用 Lilac Parser 代替正则来抓取文本的例子

lilac-parser 是我用 ClojureScript 实现的一个库, 可以做一些正则的功能.
看名字, 这个库设计的时候更多是一个 parser 的思路,
从使用来说, 当做一个正则也是比较顺的. 虽然不如正则简短明了.
正则的缺点主要是基于字符串形态编写, 需要转义, 规则长了就不好维护了.
而 lilac-parser 的方式, 就挺容易进行组合的, 我这边举一些例子

首先是 is+ 这个规则, 进行精确匹配,

(parse-lilac "x" (is+ "x"))      ; {:ok? true, :rest nil}
(parse-lilac "xyz" (is+ "xyz"))  ; {:ok? true, :rest nil}
(parse-lilac "xy" (is+ "x"))     ; {:ok? false}
(parse-lilac "xy" (is+ "x"))     ; {:ok? true, :rest ("y")}
(parse-lilac "y" (is+ "x"))      ; {:ok? false}

可以看到, 头部匹配上的表达式, 都返回了 true.
后边是否还有其他内容, 需要通过 :rest 字段再去单独判断了.

当然精确匹配比较简单, 然后是选择匹配,

(parse-lilac "x" (one-of+ "xyz"))  ; {:ok? true}
(parse-lilac "y" (one-of+ "xyz"))  ; {:ok? true}
(parse-lilac "z" (one-of+ "xyz"))  ; {:ok? true}
(parse-lilac "w" (one-of+ "xyz"))  ; {:ok? false}
(parse-lilac "xy" (one-of+ "xyz")) ; {:ok? true, :rest ("y")}

反过来, 可以有排除的规则,

(parse-lilac "x" (other-than+ "abc"))  ; {:ok? true, :rest nil}
(parse-lilac "xy" (other-than+ "abc")) ; {:ok? true, :rest ("y")}
(parse-lilac "a" (other-than+ "abc"))  ; {:ok? false}

在此基础上, 增加一些逻辑, 表示判断的规则可以不存在,
当然允许不存在的话, 任何时候都可以退回到 true 的结果的,

(parse-lilac "x" (optional+ (is+ "x"))) ; {:ok? true, :rest nil}
(parse-lilac "" (optional+ (is+ "x"))) ; {:ok? true, :rest nil}
(parse-lilac "x" (optional+ (is+ "y"))) ; {:ok? true, :rest("x")}

也可以设定规则, 判断多个, 也就是大于 1 个(目前不能控制具体个数),

(parse-lilac "x" (many+ (is+ "x")))
(parse-lilac "xx" (many+ (is+ "x")))
(parse-lilac "xxx" (many+ (is+ "x")))
(parse-lilac "xxxy" (many+ (is+ "x")))

如果允许 0 个的情况, 就不是 many 了, 而是 some 的规则,

(parse-lilac "" (some+ (is+ "x")))
(parse-lilac "x" (some+ (is+ "x")))
(parse-lilac "xx" (some+ (is+ "x")))
(parse-lilac "xxy" (some+ (is+ "x")))
(parse-lilac "y" (some+ (is+ "x")))

相应的, or 的规则可以写出来,

(parse-lilac "x" (or+ [(is+ "x") (is+ "y")]))
(parse-lilac "y" (or+ [(is+ "x") (is+ "y")]))
(parse-lilac "z" (or+ [(is+ "x") (is+ "y")]))

而 combine 是用来顺序组合多个规则的,

(parse-lilac "xy" (combine+ [(is+ "x") (is+ "y")]))  ; {:ok? true, :rest nil}
(parse-lilac "xyz" (combine+ [(is+ "x") (is+ "y")])) ; {:ok? true, :rest ("z")}
(parse-lilac "xy" (combine+ [(is+ "y") (is+ "x")]))  ; {:ok? flase}

而 interleave 是表示两个规则, 然后相互间隔重复,
这种场景很多都是逗号间隔的表达式的处理当中用到,

(parse-lilac "xy" (interleave+ (is+ "x") (is+ "y")))
(parse-lilac "xyx" (interleave+ (is+ "x") (is+ "y")))
(parse-lilac "xyxy" (interleave+ (is+ "x") (is+ "y")))
(parse-lilac "yxy" (interleave+ (is+ "x") (is+ "y")))

另外当前的代码还提供了几个内置的规则, 用来判断字母, 数字, 中文的情况,

(parse-lilac "a" lilac-alphabet)
(parse-lilac "A" lilac-alphabet)
(parse-lilac "." lilac-alphabet) ; {:ok? false}

(parse-lilac "1" lilac-digit)
(parse-lilac "a" lilac-digit) ; {:ok? false}

(parse-lilac "汉" lilac-chinese-char)
(parse-lilac "E" lilac-chinese-char)  ; {:ok? false}
(parse-lilac "," lilac-chinese-char)  ; {:ok? false}
(parse-lilac "，" lilac-chinese-char) ; {:ok? false}

具体某些特殊的字符的话, 暂时只能通过 unicode 范围来指定了.

