progs/tokenise.scala
changeset 754 1c9a23304b85
parent 753 d94fdbef1a4f
child 755 37b69593994c
equal deleted inserted replaced
753:d94fdbef1a4f 754:1c9a23304b85
     1 // A simple tokeniser based on the Sulzmann & Lu algorithm
       
     2 //=========================================================
       
     3 // 
       
     4 // call with 
       
     5 //
       
     6 //     scala tokenise.scala fib.while
       
     7 //
       
     8 //     scala tokenise.scala loops.while
       
     9 //
       
    10 // this will generate a .tks file that can be deserialised back
       
    11 // into a list of tokens
       
    12 // you can add -Xno-patmat-analysis in order to get rid of the
       
    13 // match-not-exhaustive warning
       
    14 
       
    15 object Tokenise {
       
    16 
       
    17 import scala.language.implicitConversions    
       
    18 import scala.language.reflectiveCalls
       
    19 
       
    20 // regular expressions including records
       
    21 abstract class Rexp 
       
    22 case object ZERO extends Rexp
       
    23 case object ONE extends Rexp
       
    24 case class CHAR(c: Char) extends Rexp
       
    25 case class ALT(r1: Rexp, r2: Rexp) extends Rexp 
       
    26 case class SEQ(r1: Rexp, r2: Rexp) extends Rexp 
       
    27 case class STAR(r: Rexp) extends Rexp 
       
    28 case class RECD(x: String, r: Rexp) extends Rexp
       
    29   
       
    30 // values  
       
    31 abstract class Val
       
    32 case object Empty extends Val
       
    33 case class Chr(c: Char) extends Val
       
    34 case class Sequ(v1: Val, v2: Val) extends Val
       
    35 case class Left(v: Val) extends Val
       
    36 case class Right(v: Val) extends Val
       
    37 case class Stars(vs: List[Val]) extends Val
       
    38 case class Rec(x: String, v: Val) extends Val
       
    39    
       
    40 // some convenience for typing in regular expressions
       
    41 def charlist2rexp(s : List[Char]): Rexp = s match {
       
    42   case Nil => ONE
       
    43   case c::Nil => CHAR(c)
       
    44   case c::s => SEQ(CHAR(c), charlist2rexp(s))
       
    45 }
       
    46 implicit def string2rexp(s : String) : Rexp = 
       
    47   charlist2rexp(s.toList)
       
    48 
       
    49 implicit def RexpOps(r: Rexp) = new {
       
    50   def | (s: Rexp) = ALT(r, s)
       
    51   def % = STAR(r)
       
    52   def ~ (s: Rexp) = SEQ(r, s)
       
    53 }
       
    54 
       
    55 implicit def stringOps(s: String) = new {
       
    56   def | (r: Rexp) = ALT(s, r)
       
    57   def | (r: String) = ALT(s, r)
       
    58   def % = STAR(s)
       
    59   def ~ (r: Rexp) = SEQ(s, r)
       
    60   def ~ (r: String) = SEQ(s, r)
       
    61   def $ (r: Rexp) = RECD(s, r)
       
    62 }
       
    63 
       
    64 def nullable(r: Rexp) : Boolean = r match {
       
    65   case ZERO => false
       
    66   case ONE => true
       
    67   case CHAR(_) => false
       
    68   case ALT(r1, r2) => nullable(r1) || nullable(r2)
       
    69   case SEQ(r1, r2) => nullable(r1) && nullable(r2)
       
    70   case STAR(_) => true
       
    71   case RECD(_, r1) => nullable(r1)
       
    72 }
       
    73 
       
    74 def der(c: Char, r: Rexp) : Rexp = r match {
       
    75   case ZERO => ZERO
       
    76   case ONE => ZERO
       
    77   case CHAR(d) => if (c == d) ONE else ZERO
       
    78   case ALT(r1, r2) => ALT(der(c, r1), der(c, r2))
       
    79   case SEQ(r1, r2) => 
       
    80     if (nullable(r1)) ALT(SEQ(der(c, r1), r2), der(c, r2))
       
    81     else SEQ(der(c, r1), r2)
       
    82   case STAR(r) => SEQ(der(c, r), STAR(r))
       
    83   case RECD(_, r1) => der(c, r1)
       
    84 }
       
    85 
       
    86 
       
    87 // extracts a string from value
       
    88 def flatten(v: Val) : String = v match {
       
    89   case Empty => ""
       
    90   case Chr(c) => c.toString
       
    91   case Left(v) => flatten(v)
       
    92   case Right(v) => flatten(v)
       
    93   case Sequ(v1, v2) => flatten(v1) + flatten(v2)
       
    94   case Stars(vs) => vs.map(flatten).mkString
       
    95   case Rec(_, v) => flatten(v)
       
