这里的设置:
create table t (id serial primary key, content json);
insert into t set content = '{"tags": ["tag_1", "tag_2"]}';
insert into t set content = '{"tags": ["tag_3", "tag_2"]}';
insert into t set content = '{"tags": ["tag_1", "tag_2"]}';
如果你知道代码会在任何标记阵列的最大数量,你可以使用UNION提取所有标签:
select id, json_extract(content, '$.tags[0]') AS tag from t
union
select id, json_extract(content, '$.tags[1]') from t;
+----+---------+
| id | tag |
+----+---------+
| 1 | "tag_1" |
| 2 | "tag_3" |
| 3 | "tag_1" |
| 1 | "tag_2" |
| 2 | "tag_2" |
| 3 | "tag_2" |
+----+---------+
你需要尽可能多的被联合子查询的最长阵列中的标签数量。
然后,你可以把它放进派生表,并对其执行聚合:
select tag, count(*) as count
from (
select id, json_extract(content, '$.tags[0]') as tag from t
union
select id, json_extract(content, '$.tags[1]') from t
) as t2
group by tag
order by count desc;
+---------+-------+
| tag | count |
+---------+-------+
| "tag_2" | 3 |
| "tag_1" | 2 |
| "tag_3" | 1 |
+---------+-------+
这将是更容易,如果你存储在第二表,而不是在一个JSON数组变量:
create table tags (id bigint unsigned, tag varchar(20) not null, primary key (id, tag));
insert into tags set id = 1, tag = 'tag_1';
insert into tags set id = 1, tag = 'tag_2';
insert into tags set id = 2, tag = 'tag_3';
insert into tags set id = 2, tag = 'tag_2';
insert into tags set id = 3, tag = 'tag_1';
insert into tags set id = 3, tag = 'tag_2';
select tag, count(*) as count
from tags
group by tag
order by count desc;
+-------+-------+
| tag | count |
+-------+-------+
| tag_2 | 3 |
| tag_1 | 2 |
| tag_3 | 1 |
+-------+-------+
无论您拥有多少个标签,此解决方案都能正常工作。您不需要知道每个ID的标签列表的最大长度。
当您需要存储半结构化数据的“文档”时,但只有当您将文档视为不可简化的数据值时,JSON才是很好的选择。只要您需要访问文档的元素并对其应用关系操作,面向文档的方法就显示出其弱点。
谢谢!很有帮助。在这工作一个星期! – user43857