分针网——每日分享:MySQL实现树的遍历
Posted
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了分针网——每日分享:MySQL实现树的遍历相关的知识,希望对你有一定的参考价值。
更多文章:www.f-z.cn
经常在一个表中有父子关系的两个字段,比如empno与manager,这种结构中需要用到树的遍历。在Oracle 中可以使用connect by简单解决问题,参见http://blog.csdn.net/wzy0623/archive/2007/06/18/1656345.aspx,但mysql 5.1中还不支持(据说已纳入to do中),要自己写过程或函数来实现。
一、建立测试表和数据
[c-sharp] view plain copy
DROP TABLE IF EXISTS `channel`;
CREATE TABLE `channel` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`cname` varchar(200) DEFAULT NULL,
`parent_id` int(11) DEFAULT NULL,
PRIMARY KEY (`id`)
) ENGINE=MyISAM AUTO_INCREMENT=19 DEFAULT CHARSET=utf8;
/*Data for the table `channel` */
insert into `channel`(`id`,`cname`,`parent_id`)
values (13,‘首页‘,-1),
(14,‘TV580‘,-1),
(15,‘生活580‘,-1),
(16,‘左上幻灯片‘,13),
(17,‘帮忙‘,14),
(18,‘栏目简介‘,17);
二、利用临时表和递归过程实现树的遍历(MySQL的UDF不能递归调用)
[c-sharp] view plain copy
DELIMITER $$
USE `db1`$$
-- 从某节点向下遍历子节点
-- 递归生成临时表数据
DROP PROCEDURE IF EXISTS `createChildLst`$$
CREATE PROCEDURE `createChildLst`(IN rootId INT,IN nDepth INT)
BEGIN
DECLARE done INT DEFAULT 0;
DECLARE b INT;
DECLARE cur1 CURSOR FOR SELECT id FROM channel WHERE parent_id=rootId;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
SET max_sp_recursion_depth=12;
INSERT INTO tmpLst VALUES (NULL,rootId,nDepth);
OPEN cur1;
FETCH cur1 INTO b;
WHILE done=0 DO
CALL createChildLst(b,nDepth+1);
FETCH cur1 INTO b;
END WHILE;
CLOSE cur1;
END$$
-- 从某节点向上追溯根节点
-- 递归生成临时表数据
DROP PROCEDURE IF EXISTS `createParentLst`$$
CREATE PROCEDURE `createParentLst`(IN rootId INT,IN nDepth INT)
BEGIN
DECLARE done INT DEFAULT 0;
DECLARE b INT;
DECLARE cur1 CURSOR FOR SELECT parent_id FROM channel WHERE id=rootId;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
SET max_sp_recursion_depth=12;
INSERT INTO tmpLst VALUES (NULL,rootId,nDepth);
OPEN cur1;
FETCH cur1 INTO b;
WHILE done=0 DO
CALL createParentLst(b,nDepth+1);
FETCH cur1 INTO b;
END WHILE;
CLOSE cur1;
END$$
-- 实现类似Oracle SYS_CONNECT_BY_PATH的功能
-- 递归过程输出某节点id路径
DROP PROCEDURE IF EXISTS `createPathLst`$$
CREATE PROCEDURE `createPathLst`(IN nid INT,IN delimit VARCHAR(10),INOUT pathstr VARCHAR(1000))
BEGIN
DECLARE done INT DEFAULT 0;
DECLARE parentid INT DEFAULT 0;
DECLARE cur1 CURSOR FOR
SELECT t.parent_id,CONCAT(CAST(t.parent_id AS CHAR),delimit,pathstr)
FROM channel AS t WHERE t.id = nid;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
SET max_sp_recursion_depth=12;
OPEN cur1;
FETCH cur1 INTO parentid,pathstr;
WHILE done=0 DO
CALL createPathLst(parentid,delimit,pathstr);
FETCH cur1 INTO parentid,pathstr;
END WHILE;
CLOSE cur1;
END$$
-- 递归过程输出某节点name路径
DROP PROCEDURE IF EXISTS `createPathnameLst`$$
