updated
authorChristian Urban <christian.urban@kcl.ac.uk>
Sun, 17 Sep 2023 19:12:57 +0100
changeset 919 53f08d873e09
parent 918 53e7da9f372a
child 920 7af2eea19646
updated
cws/build.sc
cws/cw01.pdf
cws/cw02.pdf
cws/cw03.pdf
cws/cw04.pdf
cws/cw05.pdf
cws/upload
progs/matcher/re1.sc
progs/matcher/re2.sc
progs/matcher/re3.sc
progs/parser-combinators/comb1.sc
progs/parser-combinators/comb2.sc
solutions/cw1/cw1.scala
solutions/cw1/matcher.sc
solutions/cw2/lexer.sc
solutions/cw3/lexer.sc
solutions/cw3/parser.sc
solutions/cw3/parser2.sc
--- a/cws/build.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/cws/build.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -1,10 +1,10 @@
 #!/usr/bin/env amm
 
 val files = Seq("cw01.tex",
-	         "cw02.tex",
-	         "cw03.tex",
-	         "cw04.tex",
-	         "cw05.tex")
+	        "cw02.tex",
+	        "cw03.tex",
+	        "cw04.tex",
+	        "cw05.tex")
 
 val pdf_files = files.map(s => s.stripSuffix("tex") ++ "pdf")
 
@@ -13,8 +13,8 @@
 def make() = {
   for (f <- files) {
     println(s"Processing $f ...")
-    os.proc("xelatex", f).call(stdout = os.Inherit, stdin = os.Inherit)
-    os.proc("xelatex", f).call(stdout = os.Inherit, stdin = os.Inherit)
+    os.proc("lualatex", f).call(stdout = os.Inherit, stdin = os.Inherit)
+    os.proc("lualatex", f).call(stdout = os.Inherit, stdin = os.Inherit)
   }
 }
 
Binary file cws/cw01.pdf has changed
Binary file cws/cw02.pdf has changed
Binary file cws/cw03.pdf has changed
Binary file cws/cw04.pdf has changed
Binary file cws/cw05.pdf has changed
--- a/cws/upload	Fri Sep 15 10:49:33 2023 +0100
+++ b/cws/upload	Sun Sep 17 19:12:57 2023 +0100
@@ -8,5 +8,10 @@
 done    
 
 
-scp $fls k1192855@bastion:public_html/cfl/cws/
+scp $fls k1192855@bastion:public_html/cfl/cws
+
 
+#hg commit -m "updated jars"
+
+# prevent PDF from being copied
+###gs -o output.pdf -dNoOutputFonts -sDEVICE=pdfwrite main_cw01.pdf 
--- a/progs/matcher/re1.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/progs/matcher/re1.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -1,6 +1,6 @@
 // A simple matcher for basic regular expressions
 //
-// Call the test cases with X = {1,2,3}
+// Call the testcases with X = {1,2,3}
 //
 //   amm re1.sc testX
 //
@@ -19,9 +19,9 @@
 case class SEQ(r1: Rexp, r2: Rexp) extends Rexp  // sequence
 case class STAR(r: Rexp) extends Rexp            // star
 
+
 // nullable function: tests whether a regular 
-// expression can recognise the empty string
-  
+// expression can recognise the empty string  
 def nullable(r: Rexp) : Boolean = r match {
   case ZERO => false
   case ONE => true
@@ -54,13 +54,12 @@
   nullable(ders(s.toList, r))
 
 
+// some examples from the homework
 
 val r = SEQ(CHAR('b'), CHAR('c'))
 matcher(r, "ac")
 
-// some examples from the homework
-
-val r = STAR(ALT(SEQ(CHAR('a'), CHAR('b')), CHAR('b')))
+val r1 = STAR(ALT(SEQ(CHAR('a'), CHAR('b')), CHAR('b')))
 der('a', r)
 der('b', r)
 der('c', r)
@@ -71,6 +70,9 @@
 der('z', der('y', der('x', r2)))
 
 
+// Test Cases
+//============
+
 // the optional regular expression (one or zero times)
 def OPT(r: Rexp) = ALT(r, ONE)   // r + 1
 
@@ -81,10 +83,6 @@
   case n => SEQ(r, NTIMES(r, n - 1))
 }
 
-
-// Test Cases
-//============
-
 // the evil regular expression  (a?){n} a{n}
 def EVIL1(n: Int) = 
   SEQ(NTIMES(OPT(CHAR('a')), n), NTIMES(CHAR('a'), n))
--- a/progs/matcher/re2.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/progs/matcher/re2.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -1,6 +1,8 @@
-// A Version with an explicit n-times regular expression;
+// A Version of the matcher with an explicit 
+// n-times regular expression
+//
 // this keeps the size of the regular expression in the
-// EVIL1 test-case quite small
+// EVIL1 testcase small
 //
 // call the test cases with X = {1,2}
 //
@@ -31,7 +33,6 @@
   case NTIMES(r, n) => if (n == 0) true else nullable(r)
 }
 
-
 def der(c: Char, r: Rexp) : Rexp = r match {
   case ZERO => ZERO
   case ONE => ZERO
@@ -54,13 +55,13 @@
   nullable(ders(s.toList, r))
 
 
+// Test Cases
+
 // the optional regular expression: one or zero times
 // this regular expression is still defined in terms of ALT
 def OPT(r: Rexp) = ALT(r, ONE)
 
 
-// Test Cases
-
 // evil regular expressions
 def EVIL1(n: Int) = SEQ(NTIMES(OPT(CHAR('a')), n), NTIMES(CHAR('a'), n))
 val EVIL2 = SEQ(STAR(STAR(CHAR('a'))), CHAR('b'))
@@ -90,7 +91,7 @@
 def test2() = {
   println("Test (a*)* b")
 
