| author | Christian Urban <urbanc@in.tum.de> | 
| Sun, 23 Sep 2018 09:04:24 +0100 | |
| changeset 557 | 6d0e8b6f4243 | 
| parent 556 | 4b0fffaef849 | 
| child 564 | 638663141103 | 
| permissions | -rw-r--r-- | 
| 477 | 1 | // Version with simplification during derivatives; | 
| 2 | // this keeps the regular expressions small, which | |
| 3 | // is good for run-time | |
| 517 | 4 | |
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 5 | |
| 422 | 6 | abstract class Rexp | 
| 407 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 7 | case object ZERO extends Rexp | 
| 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 8 | case object ONE extends Rexp | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 9 | case class CHAR(c: Char) extends Rexp | 
| 422 | 10 | case class ALT(r1: Rexp, r2: Rexp) extends Rexp | 
| 11 | case class SEQ(r1: Rexp, r2: Rexp) extends Rexp | |
| 12 | case class STAR(r: Rexp) extends Rexp | |
| 13 | case class NTIMES(r: Rexp, n: Int) extends Rexp | |
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 14 | |
| 467 
3fc9b036321d
fixed bug
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
458diff
changeset | 15 | |
| 550 | 16 | // string of a regular expressions - for testing purposes | 
| 17 | def string(r: Rexp) : String = r match {
 | |
| 18 | case ZERO => "0" | |
| 19 | case ONE => "1" | |
| 20 | case CHAR(c) => c.toString | |
| 21 |   case ALT(r1, r2) => s"(${string(r1)} + ${string(r2)})"
 | |
| 22 |   case SEQ(CHAR(c), CHAR(d)) => s"${c}${d}"
 | |
| 23 |   case SEQ(ONE, CHAR(c)) => s"1${c}"
 | |
| 24 |   case SEQ(r1, r2) => s"(${string(r1)} ~ ${string(r2)})"
 | |
| 25 |   case STAR(r) => s"(${string(r)})*"
 | |
| 26 |   case NTIMES(r, n) =>  s"(${string(r)}){${n}}"
 | |
| 27 | } | |
| 28 | ||
| 29 | // size of a regular expressions - for testing purposes | |
| 30 | def size(r: Rexp) : Int = r match {
 | |
| 31 | case ZERO => 1 | |
| 32 | case ONE => 1 | |
| 33 | case CHAR(_) => 1 | |
| 34 | case ALT(r1, r2) => 1 + size(r1) + size(r2) | |
| 35 | case SEQ(r1, r2) => 1 + size(r1) + size(r2) | |
| 36 | case STAR(r) => 1 + size(r) | |
| 37 | case NTIMES(r, _) => 1 + size(r) | |
| 38 | } | |
| 39 | ||
| 40 | ||
| 41 | ||
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 42 | // nullable function: tests whether the regular | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 43 | // expression can recognise the empty string | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 44 | def nullable (r: Rexp) : Boolean = r match {
 | 
| 407 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 45 | case ZERO => false | 
| 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 46 | case ONE => true | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 47 | case CHAR(_) => false | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 48 | case ALT(r1, r2) => nullable(r1) || nullable(r2) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 49 | case SEQ(r1, r2) => nullable(r1) && nullable(r2) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 50 | case STAR(_) => true | 
| 121 
43c116860e47
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
93diff
changeset | 51 | case NTIMES(r, i) => if (i == 0) true else nullable(r) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 52 | } | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 53 | |
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 54 | // derivative of a regular expression w.r.t. a character | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 55 | def der (c: Char, r: Rexp) : Rexp = r match {
 | 
| 407 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 56 | case ZERO => ZERO | 
| 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 57 | case ONE => ZERO | 
| 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 58 | case CHAR(d) => if (c == d) ONE else ZERO | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 59 | case ALT(r1, r2) => ALT(der(c, r1), der(c, r2)) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 60 | case SEQ(r1, r2) => | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 61 | if (nullable(r1)) ALT(SEQ(der(c, r1), r2), der(c, r2)) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 62 | else SEQ(der(c, r1), r2) | 
| 440 
e14cd32ad497
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
434diff
changeset | 63 | case STAR(r1) => SEQ(der(c, r1), STAR(r1)) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 64 | case NTIMES(r, i) => | 