(parse-lilac "a" (unicode-range+ 97 122))
(parse-lilac "z" (unicode-range+ 97 122))
(parse-lilac "A" (unicode-range+ 97 122))

有了这些规则, 就可以组合来模拟正则的功能了, 比如查找匹配项有多少,

(find-lilac "write cumulo and respo" (or+ [(is+ "cumulo") (is+ "respo")]))
; find 2
(find-lilac "write cumulo and phlox" (or+ [(is+ "cumulo") (is+ "respo")]))
; find 1
(find-lilac "write cumulo and phlox" (or+ [(is+ "cirru") (is+ "respo")]))
; find 0

或者直接进行字符串替换, 这就跟正则差不多了.

(replace-lilac "cumulo project" (or+ [(is+ "cumulo") (is+ "respo")]) (fn [x] "my"))
; "my project"
(replace-lilac "respo project" (or+ [(is+ "cumulo") (is+ "respo")]) (fn [x] "my"))
; "my project"
(replace-lilac "phlox project" (or+ [(is+ "cumulo") (is+ "respo")]) (fn [x] "my"))
; "phlox project"

可以看到, 这个写法就是组合出来的, 写起来比正则长, 但是可以定义变量, 做一些抽象.

简单的例子可能看不出这样做有什么用, 可能就是觉得搞得反而更长了, 而且性能更差.
我的项目当中有个简单的 JSON 解析的例子, 这个用正则就搞不定了吧...
直接搬运代码如下:

; 判断 true false 两种情况, 返回的是 boolean
(def boolean-parser
  (label+ "boolean" (or+ [(is+ "true") (is+ "false")] (fn [x] (if (= x "true") true false)))))

(def space-parser (label+ "space" (some+ (is+ " ") (fn [x] nil))))

; 组合一个包含空白和逗号的解析器, label 只是注释, 可以忽略
(def comma-parser
  (label+ "comma" (combine+ [space-parser (is+ ",") space-parser] (fn [x] nil))))

(def digits-parser (many+ (one-of+ "0123456789") (fn [xs] (string/join "" xs))))

; 为了简单, null 和 undefined 直接返回 nil 了
(def nil-parser (label+ "nil" (or+ [(is+ "null") (is+ "undefined")] (fn [x] nil))))

; number 的情况, 需要考虑前面可能有负号, 后面可能有小数点
; 这边偷懒没考虑科学记数法了...
(def number-parser
  (label+
   "number"
   (combine+
     ; 负号.. 可选的
    [(optional+ (is+ "-"))
     digits-parser
                ; 组合出来小数部分, 这也是可选的
     (optional+ (combine+ [(is+ ".") digits-parser] (fn [xs] (string/join "" xs))))]
    (fn [xs] (js/Number (string/join "" xs))))))

(def string-parser
  (label+
   "string"
   (combine+
     ; 字符串的解析, 引号开头引号结尾
    [(is+ "\"")
            ; 中间是非引号的字符串, 或者转义符号的情况
     (some+ (or+ [(other-than+ "\"\\") (is+ "\\\"") (is+ "\\\\") (is+ "\\n")]))
     (is+ "\"")]
    (fn [xs] (string/join "" (nth xs 1))))))

(defparser
 value-parser+
 ()
 identity
 (or+
  [number-parser string-parser nil-parser boolean-parser (array-parser+) (object-parser+)]))

(defparser
 object-parser+
 ()
 identity
 (combine+
  [(is+ "{")
   (optional+
     ; 对象就比较复杂了, 主要看 interleave 部分吧, 外边只是花括号的处理
    (interleave+
     (combine+
      [string-parser space-parser (is+ ":") space-parser (value-parser+)]
      (fn [xs] [(nth xs 0) (nth xs 4)]))
     comma-parser
     (fn [xs] (take-nth 2 xs))))
   (is+ "}")]
  (fn [xs] (into {} (nth xs 1)))))

(defparser
 array-parser+
 ()
 (fn [x] (vec (first (nth x 1))))
 (combine+
  [(is+ "[")
          ; 数组, 同样是 interleave 的情况
   (some+ (interleave+ (value-parser+) comma-parser (fn [xs] (take-nth 2 xs))))
   (is+ "]")]))

可以看到, 通过 lilac-parser 构造规则的当时, 比较容易就生成了一个 JSON Parser.
虽然支持的规则比较简单, 而且性能不大理想, 但是比起正则来说, 这个代码可读很多了.
相信可以作为一种思路, 用在很多文本处理的场景当中.
为了也许可以提供简化一些的版本, 在 JavaScript 直接使用, 代替正则.