-  for (i <- 0 to 30 by 2) {
+  for (i <- 0 to 22 by 2) {
     println(f"$i: ${time_needed(1, matcher(EVIL2, "a" * i))}%.5f")
   }
 }
--- a/progs/matcher/re3.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/progs/matcher/re3.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -1,4 +1,6 @@
-// A version with simplification of derivatives;
+// A version of the matcher with simplification 
+// of derivatives
+//
 // this keeps the regular expressions small, which
 // is good for the run-time
 // 
@@ -48,6 +50,7 @@
     if (i == 0) ZERO else SEQ(der(c, r), NTIMES(r, i - 1))
 }
 
+// simplification
 def simp(r: Rexp) : Rexp = r match {
   case ALT(r1, r2) => (simp(r1), simp(r2)) match {
     case (ZERO, r2s) => r2s
@@ -64,26 +67,23 @@
   case r => r
 }
 
-
-
 // the derivative w.r.t. a string (iterates der)
 def ders(s: List[Char], r: Rexp) : Rexp = s match {
   case Nil => r
   case c::s => ders(s, simp(der(c, r)))
 }
 
-
 // the main matcher function
 def matcher(r: Rexp, s: String) : Boolean = 
   nullable(ders(s.toList, r))
 
 
+// Test Cases
+//============
+
 // one or zero
 def OPT(r: Rexp) = ALT(r, ONE)
 
-
-// Test Cases
-
 // evil regular expressions: (a?){n} a{n}  and (a*)* b
 def EVIL1(n: Int) = SEQ(NTIMES(OPT(CHAR('a')), n), NTIMES(CHAR('a'), n))
 val EVIL2 = SEQ(STAR(STAR(CHAR('a'))), CHAR('b'))
--- a/progs/parser-combinators/comb1.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/progs/parser-combinators/comb1.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -89,12 +89,12 @@
 extension [I, T](p: Parser[I, T])(using I => Seq[_]) {
   def ||(q : => Parser[I, T]) = new AltParser[I, T](p, q)
   def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q)
-  def mapp[S](f: => T => S) = new MapParser[I, T, S](p, f)
+  def map[S](f: => T => S) = new MapParser[I, T, S](p, f)
 }
 
 def toU(s: String) = s.map(_.toUpper)
 
-(p"ELSE").mapp(toU(_)).parse("ELSEifthen")  
+(p"ELSE").map(toU(_)).parse("ELSEifthen")  
 
 // these implicits allow us to use an infix notation for
 // sequences and alternatives; we also can write the usual
@@ -109,8 +109,8 @@
 
 // A parser for palindromes (just returns them as string)
 lazy val Pal : Parser[String, String] = {
-   (p"a" ~ Pal ~ p"a").mapp{ case ((x, y), z) => s"$x$y$z" } || 
-   (p"b" ~ Pal ~ p"b").mapp{ case ((x, y), z) => s"$x$y$z" } || 
+   (p"a" ~ Pal ~ p"a").map{ case ((x, y), z) => s"$x$y$z" } || 
+   (p"b" ~ Pal ~ p"b").map{ case ((x, y), z) => s"$x$y$z" } || 
     p"a" || p"b" || p""
 }  
 
@@ -126,7 +126,7 @@
 //
 //   (transforms '(' -> '{' , ')' -> '}' )
 lazy val P : Parser[String, String] = {
-  (p"(" ~ P ~ p")" ~ P).mapp{ case (((_, x), _), y) => "{" + x + "}" + y } ||
+  (p"(" ~ P ~ p")" ~ P).map{ case (((_, x), _), y) => "{" + x + "}" + y } ||
   p""
 }  
 
@@ -224,32 +224,32 @@
 // a problem with the arithmetic expression parser: it 
 // gets very slow with deeply nested parentheses
 
-println("Runtime problem")
+println("A runtime problem")
 println(E.parse("1"))
 println(E.parse("(1)"))
 println(E.parse("((1))"))
 println(E.parse("(((1)))"))
 println(E.parse("((((1))))"))
-//println(E.parse("((((((1))))))"))
-//println(E.parse("(((((((1)))))))"))
+println(E.parse("((((((1))))))"))
+println(E.parse("(((((((1)))))))"))
 //println(E.parse("((((((((1))))))))"))
 
 
-// faster because of merge
+// faster because of merge in the +/- case
 lazy val E2: Parser[String, Int] = {
-  (T2 ~ (p"+" || p"-") ~ E2).mapp[Int]{ case ((x, y), z) => if (y == "+") x + z else x - z} || T2 }
+  (T2 ~ (p"+" || p"-") ~ E2).map[Int]{ case ((x, y), z) => if (y == "+") x + z else x - z} || T2 }
 lazy val T2: Parser[String, Int] = {
-  (F2 ~ p"*" ~ T2).mapp[Int]{ case ((x, _), z) => x * z } || F2 }
+  (F2 ~ p"*" ~ T2).map[Int]{ case ((x, _), z) => x * z } || F2 }
 lazy val F2: Parser[String, Int] = {
-  (p"(" ~ E2 ~ p")").mapp[Int]{ case ((_, y), _) => y } || NumParserInt }
+  (p"(" ~ E2 ~ p")").map[Int]{ case ((_, y), _) => y } || NumParserInt }
 
 
-println("Runtime problem")
+println("mitigated by merging clauses")
 println(E2.parse("1"))
 println(E2.parse("(1)"))
 println(E2.parse("((1))"))
 println(E2.parse("(((1)))"))
 println(E2.parse("((((1))))"))
-//println(E2.parse("((((((1))))))"))
-//println(E2.parse("(((((((1)))))))"))
-//println(E2.parse("((((((((1))))))))"))
\ No newline at end of file
+println(E2.parse("((((((1))))))"))
+println(E2.parse("(((((((1)))))))"))
+println(E2.parse("((((((((1))))))))"))
\ No newline at end of file
--- a/progs/parser-combinators/comb2.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/progs/parser-combinators/comb2.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -43,7 +43,7 @@
 }
 