| 477 | 65 | if (i == 0) ZERO else SEQ(der(c, r), NTIMES(r, i - 1)) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 66 | } | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 67 | |
| 550 | 68 | |
| 69 | ||
| 70 | def simp(r: Rexp, seen: Set[Rexp]) : (Rexp, Set[Rexp]) = r match {
 | |
| 71 |   case ALT(r1, r2) => {
 | |
| 72 | val (r1s, seen1) = simp(r1, seen) | |
| 73 | val (r2s, seen2) = simp(r2, seen1) | |
| 74 |     (r1s, r2s) match {
 | |
| 75 | case (ZERO, r2s) => (r2s, seen2) | |
| 76 | case (r1s, ZERO) => (r1s, seen2) | |
| 77 | case (r1s, r2s) => (ALT(r1s, r2s), seen2) | |
| 78 | } | |
| 422 | 79 | } | 
| 550 | 80 |   case SEQ(r1, r2) =>  {
 | 
| 81 | val (r1s, _) = simp(r1, Set()) | |
| 82 | val (r2s, _) = simp(r2, Set()) | |
| 83 | if (seen.contains(SEQ(r1s, r2s))) (ZERO, seen) | |
| 84 |     else (r1s, r2s) match {
 | |
| 85 | case (ZERO, _) => (ZERO, seen) | |
| 86 | case (_, ZERO) => (ZERO, seen) | |
| 87 | case (ONE, r2s) => (r2s, seen + r2s) | |
| 88 | case (r1s, ONE) => (r1s, seen + r1s) | |
| 89 | case (r1s, r2s) => (SEQ(r1s, r2s), seen + SEQ(r1s, r2s)) | |
| 90 | } | |
| 422 | 91 | } | 
| 550 | 92 | case r => if (seen.contains(r)) (ZERO, seen) else (r, seen + r) | 
| 422 | 93 | } | 
| 94 | ||
| 95 | ||
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 96 | // derivative w.r.t. a string (iterates der) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 97 | def ders (s: List[Char], r: Rexp) : Rexp = s match {
 | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 98 | case Nil => r | 
| 550 | 99 | case c::s => ders(s, simp(der(c, r), Set())._1) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 100 | } | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 101 | |
| 407 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 102 | |
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 103 | // main matcher function | 
| 414 | 104 | def matcher(r: Rexp, s: String) : Boolean = nullable(ders(s.toList, r)) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 105 | |
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 106 | |
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 107 | //one or zero | 
| 407 
4b454a6d1814
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
121diff
changeset | 108 | def OPT(r: Rexp) = ALT(r, ONE) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 109 | |
| 477 | 110 | |
| 550 | 111 | |
| 112 | ||
| 113 | ||
| 114 | ||
| 115 | ||
| 116 | ||
| 117 | ||
| 118 | def time_needed[T](i: Int, code: => T) = {
 | |
| 119 | val start = System.nanoTime() | |
| 120 | for (j <- 1 to i) code | |
| 121 | val end = System.nanoTime() | |
| 122 | (end - start)/(i * 1.0e9) | |
| 123 | } | |
| 124 | ||
| 125 | ||
| 126 | // star example | |
| 127 | val Tstar = STAR(STAR(STAR(CHAR('a'))))
 | |
| 128 | ||
| 129 | string(ders("".toList, Tstar))
 | |
| 130 | size(ders("".toList, Tstar))      // 4
 | |
| 131 | string(ders("a".toList, Tstar))
 | |
| 132 | size(ders("a".toList, Tstar))     // 11
 | |
| 133 | string(ders("aa".toList, Tstar))
 | |
| 134 | size(ders("aa".toList, Tstar))    // 11
 | |
| 135 | string(ders("aaa".toList, Tstar))
 | |
| 136 | size(ders("aaa".toList, Tstar))   // 11
 | |
| 137 | string(ders("aaaa".toList, Tstar))
 | |
| 138 | size(ders("aaaa".toList, Tstar))  // 11
 | |
| 139 | string(ders("aaaa".toList, Tstar))
 | |
| 140 | size(ders("aaaaa".toList, Tstar)) // 11
 | |
| 141 | string(ders("aaaab".toList, Tstar))
 | |
| 142 | size(ders("aaaaab".toList, Tstar)) // 1
 | |
| 143 | ||
| 144 | // test: ("a" | "aa")*
 | |
| 145 | val EVIL3 = STAR(ALT(CHAR('a'), SEQ(CHAR('a'), CHAR('a'))))
 | |
| 146 | ||
| 147 | for (i <- 1 to 100 by 1) {
 | |
| 148 | println(i + " " + "%.5f".format(time_needed(2, matcher(EVIL3, "a" * i))) + | |
| 149 | 	  " size: " + size(ders(("a" * i).toList, EVIL3)))
 | |
| 150 | } | |
| 151 | ||
| 152 | ||
| 153 | println("start " + string(EVIL3) + "    " + size(EVIL3))
 | |
| 154 | val t1  = der('a', EVIL3)
 | |
| 155 | println(string(t1) + " " + size(t1)) | |
| 156 | val t1s = simp(t1, Set())._1 | |
| 157 | println("simplified " + string(t1s) + "    " + size(t1s))
 | |
| 158 | val t2  = der('a', t1s)
 | |
| 159 | println(string(t2) + " " + size(t2)) | |
| 160 | val t2s = simp(t2, Set())._1 | |
| 161 | println("simplified " + string(t2s) + "    " + size(t2s))
 | |
| 162 | val t3  = der('a', t2s)
 | |
| 163 | println(string(t3) + " " + size(t3)) | |
| 164 | val t3s = simp(t3, Set())._1 | |
| 165 | println("simplified " + string(t3s) + "    " + size(t3s))
 | |
| 166 | val t4  = der('a', t3s)
