在Postgres的json列中将LocalDate转换为Timestamp

Question

我必须在PostgresDB 9.5中迁移一个列。

表名是可测试的，相关列是examplecolumn。

在examplecolumn中，我有这个JSON数组，例如：

[{"key":"id","before":null,"after":4371},
{"key":"status","before":null,"after":1},
{"key":"startDate","before":null,"after":1514761200000},
{"key":"endDate","before":null,"after":1546297199000},
{"key":"billingDate","before":null,"after":{"year":2018,"month":"JANUARY","dayOfMonth":15,"dayOfWeek":"MONDAY","era":"CE","dayOfYear":15,"leapYear":false,"monthValue":1,"chronology":{"calendarType":"iso8601","id":"ISO"}}}]

我想将具有“billingDate”的孩子转换为以下格式的时间戳：

[{"key":"id","before":null,"after":4371},
{"key":"status","before":null,"after":1},
{"key":"startDate","before":null,"after":1514761200000},
{"key":"endDate","before":null,"after":1546297199000},
{"billingDate":1515974400}]

我想保留其他一切。首先，我试着像这样获得这些行。

SELECT *, examplecolumn::json->4->'after' as export
FROM testtable
WHERE examplecolumn like '%after":{%';

但我不能用修复索引（例如4）做这个，我不知道哪个行有这个LocalDate的子对象，这就是我写这个where子句的原因。

我怎样才能做到这一点？我尝试了几次Postgres JSON函数，但我很厌倦了。

谢谢。

Answer 1

我不清楚问题是关于根据问题的标题将数据转换为时间戳，还是根据问题提取数据。所以这个答案是关于提取数据。

首先，CTE只是随机化JSON数组中项目的顺序，以显示它绕过索引问题。显然你会在你的桌子上执行查询。然后带有“billingDate”的项目仅包含在偶数结果中;奇数包含其余的JSON项并排除其中一项。

WITH cte AS (
        SELECT ('[' || STRING_AGG(CASE WHEN id = 2 THEN CASE WHEN g % 2 = 0 THEN j::TEXT END ELSE j::TEXT END, ',' ORDER BY RANDOM()) || ']')::JSON AS j
        FROM (
                VALUES
                (1, '{"key":"id","before":null,"after":4371}'),
                (1, '{"key":"status","before":null,"after":1}'),
                (1, '{"key":"startDate","before":null,"after":1514761200000}'),
                (1, '{"key":"endDate","before":null,"after":1546297199000}'),
                (2, '{"key":"billingDate","before":null,"after":{"year":2018,"month":"JANUARY","dayOfMonth":15,"dayOfWeek":"MONDAY","era":"CE","dayOfYear":15,"leapYear":false,"monthValue":1,"chronology":{"calendarType":"iso8601","id":"ISO"}}}')
        ) AS v(id, j), generate_series(1, 10) AS g
        GROUP BY g
)
SELECT j, (SELECT je->'after' FROM JSON_ARRAY_ELEMENTS(j) AS je WHERE je->>'key' = 'billingDate' AND (je->>'after') IS NOT NULL) AS export
FROM cte

结果：

如果您只想获取具有“billingDate”的项目，那么您可以从您的问题中添加WHERE子句：

WITH cte AS (
        SELECT ('[' || STRING_AGG(CASE WHEN id = 2 THEN CASE WHEN g % 2 = 0 THEN j::TEXT END ELSE j::TEXT END, ',' ORDER BY RANDOM()) || ']')::JSON AS j
        FROM (
                VALUES
                (1, '{"key":"id","before":null,"after":4371}'),
                (1, '{"key":"status","before":null,"after":1}'),
                (1, '{"key":"startDate","before":null,"after":1514761200000}'),
                (1, '{"key":"endDate","before":null,"after":1546297199000}'),
                (2, '{"key":"billingDate","before":null,"after":{"year":2018,"month":"JANUARY","dayOfMonth":15,"dayOfWeek":"MONDAY","era":"CE","dayOfYear":15,"leapYear":false,"monthValue":1,"chronology":{"calendarType":"iso8601","id":"ISO"}}}')
        ) AS v(id, j), generate_series(1, 10) AS g
        GROUP BY g
)
SELECT j, (SELECT je->'after' FROM JSON_ARRAY_ELEMENTS(j) AS je WHERE je->>'key' = 'billingDate' AND (je->>'after') IS NOT NULL) AS export
FROM cte
WHERE j::TEXT LIKE '%after":{%'

结果：