 // map parser
-class MapParser[I, T, S](p: => Parser[I, T], 
+class maparser[I, T, S](p: => Parser[I, T], 
                          f: T => S)(using I => Seq[_]) extends Parser[I, S] {
   def parse(in: I) = for ((hd, tl) <- p.parse(in)) yield (f(hd), tl)
 }
@@ -93,7 +93,7 @@
 extension [I, T](p: Parser[I, T])(using I => Seq[_]) {
   def ||(q : => Parser[I, T]) = new AltParser[I, T](p, q)
   def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q)
-  def mapp[S](f: => T => S) = new MapParser[I, T, S](p, f)
+  def map[S](f: => T => S) = new maparser[I, T, S](p, f)
 }
 
 
@@ -125,47 +125,47 @@
 
 // arithmetic expressions
 lazy val AExp: Parser[String, AExp] = 
-  (Te ~ p"+" ~ AExp).mapp[AExp]{ case x ~ _ ~ z => Aop("+", x, z) } ||
-  (Te ~ p"-" ~ AExp).mapp[AExp]{ case x ~ _ ~ z => Aop("-", x, z) } || Te
+  (Te ~ p"+" ~ AExp).map[AExp]{ case x ~ _ ~ z => Aop("+", x, z) } ||
+  (Te ~ p"-" ~ AExp).map[AExp]{ case x ~ _ ~ z => Aop("-", x, z) } || Te
 lazy val Te: Parser[String, AExp] = 
-  (Fa ~ p"*" ~ Te).mapp[AExp]{ case x ~ _ ~ z => Aop("*", x, z) } || 
-  (Fa ~ p"/" ~ Te).mapp[AExp]{ case x ~ _ ~ z => Aop("/", x, z) } || Fa  
+  (Fa ~ p"*" ~ Te).map[AExp]{ case x ~ _ ~ z => Aop("*", x, z) } || 
+  (Fa ~ p"/" ~ Te).map[AExp]{ case x ~ _ ~ z => Aop("/", x, z) } || Fa  
 lazy val Fa: Parser[String, AExp] = 
-   (p"(" ~ AExp ~ p")").mapp{ case _ ~ y ~ _ => y } || 
-   IdParser.mapp(Var) || 
-   NumParser.mapp(Num)
+   (p"(" ~ AExp ~ p")").map{ case _ ~ y ~ _ => y } || 
+   IdParser.map(Var) || 
+   NumParser.map(Num)
 
 // boolean expressions with some simple nesting
 lazy val BExp: Parser[String, BExp] = 
-   (AExp ~ p"==" ~ AExp).mapp[BExp]{ case x ~ _ ~ z => Bop("==", x, z) } || 
-   (AExp ~ p"!=" ~ AExp).mapp[BExp]{ case x ~ _ ~ z => Bop("!=", x, z) } || 
-   (AExp ~ p"<" ~ AExp).mapp[BExp]{ case x ~ _ ~ z => Bop("<", x, z) } || 
-   (AExp ~ p">" ~ AExp).mapp[BExp]{ case x ~ _ ~ z => Bop(">", x, z) } ||
-   (p"(" ~ BExp ~ p")" ~ p"&&" ~ BExp).mapp[BExp]{ case _ ~ y ~ _ ~ _ ~ v => And(y, v) } ||
-   (p"(" ~ BExp ~ p")" ~ p"||" ~ BExp).mapp[BExp]{ case _ ~ y ~ _ ~ _ ~ v => Or(y, v) } ||
-   (p"true".mapp[BExp]{ _ => True }) || 
-   (p"false".mapp[BExp]{ _ => False }) ||
-   (p"(" ~ BExp ~ p")").mapp[BExp]{ case _ ~ x ~ _ => x }
+   (AExp ~ p"==" ~ AExp).map[BExp]{ case x ~ _ ~ z => Bop("==", x, z) } || 
+   (AExp ~ p"!=" ~ AExp).map[BExp]{ case x ~ _ ~ z => Bop("!=", x, z) } || 
+   (AExp ~ p"<" ~ AExp).map[BExp]{ case x ~ _ ~ z => Bop("<", x, z) } || 
+   (AExp ~ p">" ~ AExp).map[BExp]{ case x ~ _ ~ z => Bop(">", x, z) } ||
+   (p"(" ~ BExp ~ p")" ~ p"&&" ~ BExp).map[BExp]{ case _ ~ y ~ _ ~ _ ~ v => And(y, v) } ||
+   (p"(" ~ BExp ~ p")" ~ p"||" ~ BExp).map[BExp]{ case _ ~ y ~ _ ~ _ ~ v => Or(y, v) } ||
+   (p"true".map[BExp]{ _ => True }) || 
+   (p"false".map[BExp]{ _ => False }) ||
+   (p"(" ~ BExp ~ p")").map[BExp]{ case _ ~ x ~ _ => x }
 
 // a single statement 
 lazy val Stmt: Parser[String, Stmt] =
-  ((p"skip".mapp[Stmt]{_ => Skip }) ||
-   (IdParser ~ p":=" ~ AExp).mapp[Stmt]{ case x ~ _ ~ z => Assign(x, z) } ||
-   (p"write(" ~ IdParser ~ p")").mapp[Stmt]{ case _ ~ y ~ _ => Write(y) } ||
+  ((p"skip".map[Stmt]{_ => Skip }) ||
+   (IdParser ~ p":=" ~ AExp).map[Stmt]{ case x ~ _ ~ z => Assign(x, z) } ||
+   (p"write(" ~ IdParser ~ p")").map[Stmt]{ case _ ~ y ~ _ => Write(y) } ||
    (p"if" ~ BExp ~ p"then" ~ Block ~ p"else" ~ Block)
-     .mapp[Stmt]{ case _ ~ y ~ _ ~ u ~ _ ~ w => If(y, u, w) } ||
-   (p"while" ~ BExp ~ p"do" ~ Block).mapp[Stmt]{ case _ ~ y ~ _ ~ w => While(y, w) })   
+     .map[Stmt]{ case _ ~ y ~ _ ~ u ~ _ ~ w => If(y, u, w) } ||
+   (p"while" ~ BExp ~ p"do" ~ Block).map[Stmt]{ case _ ~ y ~ _ ~ w => While(y, w) })   
  