 | |
| 167 | val t4s = simp(t4, Set())._1 | |
| 168 | ||
| 169 | ||
| 170 | ||
| 171 | ||
| 172 | ||
| 173 | ||
| 174 | ||
| 175 | ||
| 176 | println(string(t4) + " " + size(t4)) | |
| 177 | println("simplified " + string(t4s) + "    " + size(t4s))
 | |
| 178 | ||
| 179 | ||
| 477 | 180 | // Test Cases | 
| 181 | ||
| 544 | 182 | //evil regular expressions: (a?){n} a{n}  and (a*)* b
 | 
| 183 | 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: 
434diff
changeset | 184 | val EVIL2 = SEQ(STAR(STAR(CHAR('a'))), CHAR('b'))
 | 
| 422 | 185 | |
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 186 | |
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 187 | def time_needed[T](i: Int, code: => T) = {
 | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 188 | val start = System.nanoTime() | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 189 | for (j <- 1 to i) code | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 190 | val end = System.nanoTime() | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 191 | (end - start)/(i * 1.0e9) | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 192 | } | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 193 | |
| 434 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 194 | |
| 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 195 | //test: (a?{n}) (a{n})
 | 
| 519 | 196 | for (i <- 1 to 8001 by 1000) {
 | 
| 434 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 197 | println(i + " " + "%.5f".format(time_needed(2, matcher(EVIL1(i), "a" * i)))) | 
| 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 198 | } | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 199 | |
| 519 | 200 | for (i <- 1 to 8001 by 1000) {
 | 
| 415 | 201 | println(i + " " + "%.5f".format(time_needed(2, matcher(EVIL1(i), "a" * i)))) | 
| 92 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 202 | } | 
| 
e85600529ca5
moved scala files
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: diff
changeset | 203 | |
| 434 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 204 | //test: (a*)* b | 
| 477 | 205 | for (i <- 1 to 6000001 by 500000) {
 | 
| 206 | println(i + " " + "%.5f".format(time_needed(2, matcher(EVIL2, "a" * i)))) | |
| 207 | } | |
| 208 | ||
| 209 | for (i <- 1 to 6000001 by 500000) {
 | |
| 434 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 210 | println(i + " " + "%.5f".format(time_needed(2, matcher(EVIL2, "a" * i)))) | 
| 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 211 | } | 
| 
8664ff87cd77
updated
 Christian Urban <christian dot urban at kcl dot ac dot uk> parents: 
422diff
changeset | 212 | |
| 477 | 213 | |
| 214 | ||
| 215 | // size of a regular expressions - for testing purposes | |
| 216 | def size(r: Rexp) : Int = r match {
 | |
| 217 | case ZERO => 1 | |
| 218 | case ONE => 1 | |
| 219 | case CHAR(_) => 1 | |
| 220 | case ALT(r1, r2) => 1 + size(r1) + size(r2) | |
| 221 | case SEQ(r1, r2) => 1 + size(r1) + size(r2) | |
| 222 | case STAR(r) => 1 + size(r) | |
| 223 | case NTIMES(r, _) => 1 + size(r) | |
| 415 | 224 | } | 
| 225 | ||
| 477 | 226 | |
| 227 | // now the size of the derivatives grows | |
| 228 | // much, much slower | |
| 229 | ||
| 230 | size(ders("".toList, EVIL2))      // 5
 | |
| 231 | size(ders("a".toList, EVIL2))     // 8
 | |
| 232 | size(ders("aa".toList, EVIL2))    // 8
 | |
| 233 | size(ders("aaa".toList, EVIL2))   // 8
 | |
| 234 | size(ders("aaaa".toList, EVIL2))  // 8
 | |
| 235 | size(ders("aaaaa".toList, EVIL2)) // 8
 | |
| 544 | 236 | |
| 237 | ||
| 238 | ||
| 239 | ||
| 240 | ||
| 241 | // Examples from the Sulzmann paper | |
| 242 | val sulzmann = STAR(ALT(ALT(CHAR('a'), CHAR('b')), SEQ(CHAR('a'), CHAR('b'))))
 | |
| 243 | ||
| 244 | ||
| 245 | for (i <- 1 to 4501 by 500) {
 | |
| 246 | println(i + ": " + "%.5f".format(time_needed(1, matcher(sulzmann, "a" * i)))) | |
| 247 | } | |
| 248 | ||
| 249 | for (i <- 1 to 4501 by 500) {
 | |
| 250 | println(i + ": " + "%.5f".format(time_needed(1, matcher(sulzmann, "ab" * i)))) | |
| 251 | } | |
| 550 | 252 | |
| 556 | 253 | size(ders("".toList, EVIL2))      // 5
 | 
| 254 | size(ders("a".toList, EVIL2))     // 8
 | |
| 255 | size(ders("aa".toList, EVIL2))    // 8
 | |
| 256 | size(ders("aaa".toList, EVIL2))   // 8
 | |
| 257 | size(ders("aaaa".toList, EVIL2))  // 8
 | |
| 258 | size(ders("aaaaa".toList, EVIL2)) // 8
 | |
| 550 | 259 | |
| 260 | ||
| 261 | ||
| 262 | (((1 + 1a) ~ ((a + aa))*) + (((0 + 1) ~ ((a + aa))*) + ((1 + 1a) ~ ((a + aa))*))) |