUNCTION jsonb_object_diff(in_old jsonb, in_new jsonb)
RETURNS jsonb_object_diff_result AS
$jsonb_object_diff$
DECLARE
_key text;
_value jsonb;
_old jsonb;
_new jsonb;
_same jsonb;
BEGIN
_old := in_old;
_new := in_new;
FOR _key, _value IN SELECT * FROM jsonb_each(_old) LOOP
IF (_new -> _key) = _value THEN
_old := _old - _key;
_new := _new - _key;
IF _same IS NULL THEN
_same := jsonb_build_object(_key, _value);
ELSE
_same := _same || jsonb_build_object(_key, _value);
END IF;
END IF;
END LOOP;
RETURN (_old, _new, _same);
END;
$jsonb_object_diff$
LANGUAGE plpgsql;
结果可能如下所示:
SELECT * FROM jsonb_object_diff(
'{"a": 1, "b": 5, "extra1": "woo", "old_null": null, "just_null": null}'::jsonb,
'{"a": 1, "b": 4, "extra2": "ahoj", "new_null": null, "just_null": null}'::jsonb);
-[ RECORD 1 ]--------------------------------------
old | {"b": 5, "extra1": "woo", "old_null": null}
new | {"b": 4, "extra2": "ahoj", "new_null": null}
same | {"a": 1, "just_null": null}
这是一个没有创建新功能的解决方案;
SELECT
json_object_agg(COALESCE(old.key, new.key), old.value)
FROM json_each_text('{"a":"aaa", "b":"bbb", "c":"ccc"}') old
FULL OUTER JOIN json_each_text('{"a":"aaa", "b":"jjj", "d":"ddd"}') new ON new.key = old.key
WHERE
new.value IS DISTINCT FROM old.value
结果是;
{"b" : "bbb", "c" : "ccc", "d" : null}
此方法仅比较json的第一级。它不会遍历整个对象树。