  
 // statements
 lazy val Stmts: Parser[String, Block] =
-  (Stmt ~ p";" ~ Stmts).mapp[Block]{ case x ~ _ ~ z => x :: z } ||
-  (Stmt.mapp[Block]{ s => List(s) })
+  (Stmt ~ p";" ~ Stmts).map[Block]{ case x ~ _ ~ z => x :: z } ||
+  (Stmt.map[Block]{ s => List(s) })
 
 // blocks (enclosed in curly braces)
 lazy val Block: Parser[String, Block] =
-  ((p"{" ~ Stmts ~ p"}").mapp{ case _ ~ y ~ _ => y } || 
-   (Stmt.mapp(s => List(s))))
+  ((p"{" ~ Stmts ~ p"}").map{ case _ ~ y ~ _ => y } || 
+   (Stmt.map(s => List(s))))
 
 
 // Examples
--- a/solutions/cw1/cw1.scala	Fri Sep 15 10:49:33 2023 +0100
+++ b/solutions/cw1/cw1.scala	Sun Sep 17 19:12:57 2023 +0100
@@ -1,6 +1,6 @@
 // CW 1
-import scala.language.implicitConversions    
-import scala.language.reflectiveCalls 
+//import scala.language.implicitConversions    
+//import scala.language.reflectiveCalls 
 
 abstract class Rexp
 case object ZERO extends Rexp                            // matches nothing
--- a/solutions/cw1/matcher.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/solutions/cw1/matcher.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -1,114 +1,17 @@
-// CW1
 
-abstract class Rexp
-case object ZERO extends Rexp
-case object ONE extends Rexp
-case class CHAR(c: Char) extends Rexp
-case class ALT(r1: Rexp, r2: Rexp) extends Rexp 
-case class SEQ(r1: Rexp, r2: Rexp) extends Rexp 
-case class STAR(r: Rexp) extends Rexp 
-
-// extended Rexps
-case class RANGE(s: Set[Char]) extends Rexp
-case class PLUS(r: Rexp) extends Rexp
-case class OPTIONAL(r: Rexp) extends Rexp
-case class NTIMES(r: Rexp, n: Int) extends Rexp 
-case class UPTO(r: Rexp, n: Int) extends Rexp
-case class FROM(r: Rexp, n: Int) extends Rexp
-case class BETWEEN(r: Rexp, n: Int, m: Int) extends Rexp
-case class NOT(r: Rexp) extends Rexp
-
-// functions
-case class CFUN(f: Char => Boolean) extends Rexp
-
-// abbreviations
-def FCHAR(c: Char) = CFUN((a: Char) => a == c)
-def FSET(s: Set[Char]) = CFUN((a: Char) => s.contains(a))
-val FALL = CFUN(_ => true)
-
-def nullable (r: Rexp) : Boolean = r match {
-  case ZERO => false
-  case ONE => true
-  case CHAR(_) => false
-  case ALT(r1, r2) => nullable(r1) || nullable(r2)
-  case SEQ(r1, r2) => nullable(r1) && nullable(r2)
-  case STAR(_) => true
-
-  case RANGE(_) => false 
-  case PLUS(r1) => nullable(r1)
-  case OPTIONAL(_) => true
-  case NTIMES(r1, i) => if (i == 0) true else nullable(r1)
-  case UPTO(_, _) => true
-  case FROM(r1, i) => if (i == 0) true else nullable(r1)
-  case BETWEEN(r1, i, _) => if (i == 0) true else nullable(r1)
-  case NOT(r1) => !nullable(r1)
-
-  case CFUN(f) => false
-}
-
+println("===========================")
 
-def der (c: Char, r: Rexp) : Rexp = r match {
-  case ZERO => ZERO
-  case ONE => ZERO
-  case CHAR(d) => if (c == d) ONE else ZERO
-  case ALT(r1, r2) => ALT(der(c, r1), der(c, r2))
-  case SEQ(r1, r2) => 
-    if (nullable(r1)) ALT(SEQ(der(c, r1), r2), der(c, r2))
-    else SEQ(der(c, r1), r2)
-  case STAR(r1) => SEQ(der(c, r1), STAR(r1))
-
-  case RANGE(s) =>
-    if (s.contains(c)) ONE else ZERO 
-  case PLUS(r1) => SEQ(der(c, r1), STAR(r1))
-  case OPTIONAL(r1) => der(c, r1)
-  case NTIMES(r, i) => 
-    if (i == 0) ZERO else SEQ(der(c, r), NTIMES(r, i - 1))
-  case UPTO(r1, i) => 
-    if (i == 0) ZERO else SEQ(der(c, r1), UPTO(r1, i - 1)) 
-  case FROM(r1, i) =>
-    if (i == 0) SEQ(der(c, r1), FROM(r1, 0)) else
-    SEQ(der(c, r1), FROM(r1, i - 1))
-  case BETWEEN(r1, i, j) => 
-    if (i == 0) {
-      if (j == 0) ZERO else SEQ(der(c, r1), BETWEEN(r1, 0, j - 1))
-    } else SEQ(der(c, r1), BETWEEN(r1, i - 1, j - 1))
-  case NOT(r1) => NOT(der(c, r1))
+//import $file.cw1
+//import cw1._
+import CW1._ 
 
-  case CFUN(f) => if (f(c)) ONE else ZERO
-}
-
-
-def simp(r: Rexp) : Rexp = r match {
-  case ALT(r1, r2) => (simp(r1), simp(r2)) match {
-    case (ZERO, r2s) => r2s
-    case (r1s, ZERO) => r1s
-    case (r1s, r2s) => if (r1s == r2s) r1s else ALT (r1s, r2s)
-  }
-  case SEQ(r1, r2) =>  (simp(r1), simp(r2)) match {
-    case (ZERO, _) => ZERO
-    case (_, ZERO) => ZERO
-    case (ONE, r2s) => r2s
-    case (r1s, ONE) => r1s
-    case (r1s, r2s) => SEQ(r1s, r2s)
-  }
-  case r => r
-}
-
-def ders(s: List[Char], r: Rexp) : Rexp = s match {
-  case Nil => r
-  case c::s => ders(s, simp(der(c, r)))
-}
-
-def matcher(r: Rexp, s: String) : Boolean = nullable(ders(s.toList, r))
-
-
-
+/*
 val Regex31 = NTIMES(CHAR('a'),3)
 val Regex32 = NTIMES(OPTIONAL(CHAR('a')),3)
 val Regex33 = UPTO(CHAR('a'),3)
 val Regex34 = UPTO(OPTIONAL(CHAR('a')),3)
-val Regex35 = BETWEEN(CHAR('a'),3,5)
-val Regex36 = BETWEEN(OPTIONAL(CHAR('a')),3,5)
+val Regex35 = NMTIMES(CHAR('a'),3,5)
+val Regex36 = NMTIMES(OPTIONAL(CHAR('a')),3,5)
 val string31 = ""
 val string32 = "a"
 val string33 = "aa"
@@ -323,3 +226,4 @@
 matcher(PLUS(PLUS(Q7r1)), Q7str7)
 matcher(PLUS(PLUS(Q7r2)), Q7str7)
 