CREATE PROCEDURE `createPathnameLst`(IN nid INT,IN delimit VARCHAR(10),INOUT pathstr VARCHAR(1000))
BEGIN
DECLARE done INT DEFAULT 0;
DECLARE parentid INT DEFAULT 0;
DECLARE cur1 CURSOR FOR
SELECT t.parent_id,CONCAT(t.cname,delimit,pathstr)
FROM channel AS t WHERE t.id = nid;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
SET max_sp_recursion_depth=12;
OPEN cur1;
FETCH cur1 INTO parentid,pathstr;
WHILE done=0 DO
CALL createPathnameLst(parentid,delimit,pathstr);
FETCH cur1 INTO parentid,pathstr;
END WHILE;
CLOSE cur1;
END$$
-- 调用函数输出id路径
DROP FUNCTION IF EXISTS `fn_tree_path`$$
CREATE FUNCTION `fn_tree_path`(nid INT,delimit VARCHAR(10)) RETURNS VARCHAR(2000) CHARSET utf8
BEGIN
DECLARE pathid VARCHAR(1000);
SET @pathid=CAST(nid AS CHAR);
CALL createPathLst(nid,delimit,@pathid);
RETURN @pathid;
END$$
-- 调用函数输出name路径
DROP FUNCTION IF EXISTS `fn_tree_pathname`$$
CREATE FUNCTION `fn_tree_pathname`(nid INT,delimit VARCHAR(10)) RETURNS VARCHAR(2000) CHARSET utf8
BEGIN
DECLARE pathid VARCHAR(1000);
SET @pathid=‘‘;
CALL createPathnameLst(nid,delimit,@pathid);
RETURN @pathid;
END$$
-- 调用过程输出子节点
DROP PROCEDURE IF EXISTS `showChildLst`$$
CREATE PROCEDURE `showChildLst`(IN rootId INT)
BEGIN
DROP TEMPORARY TABLE IF EXISTS tmpLst;
CREATE TEMPORARY TABLE IF NOT EXISTS tmpLst
(sno INT PRIMARY KEY AUTO_INCREMENT,id INT,depth INT);
CALL createChildLst(rootId,0);
SELECT channel.id,CONCAT(SPACE(tmpLst.depth*2),‘--‘,channel.cname) NAME,channel.parent_id,tmpLst.depth,fn_tree_path(channel.id,‘/‘) path,fn_tree_pathname(channel.id,‘/‘) pathname
FROM tmpLst,channel WHERE tmpLst.id=channel.id ORDER BY tmpLst.sno;
END$$
-- 调用过程输出父节点
DROP PROCEDURE IF EXISTS `showParentLst`$$
CREATE PROCEDURE `showParentLst`(IN rootId INT)
BEGIN
DROP TEMPORARY TABLE IF EXISTS tmpLst;
CREATE TEMPORARY TABLE IF NOT EXISTS tmpLst
(sno INT PRIMARY KEY AUTO_INCREMENT,id INT,depth INT);
CALL createParentLst(rootId,0);
SELECT channel.id,CONCAT(SPACE(tmpLst.depth*2),‘--‘,channel.cname) NAME,channel.parent_id,tmpLst.depth,fn_tree_path(channel.id,‘/‘) path,fn_tree_pathname(channel.id,‘/‘) pathname
FROM tmpLst,channel WHERE tmpLst.id=channel.id ORDER BY tmpLst.sno;
END$$
DELIMITER ;
三、测试
[c-sharp] view plain copy
CALL showChildLst(-1);
CALL showChildLst(13);
CALL showChildLst(14);
CALL showChildLst(17);
CALL showChildLst(18);
CALL showParentLst(-1);
CALL showParentLst(13);
CALL showParentLst(14);
CALL showParentLst(17);
CALL showParentLst(18);
四、遗留问题
1. 因为mysql对动态游标的支持不够,所以要想做成通用的过程或函数比较困难,可以利用两个临时表来转换(同时去掉了递归调用)参考http://jan.kneschke.de/projects/mysql/sp/sp_tree.sql,是个相对通用的实现。
2. 目前来看无论哪种实现,效率都不太好,希望mysql自己能实现Oracle 的connect by 功能,应该会比较优化。
加入职业技能圈 q群:272292492
以上是关于分针网——每日分享:MySQL实现树的遍历的主要内容,如果未能解决你的问题,请参考以下文章