|
1 // Parser Combinators: Simple Version |
|
2 //==================================== |
|
3 |
|
4 /* |
|
5 Note, in the lectures I did not show the implicit type constraint |
|
6 I : IsSeq, which means that the input type 'I' needs to be |
|
7 a sequence. |
|
8 */ |
|
9 |
|
10 type IsSeq[A] = A => Seq[_] |
|
11 |
|
12 abstract class Parser[I : IsSeq, T]{ |
|
13 def parse(in: I): Set[(T, I)] |
|
14 |
|
15 def parse_all(in: I) : Set[T] = |
|
16 for ((hd, tl) <- parse(in); |
|
17 if tl.isEmpty) yield hd |
|
18 } |
|
19 |
|
20 // parser combinators |
|
21 |
|
22 // sequence parser |
|
23 class SeqParser[I : IsSeq, T, S](p: => Parser[I, T], |
|
24 q: => Parser[I, S]) extends Parser[I, (T, S)] { |
|
25 def parse(in: I) = |
|
26 for ((hd1, tl1) <- p.parse(in); |
|
27 (hd2, tl2) <- q.parse(tl1)) yield ((hd1, hd2), tl2) |
|
28 } |
|
29 |
|
30 // alternative parser |
|
31 class AltParser[I : IsSeq, T](p: => Parser[I, T], |
|
32 q: => Parser[I, T]) extends Parser[I, T] { |
|
33 def parse(in: I) = p.parse(in) ++ q.parse(in) |
|
34 } |
|
35 |
|
36 // parser map |
|
37 class MapParser[I : IsSeq, T, S](p: => Parser[I, T], |
|
38 f: T => S) extends Parser[I, S] { |
|
39 def parse(in: I) = for ((hd, tl) <- p.parse(in)) yield (f(hd), tl) |
|
40 } |
|
41 |
|
42 // an example of an atomic parser for characters |
|
43 case class CharParser(c: Char) extends Parser[String, Char] { |
|
44 def parse(in: String) = |
|
45 if (in != "" && in.head == c) Set((c, in.tail)) else Set() |
|
46 } |
|
47 |
|
48 // an atomic parser for parsing strings according to a regex |
|
49 import scala.util.matching.Regex |
|
50 |
|
51 case class RegexParser(reg: Regex) extends Parser[String, String] { |
|
52 def parse(in: String) = reg.findPrefixMatchOf(in) match { |
|
53 case None => Set() |
|
54 case Some(m) => Set((m.matched, m.after.toString)) |
|
55 } |
|
56 } |
|
57 |
|
58 // atomic parsers for numbers and "verbatim" strings |
|
59 val NumParser = RegexParser("[0-9]+".r) |
|
60 def StrParser(s: String) = RegexParser(Regex.quote(s).r) |
|
61 |
|
62 // NumParserInt transforms a "string integer" into a propper Int |
|
63 // (needs "new" because MapParser is not a case class) |
|
64 |
|
65 val NumParserInt = new MapParser(NumParser, (s: String) => s.toInt) |
|
66 |
|
67 |
|
68 // the following string interpolation allows us to write |
|
69 // StrParser(_some_string_) more conveniently as |
|
70 // |
|
71 // p"<_some_string_>" |
|
72 |
|
73 implicit def parser_interpolation(sc: StringContext) = new { |
|
74 def p(args: Any*) = StrParser(sc.s(args:_*)) |
|
75 } |
|
76 |
|
77 // more convenient syntax for parser combinators |
|
78 implicit def ParserOps[I : IsSeq, T](p: Parser[I, T]) = new { |
|
79 def ||(q : => Parser[I, T]) = new AltParser[I, T](p, q) |
|
80 def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q) |
|
81 def map[S](f: => T => S) = new MapParser[I, T, S](p, f) |
|
82 } |
|
83 |
|
84 // these implicits allow us to use an infix notation for |
|
85 // sequences and alternatives; we also can write the usual |
|
86 // map for a MapParser |
|
87 |
|
88 |
|
89 // with this NumParserInt can now be written more conveniently |
|
90 // as: |
|
91 |
|
92 val NumParserInt2 = NumParser.map(s => s.toInt) |
|
93 |
|
94 |
|
95 // A parser for palindromes (just returns them as string) |
|
96 lazy val Pal : Parser[String, String] = { |
|
97 (p"a" ~ Pal ~ p"a").map{ case ((x, y), z) => s"$x$y$z" } || |
|
98 (p"b" ~ Pal ~ p"b").map{ case ((x, y), z) => s"$x$y$z" } || |
|
99 p"a" || p"b" || p"" |
|
100 } |
|
101 |
|
102 // examples |
|
103 Pal.parse_all("abaaaba") |
|
104 Pal.parse("abaaaba") |
|
105 |
|
106 println("Palindrome: " + Pal.parse_all("abaaaba")) |
|
107 |
|
108 // A parser for wellnested parentheses (transforms '(' -> '{' , ')' -> '}' ) |
|
109 lazy val P : Parser[String, String] = |
|
110 (p"(" ~ P ~ p")" ~ P).map{ case (((_, x), _), y) => "{" + x + "}" + y } || p"" |
|
111 |
|
112 println(P.parse_all("(((()()))())")) |