+*/
\ No newline at end of file
--- a/solutions/cw2/lexer.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/solutions/cw2/lexer.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -37,13 +37,14 @@
 implicit def string2rexp(s : String) : Rexp = 
   charlist2rexp(s.toList)
 
-implicit def RexpOps(r: Rexp) = new {
+extension (r: Rexp) {
+  def ~ (s: Rexp) = SEQ(r, s)
+  def % = STAR(r)
   def | (s: Rexp) = ALT(r, s)
-  def % = STAR(r)
-  def ~ (s: Rexp) = SEQ(r, s)
 }
 
-implicit def stringOps(s: String) = new {
+
+extension (s: String) {
   def | (r: Rexp) = ALT(s, r)
   def | (r: String) = ALT(s, r)
   def % = STAR(s)
--- a/solutions/cw3/lexer.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/solutions/cw3/lexer.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -37,13 +37,13 @@
 implicit def string2rexp(s : String) : Rexp = 
   charlist2rexp(s.toList)
 
-implicit def RexpOps(r: Rexp) = new {
+extension (r: Rexp) {
   def | (s: Rexp) = ALT(r, s)
   def % = STAR(r)
   def ~ (s: Rexp) = SEQ(r, s)
 }
 
-implicit def stringOps(s: String) = new {
+extension (s: String) {
   def | (r: Rexp) = ALT(s, r)
   def | (r: String) = ALT(s, r)
   def % = STAR(s)
@@ -82,7 +82,7 @@
   case RECD(_, r1) => der(c, r1)
   case RANGE(s) => if (s.contains(c)) ONE else ZERO 
   case PLUS(r1) => SEQ(der(c, r1), STAR(r1))
-  case OPTIONAL(r1) => ALT(der(c, r1), ZERO)
+  case OPTIONAL(r1) => der(c, r1)
   case NTIMES(r, i) => 
     if (i == 0) ZERO else SEQ(der(c, r), NTIMES(r, i - 1))
 }
@@ -119,7 +119,7 @@
   case RECD(x, r) => Rec(x, mkeps(r))
 
   case PLUS(r) => Stars(List(mkeps(r)))   // the first copy must match the empty string
-  case OPTIONAL(r) => Right(Empty)
+  case OPTIONAL(r) => if (nullable(r)) Stars(List(mkeps(r))) else Stars(Nil)
   case NTIMES(r, i) => Stars(List.fill(i)(mkeps(r)))
 }
 
@@ -136,7 +136,7 @@
 
   case (RANGE(_), Empty) => Chr(c)
   case (PLUS(r), Sequ(v1, Stars(vs))) => Stars(inj(r, c, v1)::vs)
-  case (OPTIONAL(r), Left(v1)) => Left(inj(r, c, v1))
+  case (OPTIONAL(r), Left(v1)) => Stars(List(inj(r, c, v1)))
   case (NTIMES(r, n), Sequ(v1, Stars(vs))) => Stars(inj(r, c, v1)::vs)
 }
 
@@ -201,18 +201,18 @@
 // Language specific code
 val KEYWORD : Rexp = "while" | "if" | "then" | "else" | "do" | "for" | "to" | "true" | "false" | "read" | "write" | "skip" 
 val OP : Rexp = "+" | "-" | "*" | "%" | "/" | "==" | "!=" | ">" | "<" | ">=" | "<=" | ":=" | "&&" | "||"
-val LET: Rexp = RANGE(('A' to 'Z').toSet ++ ('a' to 'z'))
-val SYM : Rexp = (LET | RANGE(Set('.', '_', '>', '<', '=', ';', ',', ':')))
+val LET: Rexp = RANGE(('A' to 'Z').toSet ++ ('a' to 'z').toSet)
+val SYM : Rexp = RANGE(Set('.', '_', '>', '<', '=', ';', ',', ':', ')', '('))
 val PARENS : Rexp = "(" | "{" | ")" | "}"
 val SEMI : Rexp = ";"
 val WHITESPACE : Rexp = PLUS(" ") | "\n" | "\t" | "\r"
 val DIGIT : Rexp = RANGE(('0' to '9').toSet)
 val DIGIT1 : Rexp = RANGE(('1' to '9').toSet)
-val STRING : Rexp = "\"" ~ (SYM | " " | "\\n" | DIGIT).% ~ "\""
+val STRING : Rexp = "\"" ~ (LET | SYM | DIGIT | " " | "\\n").% ~ "\""
 val ID : Rexp = LET ~ (LET | "_" | DIGIT).%
 val NUM : Rexp = "0" | (DIGIT1 ~ DIGIT.%)
 val EOL : Rexp = "\n" | "\r\n"
-val COMMENT : Rexp = "//" ~ (SYM | PARENS | " " | DIGIT).% ~ EOL 
+val COMMENT : Rexp = "//" ~ (LET | SYM | PARENS | " " | DIGIT).% ~ EOL 
 
 val WHILE_REGS = (("k" $ KEYWORD) | 
                   ("o" $ OP) | 
@@ -224,7 +224,18 @@
                   ("n" $ NUM) |
 		  ("c" $ COMMENT)).%
 
