概述

实际生活有很多树形结构的数据，比如公司分为多个部门，部门下分为多个组，组下分为多个员工；省市县的归属；页面菜单栏等等。

如果想查询某个节点的父节点或者子节点，一般通过表自身连接完成，但如果该节点的子节点还有多层结构，就需要使用递归调用。但如果数据量特别大，递归的次数指数级上升，而且查询数据库的次数也指数级上升，导致程序和数据库压力剧增，查询时间特别长。那数据库有没有递归查询语句呢？答案是肯定的。

start with connect by prior 递归查询

1、数据准备

create table area_test(
  id         number(10) not null,
  parent_id  number(10),
  name       varchar2(255) not null
);

alter table area_test add (constraint district_pk primary key (id));

insert into area_test (ID, PARENT_ID, NAME) values (1, null, '中国');
insert into area_test (ID, PARENT_ID, NAME) values (11, 1, '河南省'); 
insert into area_test (ID, PARENT_ID, NAME) values (12, 1, '北京市');
insert into area_test (ID, PARENT_ID, NAME) values (111, 11, '郑州市');
insert into area_test (ID, PARENT_ID, NAME) values (112, 11, '平顶山市');
insert into area_test (ID, PARENT_ID, NAME) values (113, 11, '洛阳市');
insert into area_test (ID, PARENT_ID, NAME) values (114, 11, '新乡市');
insert into area_test (ID, PARENT_ID, NAME) values (115, 11, '南阳市');
insert into area_test (ID, PARENT_ID, NAME) values (121, 12, '朝阳区');
insert into area_test (ID, PARENT_ID, NAME) values (122, 12, '昌平区');
insert into area_test (ID, PARENT_ID, NAME) values (1111, 111, '二七区');
insert into area_test (ID, PARENT_ID, NAME) values (1112, 111, '中原区');
insert into area_test (ID, PARENT_ID, NAME) values (1113, 111, '新郑市');
insert into area_test (ID, PARENT_ID, NAME) values (1114, 111, '经开区');
insert into area_test (ID, PARENT_ID, NAME) values (1115, 111, '金水区');
insert into area_test (ID, PARENT_ID, NAME) values (1121, 112, '湛河区');
insert into area_test (ID, PARENT_ID, NAME) values (1122, 112, '舞钢市');
insert into area_test (ID, PARENT_ID, NAME) values (1123, 112, '宝丰市');
insert into area_test (ID, PARENT_ID, NAME) values (11221, 1122, '尚店镇');

2 start with connect by prior递归查询

• start with 子句：遍历起始条件。如果要查父结点，这里可以用子结点的列，反之亦然。
• connect by 子句：连接条件。prior 跟父节点列parentid放在一起，就是往父结点方向遍历；prior 跟子结点列subid放在一起，则往叶子结点方向遍历。parent_id、id两列谁放在 “=” 前都无所谓，关键是prior跟谁在一起。
• order by 子句：排序。

常用的select项：

LEVEL：级别
connect_by_root：根节点
sys_connect_by_path：递归路径

2.1 查询所有子节点

select t.*,LEVEL
from area_test t
start with name ='郑州市'
connect by prior id=parent_id

其实，如果单层结构，使用表自身连接也可以实现：

select * from area_test t1,area_test t2
where t1.ID = t2.PARENT_ID and t1.NAME='郑州市'

当查询节点下有多层数据：

select t.*,LEVEL
from area_test t
start with name ='河南省'
connect by prior id=parent_id

select * from area_test t1,area_test t2 
where t1.PARENT_ID = t2.ID and t2.name='河南省';

如果使用自身连接，也只能查到子一级节点的数据，需要遍历子一级节点，递归查询每个子一级节点下的子节点。明显麻烦很多！！！

2.2 查询所有父节点

select t.*,level
from area_test t
start with name ='郑州市'
connect by prior t.parent_id=t.id
order by level asc;

2.3 查询指定节点的根节点

select d.*,
	   connect_by_root(d.id) rootid,
	   connect_by_root(d.name) rootname
from area_test d
where name='二七区'
start with d.parent_id IS NULL
connect by prior d.id=d.parent_id

select d.*,
	   connect_by_root(d.id) rootid,
	   connect_by_root(d.name) rootname
from area_test d
start with d.parent_id IS NULL
connect by prior d.id=d.parent_id

2.4 查询下行政组织递归路径

select id, parent_id, name, sys_connect_by_path(name, '->') namepath, level
from area_test
start with name = '平顶山市'
connect by prior id = parent_id

3 with递归查询

3.1 with递归子类

with tmp(id, parent_id, name) 
as (
	select id, parent_id, name
    from area_test
    where name = '平顶山市'
    union all
    select d.id, d.parent_id, d.name
    from tmp, area_test d
    where tmp.id = d.parent_id
   )
select * from tmp;

3.2 递归父类

with tmp(id, parent_id, name) 
as
  (
   select id, parent_id, name
   from area_test
   where name = '二七区'
   union all
   select d.id, d.parent_id, d.name
   from tmp, area_test d
   where tmp.parent_id = d.id
   )
select * from tmp;

4 MySQL 递归查找树形结构

参考文章：MySQL 递归查找树形结构，这个方法太实用了

参考文章：Oracle递归查询