    96 }
       
    97 
       
    98 
       
    99 // extracts an environment from a value;
       
   100 // used for tokenise a string
       
   101 def env(v: Val) : List[(String, String)] = v match {
       
   102   case Empty => Nil
       
   103   case Chr(c) => Nil
       
   104   case Left(v) => env(v)
       
   105   case Right(v) => env(v)
       
   106   case Sequ(v1, v2) => env(v1) ::: env(v2)
       
   107   case Stars(vs) => vs.flatMap(env)
       
   108   case Rec(x, v) => (x, flatten(v))::env(v)
       
   109 }
       
   110 
       
   111 // The Injection Part of the lexer
       
   112 
       
   113 def mkeps(r: Rexp) : Val = r match {
       
   114   case ONE => Empty
       
   115   case ALT(r1, r2) => 
       
   116     if (nullable(r1)) Left(mkeps(r1)) else Right(mkeps(r2))
       
   117   case SEQ(r1, r2) => Sequ(mkeps(r1), mkeps(r2))
       
   118   case STAR(r) => Stars(Nil)
       
   119   case RECD(x, r) => Rec(x, mkeps(r))
       
   120 }
       
   121 
       
   122 def inj(r: Rexp, c: Char, v: Val) : Val = (r, v) match {
       
   123   case (STAR(r), Sequ(v1, Stars(vs))) => Stars(inj(r, c, v1)::vs)
       
   124   case (SEQ(r1, r2), Sequ(v1, v2)) => Sequ(inj(r1, c, v1), v2)
       
   125   case (SEQ(r1, r2), Left(Sequ(v1, v2))) => Sequ(inj(r1, c, v1), v2)
       
   126   case (SEQ(r1, r2), Right(v2)) => Sequ(mkeps(r1), inj(r2, c, v2))
       
   127   case (ALT(r1, r2), Left(v1)) => Left(inj(r1, c, v1))
       
   128   case (ALT(r1, r2), Right(v2)) => Right(inj(r2, c, v2))
       
   129   case (CHAR(d), Empty) => Chr(c) 
       
   130   case (RECD(x, r1), _) => Rec(x, inj(r1, c, v))
       
   131 }
       
   132 
       
   133 // some "rectification" functions for simplification
       
   134 def F_ID(v: Val): Val = v
       
   135 def F_RIGHT(f: Val => Val) = (v:Val) => Right(f(v))
       
   136 def F_LEFT(f: Val => Val) = (v:Val) => Left(f(v))
       
   137 def F_ALT(f1: Val => Val, f2: Val => Val) = (v:Val) => v match {
       
   138   case Right(v) => Right(f2(v))
       
   139   case Left(v) => Left(f1(v))
       
   140 }
       
   141 def F_SEQ(f1: Val => Val, f2: Val => Val) = (v:Val) => v match {
       
   142   case Sequ(v1, v2) => Sequ(f1(v1), f2(v2))
       
   143 }
       
   144 def F_SEQ_Empty1(f1: Val => Val, f2: Val => Val) = 
       
   145   (v:Val) => Sequ(f1(Empty), f2(v))
       
   146 def F_SEQ_Empty2(f1: Val => Val, f2: Val => Val) = 
       
   147   (v:Val) => Sequ(f1(v), f2(Empty))
       
   148 def F_RECD(f: Val => Val) = (v:Val) => v match {
       
   149   case Rec(x, v) => Rec(x, f(v))
       
   150 }
       
   151 def F_ERROR(v: Val): Val = throw new Exception("error")
       
   152 
       
   153 def simp(r: Rexp): (Rexp, Val => Val) = r match {
       
   154   case ALT(r1, r2) => {
       
   155     val (r1s, f1s) = simp(r1)
       
   156     val (r2s, f2s) = simp(r2)
       
   157     (r1s, r2s) match {
       
   158       case (ZERO, _) => (r2s, F_RIGHT(f2s))
       
   159       case (_, ZERO) => (r1s, F_LEFT(f1s))
       
   160       case _ => if (r1s == r2s) (r1s, F_LEFT(f1s))
       
   161                 else (ALT (r1s, r2s), F_ALT(f1s, f2s)) 
       
   162     }
       
   163   }
       
   164   case SEQ(r1, r2) => {
       
   165     val (r1s, f1s) = simp(r1)
       
   166     val (r2s, f2s) = simp(r2)
       
   167     (r1s, r2s) match {
       
   168       case (ZERO, _) => (ZERO, F_ERROR)
       
   169       case (_, ZERO) => (ZERO, F_ERROR)
       
   170       case (ONE, _) => (r2s, F_SEQ_Empty1(f1s, f2s))
       
   171       case (_, ONE) => (r1s, F_SEQ_Empty2(f1s, f2s))
       
   172       case _ => (SEQ(r1s,r2s), F_SEQ(f1s, f2s))
       