-// Token
+
+
+def esc(raw: String): String = {
+  import scala.reflect.runtime.universe._
+  Literal(Constant(raw)).toString
+}
+
+def escape(tks: List[(String, String)]) =
+  tks.map{ case (s1, s2) => (s1, esc(s2))}
+
+
+// Tokens
 abstract class Token extends Serializable 
 case class T_KEYWORD(s: String) extends Token
 case class T_OP(s: String) extends Token
@@ -247,21 +258,3 @@
 // Tokenise
 def tokenise(s: String) : List[Token] = 
   lexing_simp(WHILE_REGS, s).collect(token)
-
-
-val fact = """
-write "Input n please";
-read n;
-write "The factors of n are";
-f := 2;
-while n != 1 do {
-    while (n / f) * f == n do {
-        write f;
-        n := n / f
-    };
-    f := f + 1
-}
-"""
-
-//println(tokenise(fact))
-
--- a/solutions/cw3/parser.sc	Fri Sep 15 10:49:33 2023 +0100
+++ b/solutions/cw3/parser.sc	Sun Sep 17 19:12:57 2023 +0100
@@ -4,32 +4,67 @@
 import lexer._ 
 
 
-case class ~[+A, +B](_1: A, _2: B)
-type IsSeq[A] = A => Seq[_]
+case class ~[+A, +B](x: A, y: B)
+
+// parser combinators
+
+abstract class Parser[I, T](using is: I => Seq[_])  {
+  def parse(in: I): Set[(T, I)]  
+
+  def parse_all(in: I) : Set[T] =
+    for ((hd, tl) <- parse(in); 
+        if is(tl).isEmpty) yield hd
+}
 
-abstract class Parser[I : IsSeq, T] {
-  def parse(ts: I): Set[(T, I)]
+// alternative parser
+class AltParser[I, T](p: => Parser[I, T], 
+                      q: => Parser[I, T])(using I => Seq[_]) extends Parser[I, T] {
+  def parse(in: I) = p.parse(in) ++ q.parse(in)   
+}
 
-  def parse_all(ts: I) : Set[T] =
-    for ((head, tail) <- parse(ts); if tail.isEmpty) yield head
+// sequence parser
+class SeqParser[I, T, S](p: => Parser[I, T], 
+                         q: => Parser[I, S])(using I => Seq[_]) extends Parser[I, ~[T, S]] {
+  def parse(in: I) = 
+    for ((hd1, tl1) <- p.parse(in); 
+         (hd2, tl2) <- q.parse(tl1)) yield (new ~(hd1, hd2), tl2)
 }
 
-class SeqParser[I : IsSeq, T, S](p: => Parser[I, T], q: => Parser[I, S]) extends Parser[I, ~[T, S]] {
-  def parse(sb: I) = 
-    for ((head1, tail1) <- p.parse(sb); 
-         (head2, tail2) <- q.parse(tail1)) yield (new ~(head1, head2), tail2)
+// map parser
+class MapParser[I, T, S](p: => Parser[I, T], 
+                         f: T => S)(using I => Seq[_]) extends Parser[I, S] {
+  def parse(in: I) = for ((hd, tl) <- p.parse(in)) yield (f(hd), tl)
 }
 
-class AltParser[I : IsSeq, T](p: => Parser[I, T], q: => Parser[I, T]) extends Parser[I, T] {
-  def parse(sb: I) = p.parse(sb) ++ q.parse(sb)   
+
+/*
+// atomic parser for (particular) strings
+case class StrParser(s: String) extends Parser[String, String] {
+  def parse(sb: String) = {
+    val (prefix, suffix) = sb.splitAt(s.length)
+    if (prefix == s) Set((prefix, suffix)) else Set()
+  }
 }
 
-class FunParser[I : IsSeq, T, S](p: => Parser[I, T], f: T => S) extends Parser[I, S] {
-  def parse(sb: I) = 
-    for ((head, tail) <- p.parse(sb)) yield (f(head), tail)
+extension (sc: StringContext) 
+  def p(args: Any*) = StrParser(sc.s(args:_*))
+*/
+
+// more convenient syntax for parser combinators
+extension [I, T](p: Parser[I, T])(using I => Seq[_]) {
+  def ||(q : => Parser[I, T]) = new AltParser[I, T](p, q)
+  def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q)
+  def map[S](f: => T => S) = new MapParser[I, T, S](p, f)
 }
 
 // New parser that takes as input a list of tokens
+case class TokenParser(t: Token) extends Parser[List[Token], Token] {
+    def parse(in: List[Token]) = {
+      // an example of an atomic parser for characters
+      if (!in.isEmpty && in.head == t) Set((t, in.tail)) else Set()
+    }
+}   
+
 case class TokenListParser(ts: List[Token]) extends Parser[List[Token], List[Token]] {
     def parse(tsb: List[Token]) = {
         val (prefix, suffix) = tsb.splitAt(ts.length)
@@ -39,34 +74,16 @@
 
 // Implicit definitions to go from a token 
 // or a list of tokens to a TokenListParser
-implicit def token2parser(t: Token) = TokenListParser(List(t))
-implicit def tokenList2parser(ts: List[Token]) = TokenListParser(ts)
+implicit def token2parser(t: Token) : Parser[List[Token], Token] = 
+  TokenParser(t)
 
-implicit def ParserOps[I : IsSeq, T](p: Parser[I, T]) = new {
-  def || (q : => Parser[I, T]) = new AltParser[I, T](p, q)
-  def ==>[S] (f: => T => S) = new FunParser[I, T, S](p, f)
-  def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q)
+extension (t: Token) {
+    def || (q : => Parser[List[Token], Token]) = 
+      new AltParser[List[Token], Token](t, q)
+    def ~[S](q : => Parser[List[Token], S]) = 
+      new SeqParser[List[Token], Token, S](t, q)  
 }
 
