| author | Christian Urban <christian.urban@kcl.ac.uk> | 
| Sat, 04 Dec 2021 00:41:31 +0000 | |
| changeset 859 | 2c6fa627df78 | 
| parent 742 | 155426396b5f | 
| permissions | -rw-r--r-- | 
| 623 | 1  | 
// A Version with an explicit n-times regular expression;  | 
| 513 | 2  | 
// this keeps the size of the regular expression in the  | 
3  | 
// EVIL1 test-case quite small  | 
|
| 422 | 4  | 
|
| 
120
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
5  | 
abstract class Rexp  | 
| 
424
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
6  | 
case object ZERO extends Rexp  | 
| 
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
7  | 
case object ONE extends Rexp  | 
| 49 | 8  | 
case class CHAR(c: Char) extends Rexp  | 
| 
120
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
9  | 
case class ALT(r1: Rexp, r2: Rexp) extends Rexp  | 
| 
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
10  | 
case class SEQ(r1: Rexp, r2: Rexp) extends Rexp  | 
| 
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
11  | 
case class STAR(r: Rexp) extends Rexp  | 
| 477 | 12  | 
case class NTIMES(r: Rexp, n: Int) extends Rexp //explicit constructor for n-times  | 
13  | 
||
| 49 | 14  | 
|
15  | 
def nullable (r: Rexp) : Boolean = r match {
 | 
|
| 
424
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
16  | 
case ZERO => false  | 
| 
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
17  | 
case ONE => true  | 
| 49 | 18  | 
case CHAR(_) => false  | 
19  | 
case ALT(r1, r2) => nullable(r1) || nullable(r2)  | 
|
20  | 
case SEQ(r1, r2) => nullable(r1) && nullable(r2)  | 
|
21  | 
case STAR(_) => true  | 
|
| 
120
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
22  | 
case NTIMES(r, i) => if (i == 0) true else nullable(r)  | 
| 49 | 23  | 
}  | 
24  | 
||
| 477 | 25  | 
|
| 49 | 26  | 
def der (c: Char, r: Rexp) : Rexp = r match {
 | 
| 
424
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
27  | 
case ZERO => ZERO  | 
| 
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
28  | 
case ONE => ZERO  | 
| 
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
29  | 
case CHAR(d) => if (c == d) ONE else ZERO  | 
| 49 | 30  | 
case ALT(r1, r2) => ALT(der(c, r1), der(c, r2))  | 
31  | 
case SEQ(r1, r2) =>  | 
|
32  | 
if (nullable(r1)) ALT(SEQ(der(c, r1), r2), der(c, r2))  | 
|
33  | 
else SEQ(der(c, r1), r2)  | 
|
| 
440
 
e14cd32ad497
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
434 
diff
changeset
 | 
34  | 
case STAR(r1) => SEQ(der(c, r1), STAR(r1))  | 
| 
 
e14cd32ad497
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
434 
diff
changeset
 | 
35  | 
case NTIMES(r1, i) =>  | 
| 477 | 36  | 
if (i == 0) ZERO else SEQ(der(c, r1), NTIMES(r1, i - 1))  | 
| 49 | 37  | 
}  | 
38  | 
||
39  | 
def ders (s: List[Char], r: Rexp) : Rexp = s match {
 | 
|
40  | 
case Nil => r  | 
|
| 
120
 
3e71efb25ce9
added
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
93 
diff
changeset
 | 
41  | 
case c::s => ders(s, der(c, r))  | 
| 49 | 42  | 
}  | 
43  | 
||
| 631 | 44  | 
def matcher(r: Rexp, s: String) : Boolean = nullable(ders(s.toList, r))  | 
| 49 | 45  | 
|
46  | 
||
| 623 | 47  | 
// the optional regular expression: one or zero times  | 
48  | 
// this regular expression is still defined in terms of ALT  | 
|
| 
424
 
