文章目录
1 基础数据
我们以以下数据为例进行说明
2 继承关系驱动的架构设计
2.1 表结构
| id | parent_id | name |
|---|---|---|
| 1 | A | |
| 2 | 1 | AA |
| 3 | 1 | AB |
| 4 | 3 | ABA |
| 5 | 3 | ABB |
| 6 | 3 | ABC |
| 7 | 1 | AC |
| 8 | 7 | ACA |
| 9 | 8 | ACAA |
| 10 | 8 | ACAB |
2.2 方案的优点及缺点
- 优点: 设计和实现简单, 直观
- 缺点: CURD操作是低效的, 主要归根于频繁的“递归”操作导致的IO开销
- 解决方案: 在数据规模较小的情况下可以通过缓存机制来优化
3 基于左右值编码的架构设计
关于此方案的设计可以查看另一篇博客, 本人也是通过查看此篇博客学习的, 一些说明也是直接粘过来的, 所以部分细节我这里不再说明, 本篇博客与其的区别主要在于第四节
在基于数据库的一般应用中，查询的需求总要大于删除和修改。为了避免对于树形结构查询时的“递归”过程，基于Tree的前序遍历设计一种全新的无递归查询、无限分组的左右值编码方案，来保存该树的数据。
3.1 表结构
| id | left | right | name |
|---|---|---|---|
| 1 | 1 | 20 | A |
| 2 | 2 | 3 | AA |
| 3 | 4 | 11 | AB |
| 4 | 5 | 6 | ABA |
| 5 | 7 | 8 | ABB |
| 6 | 9 | 10 | ABC |
| 7 | 12 | 19 | AC |
| 8 | 13 | 18 | ACA |
| 9 | 14 | 15 | ACAA |
| 10 | 16 | 17 | ACAB |
第一次看见这种表结构，相信大部分人都不清楚左值（left）和右值（right）是如何计算出来的，而且这种表设计似乎并没有保存父子节点的继承关系。但当你用手指指着表中的数字从1数到20，你应该会发现点什么吧。对，你手指移动的顺序就是对这棵树进行前序遍历的顺序，如下图所示。当我们从根节点A左侧开始，标记为1，并沿前序遍历的方向，依次在遍历的路径上标注数字，最后我们回到了根节点A，并在右边写上了20。
3.2 方案优缺点
- 优点:
- 可以方便的查询出某个节点的所有子孙节点
- 可以方便的获取某个节点的族谱路径(即所有的上级节点)
- 可已通过自身的left, right值计算出共有多少个子孙节点
- 缺点:
- 增删及移动节点操作比较复杂
- 无法简单的获取某个节点的子节点
4 基于继承关系及左右值编码的架构设计
其实就是在第三节的基础上又加了一列parent_id, 目的是在保留上述优点的同时可以简单的获取某个节点的直属子节点
4.1 表结构
| id | parent_id | left | right | name |
|---|---|---|---|---|
| 1 | 1 | 20 | A | |
| 2 | 1 | 2 | 3 | AA |
| 3 | 1 | 4 | 11 | AB |
| 4 | 3 | 5 | 6 | ABA |
| 5 | 3 | 7 | 8 | ABB |
| 6 | 3 | 9 | 10 | ABC |
| 7 | 1 | 12 | 19 | AC |
| 8 | 7 | 13 | 18 | ACA |
| 9 | 8 | 14 | 15 | ACAA |
| 10 | 8 | 16 | 17 | ACAB |
4.2 CURD操作
4.2.1 create node
# 为id为 id_ 的节点创建名为 name_ 的子节点
CREATE PROCEDURE `tree_create_node`(IN `id_` INT, IN `name_` VARCHAR(50))
LANGUAGE SQL
NOT DETERMINISTIC
CONTAINS SQL
SQL SECURITY DEFINER
COMMENT '创建节点'
BEGIN
declare right1 int;
# 当 id_ 为 0 时表示创建根节点
if id_ = 0 then
# 此处我限制了仅允许存在一个根节点, 当然这并不是必须的
if exists(select `id` from tree_table where `left` = 1) then
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = '根节点已存在';
end if;
insert into tree_table(`parent_id`, `name`, `left`, `right`)
values(0, name_, 1, 2);
commit;
elseif exists(select `id` from tree_table where `parent_id` = id_ and `name` = name_) then
# 禁止在同一级创建同名节点
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = '已存在同名兄弟节点';
elseif exists(select `id` from tree_table where `id` = id_ and `is_delete` = 0) then
start transaction;
set right1=(select `right` from tree_table where `id` = id_);
update tree_table set `right` = `right` + 2 where `right` >= right1;
update tree_table set `left` = `left` + 2 where `left` >= right1;
insert into tree_table(`parent_id`, `name`, `left`, `right`)
values(id_, name_, right1, right1 + 1);
commit;
# 下面一行仅为了展示以下新插入记录的id, 并不是必须的
select LAST_INSERT_ID();
else
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = '父节点不存在(未创建或被删除)';