-implicit def TokenOps(t: Token) = new {
-    def || (q : => Parser[List[Token], List[Token]]) = new AltParser[List[Token], List[Token]](List(t), q)
-    def || (qs : List[Token]) = new AltParser[List[Token], List[Token]](List(t), qs)
-    def ==>[S] (f: => List[Token] => S) = new FunParser[List[Token], List[Token], S](List(t), f)
-    def ~[S](q : => Parser[List[Token], S]) =
-        new SeqParser[List[Token], List[Token], S](List(t), q)
-    def ~ (qs : List[Token]) =
-        new SeqParser[List[Token], List[Token], List[Token]](List(t), qs)
-}
-
-implicit def TokenListOps(ts: List[Token]) = new {
-    def || (q : => Parser[List[Token], List[Token]]) = new AltParser[List[Token], List[Token]](ts, q)
-    def || (qs : List[Token]) = new AltParser[List[Token], List[Token]](ts, qs)
-    def ==>[S] (f: => List[Token] => S) = new FunParser[List[Token], List[Token], S](ts, f)
-    def ~[S](q : => Parser[List[Token], S]) =
-        new SeqParser[List[Token], List[Token], S](ts, q)
-    def ~ (qs : List[Token]) =
-        new SeqParser[List[Token], List[Token], List[Token]](ts, qs)
-}
 
 // Abstract Syntax Trees
 abstract class Stmt
@@ -114,48 +131,49 @@
     }
 }
 
+
 // WHILE Language Parsing
 lazy val AExp: Parser[List[Token], AExp] = 
-  (Te ~ T_OP("+") ~ AExp) ==> { case x ~ _ ~ z => Aop("+", x, z): AExp } ||
-  (Te ~ T_OP("-") ~ AExp) ==> { case x ~ _ ~ z => Aop("-", x, z): AExp } || Te
+  (Te ~ T_OP("+") ~ AExp).map{ case x ~ _ ~ z => Aop("+", x, z): AExp } ||
+  (Te ~ T_OP("-") ~ AExp).map{ case x ~ _ ~ z => Aop("-", x, z): AExp } || Te
 lazy val Te: Parser[List[Token], AExp] = 
-  (Fa ~ T_OP("*") ~ Te) ==> { case x ~ _ ~ z => Aop("*", x, z): AExp } || 
-  (Fa ~ T_OP("/") ~ Te) ==> { case x ~ _ ~ z => Aop("/", x, z): AExp } || 
-  (Fa ~ T_OP("%") ~ Te) ==> { case x ~ _ ~ z => Aop("%", x, z): AExp } || Fa  
+  (Fa ~ T_OP("*") ~ Te).map{ case x ~ _ ~ z => Aop("*", x, z): AExp } || 
+  (Fa ~ T_OP("/") ~ Te).map{ case x ~ _ ~ z => Aop("/", x, z): AExp } || 
+  (Fa ~ T_OP("%") ~ Te).map{ case x ~ _ ~ z => Aop("%", x, z): AExp } || Fa  
 lazy val Fa: Parser[List[Token], AExp] = 
-   (T_PAREN("(") ~ AExp ~ T_PAREN(")")) ==> { case _ ~ y ~ _ => y } || 
-   IdParser() ==> Var  || 
-   NumParser() ==> Num
+   (T_PAREN("(") ~ AExp ~ T_PAREN(")")).map{ case _ ~ y ~ _ => y } || 
+   IdParser().map{Var(_)}  || 
+   NumParser().map{Num(_)}
 
 lazy val BExp: Parser[List[Token], BExp] = 
-   (AExp ~ T_OP("==") ~ AExp) ==> { case x ~ _ ~ z => Bop("==", x, z): BExp } || 
-   (AExp ~ T_OP("!=") ~ AExp) ==> { case x ~ _ ~ z => Bop("!=", x, z): BExp } || 
-   (AExp ~ T_OP("<") ~ AExp) ==> { case x ~ _ ~ z => Bop("<", x, z): BExp } || 
-   (AExp ~ T_OP(">") ~ AExp) ==> { case x ~ _ ~ z => Bop(">", x, z): BExp } ||
-   (T_PAREN("(") ~ BExp ~ T_PAREN(")") ~ T_OP("&&") ~ BExp) ==> { case _ ~ y ~ _ ~ _ ~ v => And(y, v): BExp } ||
-   (T_PAREN("(") ~ BExp ~ T_PAREN(")") ~ T_OP("||") ~ BExp) ==> { case _ ~ y ~ _ ~ _ ~ v => Or(y, v): BExp } ||
-   (T_KEYWORD("true") ==> (_ => True: BExp )) || 
-   (T_KEYWORD("false") ==> (_ => False: BExp )) ||
-   (T_PAREN("(") ~ BExp ~ T_PAREN(")")) ==> { case _ ~ x ~ _ => x }
+   (AExp ~ T_OP("==") ~ AExp).map{ case x ~ _ ~ z => Bop("==", x, z): BExp } || 
+   (AExp ~ T_OP("!=") ~ AExp).map{ case x ~ _ ~ z => Bop("!=", x, z): BExp } || 
+   (AExp ~ T_OP("<") ~ AExp).map{ case x ~ _ ~ z => Bop("<", x, z): BExp } || 
+   (AExp ~ T_OP(">") ~ AExp).map{ case x ~ _ ~ z => Bop(">", x, z): BExp } ||
+   (T_PAREN("(") ~ BExp ~ T_PAREN(")") ~ T_OP("&&") ~ BExp).map{ case _ ~ y ~ _ ~ _ ~ v => And(y, v): BExp } ||
+   (T_PAREN("(") ~ BExp ~ T_PAREN(")") ~ T_OP("||") ~ BExp).map{ case _ ~ y ~ _ ~ _ ~ v => Or(y, v): BExp } ||
+   (T_KEYWORD("true").map(_ => True: BExp )) || 
+   (T_KEYWORD("false").map(_ => False: BExp )) ||
+   (T_PAREN("(") ~ BExp ~ T_PAREN(")")).map{ case _ ~ x ~ _ => x }
 