|
113 println(P.parse_all("(((()()))()))")) |
|
114 println(P.parse_all(")(")) |
|
115 println(P.parse_all("()")) |
|
116 |
|
117 // A parser for arithmetic expressions (Terms and Factors) |
|
118 |
|
119 lazy val E: Parser[String, Int] = |
|
120 (T ~ p"+" ~ E).map{ case ((x, _), z) => x + z } || |
|
121 (T ~ p"-" ~ E).map{ case ((x, _), z) => x - z } || T |
|
122 lazy val T: Parser[String, Int] = |
|
123 (F ~ p"*" ~ T).map{ case ((x, _), z) => x * z } || F |
|
124 lazy val F: Parser[String, Int] = |
|
125 (p"(" ~ E ~ p")").map{ case ((_, y), _) => y } || NumParserInt |
|
126 |
|
127 /* same parser but producing a string |
|
128 lazy val E: Parser[String, String] = |
|
129 (T ~ "+" ~ E).map{ case x ~ y ~ z => "(" + x + ")+(" + z + ")"} || T |
|
130 lazy val T: Parser[String, String] = |
|
131 (F ~ "*" ~ T).map{ case x ~ y ~ z => "(" + x + ")*("+ z + ")"} || F |
|
132 lazy val F: Parser[String, String] = |
|
133 ("(" ~ E ~ ")").map{ case x ~ y ~ z => y } || NumParser |
|
134 */ |
|
135 |
|
136 println(E.parse_all("1+3+4")) |
|
137 println(E.parse("1+3+4")) |
|
138 println(E.parse_all("4*2+3")) |
|
139 println(E.parse_all("4*(2+3)")) |
|
140 println(E.parse_all("(4)*((2+3))")) |
|
141 println(E.parse_all("4/2+3")) |
|
142 println(E.parse("1 + 2 * 3")) |
|
143 println(E.parse_all("(1+2)+3")) |
|
144 println(E.parse_all("1+2+3")) |
|
145 |
|
146 |
|
147 // with parser combinators (and other parsing algorithms) |
|
148 // no left-recursion is allowed, otherwise the will loop |
|
149 |
|
150 lazy val EL: Parser[String, Int] = |
|
151 ((EL ~ p"+" ~ EL).map{ case ((x, y), z) => x + z} || |
|
152 (EL ~ p"*" ~ EL).map{ case ((x, y), z) => x * z} || |
|
153 (p"(" ~ EL ~ p")").map{ case ((x, y), z) => y} || |
|
154 NumParserInt) |
|
155 |
|
156 // this will run forever: |
|
157 //println(EL.parse_all("1+2+3")) |
|
158 |
|
159 |
|
160 // non-ambiguous vs ambiguous grammars |
|
161 |
|
162 // ambiguous |
|
163 lazy val S : Parser[String, String] = |
|
164 (p"1" ~ S ~ S).map{ case ((x, y), z) => x + y + z } || p"" |
|
165 |
|
166 println(time(S.parse("1" * 10))) |
|
167 println(time(S.parse_all("1" * 10))) |
|
168 |
|
169 // non-ambiguous |
|
170 lazy val U : Parser[String, String] = |
|
171 (p"1" ~ U).map{ case (x, y) => x + y } || p"" |
|
172 |
|
173 println(time(U.parse("1" * 10))) |
|
174 println(time(U.parse_all("1" * 10))) |
|
175 println(U.parse("1" * 25)) |
|
176 |
|
177 U.parse("11") |
|
178 U.parse("11111") |
|
179 U.parse("11011") |
|
180 |
|
181 U.parse_all("1" * 100) |
|
182 U.parse_all("1" * 100 + "0") |
|
183 |
|
184 // you can see the difference in second example |
|
185 //S.parse_all("1" * 100) // succeeds |
|
186 //S.parse_all("1" * 100 + "0") // fails |
|
187 |
|
188 |
|
189 // A variant which counts how many 1s are parsed |
|
190 lazy val UCount : Parser[String, Int] = |
|
191 (p"1" ~ UCount).map[Int]{ case (_, y) => y + 1 } || p"".map[Int]{ _ => 0 } |
|
192 |
|
193 println(UCount.parse("11111")) |
|
194 println(UCount.parse_all("11111")) |
|
195 |
|
196 // Two single character parsers |
|
197 lazy val One : Parser[String, String] = p"a" |
|
198 lazy val Two : Parser[String, String] = p"b" |
|
199 |
|
200 One.parse("a") |
|
201 One.parse("aaa") |
|
202 |
|
203 // note how the pairs nest to the left with sequence parsers |
|
204 (One ~ One).parse("aaa") |
|
205 (One ~ One ~ One).parse("aaa") |
|
206 (One ~ One ~ One ~ One).parse("aaaa") |
|
207 |
|
208 (One || Two).parse("aaa") |
|
209 |
|
210 |
|
211 |
|
212 // a problem with the arithmetic expression parser: it |
|
213 // gets vert slow with deeply nested parentheses |
|
214 |
|
215 println("Runtime problem") |
|
216 println(E.parse("1")) |
|
217 println(E.parse("(1)")) |
|
218 println(E.parse("((1))")) |
|
219 //println(E.parse("(((1)))")) |
|
220 //println(E.parse("((((1))))")) |
|
221 //println(E.parse("((((((1))))))")) |
|
222 //println(E.parse("(((((((1)))))))")) |
|
223 //println(E.parse("((((((((1)))))))")) |