废话哔哔
都说算法是程序的灵魂，算法源于数学，数学是描述宇宙万物的语言，这话一点不假，开发出身算法用的较少，回过头看算法，用到了，递归、循环、分支、分治、合并、取舍调优的思想，确实精彩，烧脑还挺有意思。
好的技术博客必须有做到有No BB,Show Code的干货，也得有说明辅助理解，因此写了这篇博客。
需求
- 概括：将一个字符字符串，拆分单个的拼音，例如bianchengyuyan（编程语言），拆分成bian、cheng、yu、yan。如果是zhang1san（张1三），zhangssan（张s三），则返回空数组。
- 应用场景：这算法并非百无一用，拼音分词，词义分析，品相判别等场景能够用到。
成品代码
<?php
/*** @function 返回拼音组合，由于冗长，单独拆分出来，拼音手动总结，可能存在遗漏* @return array*/
function pinYinArr() {return [1 => ['a', 'e', 'm', 'n', 'o'],2 => ['ai','an','ao','ba','bi','bo','bu','ca','ce','ci','cu','da','de','di','du','en','er','fa','fo','fu','ga','ge','gu','ha','he','hm','hu','ji','ju','ka','ke','ku','la','le','li','lo','lu','lv','ma','me','mi','mo','mu','na','ne','ng','ni','nu','nv','ou','pa','pi','po','pu','qi','qu','re','ri','ru','sa','se','si','su','ta','te','ti','tu','wa','wo','wu','xi','xu','ya','ye','yi','yo','yu','za','ze','zi','zu'],3 => ['ang','bai','ban','bao','bei','ben','bie','bin','cai','can','cao','cen','cha','che','chi','chu','cou','cui','cun','cuo','dai','dan','dao','dei','den','dia','die','diu','dou','dui','dun','duo','eng','fan','fei','fen','fou','gai','gan','gao','gei','gen','gou','gua','gui','gun','guo','hai','han','hao','hei','hen','hng','hou','hua','hui','hun','huo','jia','jie','jin','jiu','jue','jun','kai','kan','kao','kei','ken','kou','kua','kui','kun','kuo','lai','lan','lao','lei','lia','lie','lin','liu','lou','lue','lun','luo','mai','man','mao','mei','men','mie','min','miu','mou','nai','nan','nao','nei','nen','nie','nin','niu','nou','nue','nuo','pai','pan','pao','pei','pen','pie','pin','pou','qia','qie','qin','qiu','que','qun','ran','rao','ren','rou','rua','rui','run','ruo','sai','san','sao','sen','sha','she','shi','shu','sou','sui','sun','suo','tai','tan','tao','tie','tou','tui','tun','tuo','wai','wan','wei','wen','xia','xie','xin','xiu','xue','xun','yan','yao','yin','you','yue','yun','zai','zan','zao','zei','zen','zha','zhe','zhi','zhu','zou','zui','zun','zuo'],4 => ['bang','beng','bian','biao','bing','cang','ceng','chai','chan','chao','chen','chou','chua','chui','chun','chuo','cong','cuan','dang','deng','dian','diao','ding','dong','duan','fang','feng','gang','geng','gong','guai','guan','hang','heng','hong','huai','huan','jian','jiao','jing','juan','kang','keng','kong','kuai','kuan','lang','leng','lian','liao','ling','long','luan','mang','meng','mian','miao','ming','nang','neng','nian','niao','ning','nong','nuan','pang','peng','pian','piao','ping','qian','qiao','qing','quan','rang','reng','rong','ruan','sang','seng','shai','shan','shao','shei','shen','shou','shua','shui','shun','shuo','song','suan','tang','teng','tian','tiao','ting','tong','tuan','wang','weng','xian','xiao','xing','xuan','yang','ying','yong','yuan','zang','zeng','zhai','zhan','zhao','zhei','zhen','zhou','zhua','zhui','zhun','zhuo','zong','zuan'],5 => ['chang','cheng','chong','chuai','chuan','guang','huang','jiang','jiong','kuang','liang','niang','qiang','qiong','shang','sheng','shuai','shuan','xiang','xiong','zhang','zheng','zhong','zhuai','zhuan'],6 => ['chuang', 'shuang', 'zhuang'],];
}/*** @function 移除参数1中右边包含的参数2，并返回剩余的字符，例如strRemoveRightOnce('wahaha', 'ha')，返回waha* @param $string string 被操作字符串* @param $part string 要被移除的字符串* @return string*/
function strRemoveRightOnce($str, $part) {if (substr($str, -strlen($part)) == $part) {return substr($str, 0, - strlen($part));}return $str;
}/*** @function 获取字符串存在的拼音数量，不兼容-符号，从长往短了截取* @param $str string 字符* @param $result array 函数返回的结果* @return array*/
function pinYinCutLongToShort($str, $result = []) {if($str == '') {return $result;}if(($str == '') && ($result == [])) {return [];}//判断是否是纯拼音，不是直接过滤if((! preg_match('/^[a-z]+$/', $str)) && ($result == [])) {return [];}$initial_arr = pinYinArr();$initial_keys = array_keys($initial_arr);$max = max($initial_keys);$min = min($initial_keys);for($i = $max; $i >= $min; $i--) {$substring = substr($str, - $i);if(in_array($substring, $initial_arr[$i])) {array_unshift($result, $substring);//避免xiao ha ha，用ltrim函数，一次性移除掉了两个ha造成的计量有误return pinYinCutLongToShort(strRemoveRightOnce($str, $substring), $result);} else {if($i == $min) {return [];}}}return $result;
}print_r(pinYinCutLongToShort('bianchengyuyan'));
Array
([0] => bian[1] => cheng[2] => yu[3] => yan
)
算法调优注意的地方
- 如果检测到含有非拼音的字符（例如有数字），以及多余的拼音字符（例如zhangxsan张x三），会直接返回空数组。
- 代码采用从长往短切割的策略，以xianggang（香港）举例：
- 从长到短（for循环递减）：分成xiang、gang。粒度大，但是失败率小。
- 从短到长（for循环递增）：拆分成xi、an，后面的gg没法切了。粒度更小，容易出错。
- 从长到短缺点也很明显：相应的会忽略精度，所以xian（西安）会记作一个拼音，当做xian（贤）处理。
- 代码采用的从字符串右边往左切的策略，进一步避免切割出错。例如xianguang（闲逛）：
- 从左到右从长到短：xiang、uang没办法切了。
- 从左到右从短到长：xi、an、gu、an、g没办法切了。
- 从右到左从长到短：guang、xian，刚刚好。
- 从右到左从短到长：ang、gu、an、xi，也行。
- 结语：从长往短了切割用于减少失败率，从右往左切割，用于进一步避免出错，因此被采用。
- 注意：以上算法，并非适合所有场景，可能存在误差，毕竟没有NLP的AI算法加持（自然语言处理）。
- 补充：若读者想要获取最细粒度的拼音，不必再原有函数上改动，可以将返回的数组结果遍历，再次调用另一个切割函数（注意另一个函数是从短到长切割），随后汇总。