1129024b26d5
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
422 
diff
changeset
 | 
49  | 
def OPT(r: Rexp) = ALT(r, ONE)  | 
| 49 | 50  | 
|
| 477 | 51  | 
|
52  | 
// Test Cases  | 
|
53  | 
||
| 623 | 54  | 
// evil regular expressions  | 
| 414 | 55  | 
def EVIL1(n: Int) = SEQ(NTIMES(OPT(CHAR('a')), n), NTIMES(CHAR('a'), n))
 | 
| 
440
 
e14cd32ad497
updated
 
Christian Urban <christian dot urban at kcl dot ac dot uk> 
parents: 
434 
diff
changeset
 | 
56  | 
val EVIL2 = SEQ(STAR(STAR(CHAR('a'))), CHAR('b'))
 | 
| 49 | 57  | 
|
58  | 
def time_needed[T](i: Int, code: => T) = {
 | 
|
59  | 
val start = System.nanoTime()  | 
|
60  | 
for (j <- 1 to i) code  | 
|
61  | 
val end = System.nanoTime()  | 
|
| 631 | 62  | 
(end - start) / (i * 1.0e9)  | 
| 414 | 63  | 
}  | 
64  | 
||
65  | 
||
| 477 | 66  | 
|
| 623 | 67  | 
// test: (a?{n}) (a{n})
 | 
68  | 
for (i <- 0 to 1000 by 100) {
 | 
|
| 631 | 69  | 
  println(f"$i: ${time_needed(2, matcher(EVIL1(i), "a" * i))}%.5f")
 | 
| 623 | 70  | 
}  | 
71  | 
||
72  | 
// test: (a*)* b  | 
|
| 631 | 73  | 
for (i <- 0 to 20) {
 | 
74  | 
  println(f"$i: ${time_needed(2, matcher(EVIL2, "a" * i))}%.5f")
 | 
|
| 623 | 75  | 
}  | 
76  | 
||
77  | 
||
78  | 
// the size of a regular expressions - for testing purposes  | 
|
| 477 | 79  | 
def size(r: Rexp) : Int = r match {
 | 
80  | 
case ZERO => 1  | 
|
81  | 
case ONE => 1  | 
|
82  | 
case CHAR(_) => 1  | 
|
83  | 
case ALT(r1, r2) => 1 + size(r1) + size(r2)  | 
|
84  | 
case SEQ(r1, r2) => 1 + size(r1) + size(r2)  | 
|
85  | 
case STAR(r) => 1 + size(r)  | 
|
86  | 
case NTIMES(r, _) => 1 + size(r)  | 
|
87  | 
}  | 
|
88  | 
||
89  | 
// EVIL1(n) has now a constant size, no matter  | 
|
| 564 | 90  | 
// what n is; also the derivative only grows  | 
91  | 
// moderately  | 
|
| 477 | 92  | 
|
93  | 
size(EVIL1(1)) // 7  | 
|
94  | 
size(EVIL1(3)) // 7  | 
|
95  | 
size(EVIL1(5)) // 7  | 
|
96  | 
size(EVIL1(7)) // 7  | 
|
| 623 | 97  | 
size(EVIL1(20)) // 7  | 
| 477 | 98  | 
|
| 564 | 99  | 
size(ders("".toList, EVIL1(5)))       // 7
 | 
100  | 
size(ders("a".toList, EVIL1(5)))      // 16
 | 
|
101  | 
size(ders("aa".toList, EVIL1(5)))     // 35
 | 
|
102  | 
size(ders("aaa".toList, EVIL1(5)))    // 59
 | 
|
103  | 
size(ders("aaaa".toList, EVIL1(5)))   // 88
 | 
|
104  | 
size(ders("aaaaa".toList, EVIL1(5)))  // 122
 | 
|
105  | 
size(ders("aaaaaa".toList, EVIL1(5))) // 151
 | 
|
| 477 | 106  | 
|
| 478 | 107  | 
// but the size of the derivatives can still grow  | 
| 564 | 108  | 
// quite dramatically in case of EVIL2  | 
| 477 | 109  | 
|
| 564 | 110  | 
size(ders("".toList, EVIL2))       // 5
 | 
111  | 
size(ders("a".toList, EVIL2))      // 12
 | 
|
112  | 
size(ders("aa".toList, EVIL2))     // 28
 | 
|
113  | 
size(ders("aaa".toList, EVIL2))    // 58
 | 
|
114  | 
size(ders("aaaa".toList, EVIL2))   // 116
 | 
|
115  | 
size(ders("aaaaa".toList, EVIL2))  // 230
 | 
|
116  | 
size(ders("aaaaaa".toList, EVIL2)) // 456
 | 
|
| 566 | 117  | 
|
118  | 
size(ders(("a" * 20).toList, EVIL2)) // 7340068 |