end if;
END
# 创建根节点
call tree_create_node(0, 'A')
# 为节点1创建名为AB的子节点
call tree_create_node(1, 'AB')
4.2.2 delete node
CREATE PROCEDURE `tree_delete_node`(IN `id_` INT)
LANGUAGE SQL
NOT DETERMINISTIC
CONTAINS SQL
SQL SECURITY DEFINER
COMMENT ''
BEGIN
declare left1 int;
declare right1 int;
if exists(select id from tree_table where id = id_) then
start transaction;
select `left`, `right` into left1, right1 from tree_table where id = id_;
delete from tree_table where `left` >= left1 and `right` <= right1;
update tree_table set `left` = `left` - (right1-left1+1) where `left` > left1;
update tree_table set `right` = `right` - (right1-left1+1) where `right` > right1;
commit;
end if;
END
# 删除节点2, 节点2的子孙节点也会被删除
call tree_delete_node(2)
4.2.3 move node
move的原理是先删除再添加, 但涉及被移动的节点的left, right值不能乱所以需要使用临时表(由于在存储过程中无法创建临时表, 此处我使用了一张正常的表进行缓存, 欢迎提出更合理的方案)
# 此存储过程中涉及到is_delete字段, 表示数据是否被删除, 因为正式环境中删除操作一般都不会真的删除而是进行软删(即标记删除), 如果不需要此字段请自行对程序进行调整
CREATE PROCEDURE `tree_move_node`(IN `self_id` INT, IN `parent_id` INT
, IN `sibling_id` INT)
LANGUAGE SQL
NOT DETERMINISTIC
CONTAINS SQL
SQL SECURITY DEFINER
COMMENT ''
BEGIN
declare self_left int;
declare self_right int;
declare parent_left int;
declare parent_right int;
declare sibling_left int;
declare sibling_right int;
declare sibling_parent_id int;
if exists(select id from tree_table where id = parent_id and is_delete = 0) then
# 创建中间表
CREATE TABLE If Not Exists tree_table_self_ids (`id` int(10) unsigned NOT NULL);
truncate tree_table_self_ids;
start transaction; # 事务
# 获取移动对象的 left, right 值
select `left`, `right` into self_left, self_right from tree_table where id = self_id;
# 将需要移动的记录的 id 存入临时表, 以保证操作 left, right 值变化时这些记录不受影响
insert into tree_table_self_ids(id) select id from tree_table where `left` >= self_left and `right` <= self_right;
# 将被移动记录后面的记录往前移, 填充空缺位置
update tree_table set `left` = `left` - (self_right-self_left+1) where `left` > self_left and id not in (select id from tree_table_self_ids);
update tree_table set `right` = `right` - (self_right-self_left+1) where `right` > self_right and id not in (select id from tree_table_self_ids);
select `left`, `right` into parent_left, parent_right from tree_table where id = parent_id;
if sibling_id = -1 then
# 在末尾插入子节点
update tree_table set `right` = `right` + (self_right-self_left+1) where `right` >= parent_right and id not in (select id from tree_table_self_ids);
update tree_table set `left` = `left` + (self_right-self_left+1) where `left` >= parent_right and id not in (select id from tree_table_self_ids);
update tree_table set `right`=`right` + (parent_right-self_left), `left`=`left` + (parent_right-self_left) where id in (select id from tree_table_self_ids);
elseif sibling_id = 0 then