   173     }
       
   174   }
       
   175   case r => (r, F_ID)
       
   176 }
       
   177 
       
   178 // lexing functions including simplification
       
   179 def lex_simp(r: Rexp, s: List[Char]) : Val = s match {
       
   180   case Nil => if (nullable(r)) mkeps(r) else 
       
   181     { throw new Exception("lexing error") } 
       
   182   case c::cs => {
       
   183     val (r_simp, f_simp) = simp(der(c, r))
       
   184     inj(r, c, f_simp(lex_simp(r_simp, cs)))
       
   185   }
       
   186 }
       
   187 
       
   188 def lexing_simp(r: Rexp, s: String) = 
       
   189   env(lex_simp(r, s.toList))
       
   190 
       
   191 
       
   192 // The Lexing Rules for the WHILE Language
       
   193 
       
   194 // inefficient representations for some extended regular
       
   195 // expressions
       
   196 def PLUS(r: Rexp) = r ~ r.%
       
   197 
       
   198 def Range(s : List[Char]) : Rexp = s match {
       
   199   case Nil => ZERO
       
   200   case c::Nil => CHAR(c)
       
   201   case c::s => ALT(CHAR(c), Range(s))
       
   202 }
       
   203 def RANGE(s: String) = Range(s.toList)
       
   204 
       
   205 val SYM = RANGE("ABCDEFGHIJKLMNOPQRSTUVXYZabcdefghijklmnopqrstuvwxyz_")
       
   206 val DIGIT = RANGE("0123456789")
       
   207 val ID = SYM ~ (SYM | DIGIT).% 
       
   208 val NUM = PLUS(DIGIT)
       
   209 val KEYWORD : Rexp = "skip" | "while" | "do" | "if" | "then" | "else" | "read" | "write" 
       
   210 val SEMI: Rexp = ";"
       
   211 val OP: Rexp = ":=" | "=" | "-" | "+" | "*" | "!=" | "<" | ">"
       
   212 val WHITESPACE = PLUS(" " | "\n" | "\t")
       
   213 val RPAREN: Rexp = "{"
       
   214 val LPAREN: Rexp = "}"
       
   215 val STRING: Rexp = "\"" ~ SYM.% ~ "\""
       
   216 
       
   217 
       
   218 val WHILE_REGS = (("k" $ KEYWORD) | 
       
   219                   ("i" $ ID) | 
       
   220                   ("o" $ OP) | 
       
   221                   ("n" $ NUM) | 
       
   222                   ("s" $ SEMI) | 
       
   223                   ("str" $ STRING) |
       
   224                   ("p" $ (LPAREN | RPAREN)) | 
       
   225                   ("w" $ WHITESPACE)).%
       
   226 
       
   227 
       
   228 
       
   229 // Generate tokens for the WHILE language
       
   230 // and serialise them into a .tks file
       
   231 
       
   232 import java.io._
       
   233 
       
   234 abstract class Token extends Serializable 
       
   235 case object T_SEMI extends Token
       
   236 case object T_LPAREN extends Token
       
   237 case object T_RPAREN extends Token
       
   238 case class T_ID(s: String) extends Token
       
   239 case class T_OP(s: String) extends Token
       
   240 case class T_NUM(n: Int) extends Token
       
   241 case class T_KWD(s: String) extends Token
       
   242 case class T_STR(s: String) extends Token
       
   243 
       
   244 // transforms pairs into tokens
       
   245 val token : PartialFunction[(String, String), Token] = {
       
   246   case ("s", _) => T_SEMI
       
   247   case ("p", "{") => T_LPAREN
       
   248   case ("p", "}") => T_RPAREN
       
   249   case ("i", s) => T_ID(s)
       
   250   case ("o", s) => T_OP(s)
       
   251   case ("n", s) => T_NUM(s.toInt)
       
   252   case ("k", s) => T_KWD(s)
       
   253   case ("str", s) => T_STR(s)
       
   254 }
       
   255 
       
   256 // filters out all un-interesting tokens
       
   257 def tokenise(s: String) : List[Token] = 
       
   258   lexing_simp(WHILE_REGS, s).collect(token)
       
   259 
       
   260 
       
   261 def serialise[T](fname: String, data: T) = {
       
   262   import scala.util.Using
       
   263   Using(new ObjectOutputStream(new FileOutputStream(fname))) {
       
   264     out => out.writeObject(data)
       
   265   }
       
   266 }
       
   267 
       
   268 def main(args: Array[String]) : Unit = {
       
   269   val fname = args(0)
       
   270   val tname = fname.stripSuffix(".while") ++ ".tks"
       
   271   val file = io.Source.fromFile(fname).mkString
       
   272   serialise(tname, tokenise(file))
       
   273 }
       
   274 
       
   275 
       
   276 }