# 在开头插入子节点
update tree_table set `right` = `right` + (self_right-self_left+1) where `right` > parent_left and id not in (select id from tree_table_self_ids);
update tree_table set `left` = `left` + (self_right-self_left+1) where `left` > parent_left and id not in (select id from tree_table_self_ids);
update tree_table set `right`=`right` - (self_left-parent_left-1), `left`=`left` - (self_left-parent_left-1) where id in (select id from tree_table_self_ids);
else
# 插入指定节点之后
select `left`, `right`, `parent_id` into sibling_left, sibling_right, sibling_parent_id from tree_table where id = sibling_id;
if parent_id != sibling_parent_id then
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = '指定的兄弟节点不在指定的父节点中';
end if;
update tree_table set `right` = `right` + (self_right-self_left+1) where `right` > sibling_right and id not in (select id from ctree_table_self_ids);
update tree_table set `left` = `left` + (self_right-self_left+1) where `left` > sibling_right and id not in (select id from tree_table_self_ids);
update tree_table set `right`=`right` - (self_left-sibling_right-1), `left`=`left` - (self_left-sibling_right-1) where id in (select id from tree_table_self_ids);
end if;
update tree_table set `parent_id`=parent_id where `id` = self_id;
commit;
else
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = '父节点不存在(未创建或被删除)';
end if;
END
# 将节点2移动到节点1下面开头的位置
call tree_move_node(2, 1, 0)
# 将节点2移动到节点1下面末尾的位置
call tree_move_node(2, 1, -1)
# 将节点2移动到节点1下面且跟在节点3后面的位置
call tree_move_node(2, 1, 3)
4.2.4 select
# 以下sql中需要传的值全用???表示
# 根据节点id获取此节点所有子孙节点
select * from tree_table where
left > (select left from tree_table where id=???) and
right < (select right from tree_table where id=???)
# 根据节点id获取此节点的所有子孙节点(包含自己)
select * from tree_table where
left >= (select left from tree_table where id=???) and
right <= (select right from tree_table where id=???)
# 根据节点id获取此节点的所有上级节点
select * from tree_table where
left < (select left from tree_table where id=???) and
right > (select right from tree_table where id=???)
# 根据节点id获取此节点的所有上级节点(包括自己)
select * from tree_table where
left <= (select left from tree_table where id=???) and
right >= (select right from tree_table where id=???)
5 存储完整父级路径的架构设计
5.1 表结构
| id | parent_id | parents_id | name |
|---|---|---|---|
| 1 | 0 | A | |
| 2 | 1 | 1 | AA |
| 3 | 1 | 1 | AB |
| 4 | 3 | 1,3 | ABA |
| 5 | 3 | 1,3 | ABB |
| 6 | 3 | 1,3 | ABC |
| 7 | 1 | 1 | AC |
| 8 | 7 | 1,7 | ACA |
| 9 | 8 | 1,7,8 | ACAA |
| 10 | 8 | 1,7,8 | ACAB |
给parents_id字段加上全文索引就可以方便的查出某个节点的所有下级数据了
须注意的mysql全文索引默认是无法识别出短数字的, 需要先修改配置
配置参数参考另一篇文章
将上述参数改为1即可正常使用全文索引搜索数据(如果索引已经建好, 改完数据库配置需要重建索引才能生效)
当然不想使用全文索引的话, 正常的索引使用前缀查询也可以相对方便的查出数据
5.2 方案优缺点
- 创建节点时需先查出父节点的信息
- 当移动节点时更新数据较多
6 总结
此篇文章对左右值编码结构的原理介绍的不多, 需要详细了解的可以查阅末尾引用的博客
扫一扫
9万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