 lazy val Stmt: Parser[List[Token], Stmt] =
-    T_KEYWORD("skip") ==> (_ => Skip: Stmt) ||
-    (IdParser() ~ T_OP(":=") ~ AExp) ==> { case id ~ _ ~ z => Assign(id, z): Stmt } ||
-    (T_KEYWORD("if") ~ BExp ~ T_KEYWORD("then") ~ Block ~ T_KEYWORD("else") ~ Block) ==> { case _ ~ y ~ _ ~ u ~ _ ~ w => If(y, u, w): Stmt } ||
-    (T_KEYWORD("while") ~ BExp ~ T_KEYWORD("do") ~ Block) ==> { case _ ~ y ~ _ ~ w => While(y, w) : Stmt } ||
-    (T_KEYWORD("read") ~ IdParser()) ==> { case _ ~ id => Read(id): Stmt} ||
-    (T_KEYWORD("write") ~ IdParser()) ==> { case _ ~ id => WriteId(id): Stmt} ||
-    (T_KEYWORD("write") ~ StringParser()) ==> { case _ ~ s => WriteString(s): Stmt} || 
-    (T_KEYWORD("write") ~ T_PAREN("(") ~ IdParser() ~ T_PAREN(")")) ==> { case _ ~ _ ~ id ~ _ => WriteId(id): Stmt} ||
-    (T_KEYWORD("write") ~  T_PAREN("(") ~ StringParser() ~ T_PAREN(")")) ==> { case _ ~ _ ~ s ~ _ => WriteString(s): Stmt}
+    T_KEYWORD("skip").map(_ => Skip: Stmt) ||
+    (IdParser() ~ T_OP(":=") ~ AExp).map{ case id ~ _ ~ z => Assign(id, z): Stmt } ||
+    (T_KEYWORD("if") ~ BExp ~ T_KEYWORD("then") ~ Block ~ T_KEYWORD("else") ~ Block).map{ case _ ~ y ~ _ ~ u ~ _ ~ w => If(y, u, w): Stmt } ||
+    (T_KEYWORD("while") ~ BExp ~ T_KEYWORD("do") ~ Block).map{ case _ ~ y ~ _ ~ w => While(y, w) : Stmt } ||
+    (T_KEYWORD("read") ~ IdParser()).map{ case _ ~ id => Read(id): Stmt} ||
+    (T_KEYWORD("write") ~ IdParser()).map{ case _ ~ id => WriteId(id): Stmt} ||
+    (T_KEYWORD("write") ~ StringParser()).map{ case _ ~ s => WriteString(s): Stmt} || 
+    (T_KEYWORD("write") ~ T_PAREN("(") ~ IdParser() ~ T_PAREN(")")).map{ case _ ~ _ ~ id ~ _ => WriteId(id): Stmt} ||
+    (T_KEYWORD("write") ~  T_PAREN("(") ~ StringParser() ~ T_PAREN(")")).map{ case _ ~ _ ~ s ~ _ => WriteString(s): Stmt}
 
 lazy val Stmts: Parser[List[Token], Block] =
-    (Stmt ~ T_SEMI ~ Stmts) ==> { case x ~ _ ~ z => x :: z : Block } ||
-    (Stmt ==> (s => List(s) : Block))
+    (Stmt ~ T_SEMI ~ Stmts).map{ case x ~ _ ~ z => x :: z : Block } ||
+    (Stmt.map(s => List(s) : Block))
 
 lazy val Block: Parser[List[Token], Block] =
-    (T_PAREN("{") ~ Stmts ~ T_PAREN("}")) ==> { case x ~ y ~ z => y} ||
-    (Stmt ==> (s => List(s)))
+    (T_PAREN("{") ~ Stmts ~ T_PAREN("}")).map{ case x ~ y ~ z => y} ||
+    (Stmt.map(s => List(s)))
 
 // Testing with programs 2 & 3
 
--- a/solutions/cw3/parser2.sc	Fri Sep 15 10:49:33 2023 +0100
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,255 +0,0 @@
-// CW3
-
-import $file.lexer
-import lexer._ 
-
-
-case class ~[+A, +B](_1: A, _2: B)
-type IsSeq[A] = A => Seq[_]
-
-abstract class Parser[I : IsSeq, T] {
-  def parse(ts: I): Set[(T, I)]
-
-  def parse_all(ts: I) : Set[T] =
-    for ((head, tail) <- parse(ts); if tail.isEmpty) yield head
-}
-
-class SeqParser[I : IsSeq, T, S](p: => Parser[I, T], q: => Parser[I, S]) extends Parser[I, ~[T, S]] {
-  def parse(sb: I) = 
-    for ((head1, tail1) <- p.parse(sb); 
-         (head2, tail2) <- q.parse(tail1)) yield (new ~(head1, head2), tail2)
-}
-
-class AltParser[I : IsSeq, T](p: => Parser[I, T], q: => Parser[I, T]) extends Parser[I, T] {
-  def parse(sb: I) = p.parse(sb) ++ q.parse(sb)   
-}
-
-class FunParser[I : IsSeq, T, S](p: => Parser[I, T], f: T => S) extends Parser[I, S] {
-  def parse(sb: I) = 
-    for ((head, tail) <- p.parse(sb)) yield (f(head), tail)
-}
-
-// New parser that takes as input a list of tokens
-case class TokenListParser(ts: List[Token]) extends Parser[List[Token], List[Token]] {
-    def parse(tsb: List[Token]) = {
-        val (prefix, suffix) = tsb.splitAt(ts.length)
-        if (prefix == ts) Set((prefix, suffix)) else Set()
-    }
-}
-
-// Implicit definitions to go from a token 
-// or a list of tokens to a TokenListParser
-implicit def token2parser(t: Token) = TokenListParser(List(t))
-implicit def tokenList2parser(ts: List[Token]) = TokenListParser(ts)
-
-implicit def ParserOps[I : IsSeq, T](p: Parser[I, T]) = new {
-  def || (q : => Parser[I, T]) = new AltParser[I, T](p, q)
-  def ==>[S] (f: => T => S) = new FunParser[I, T, S](p, f)
-  def ~[S] (q : => Parser[I, S]) = new SeqParser[I, T, S](p, q)
-}
-
-implicit def TokenOps(t: Token) = new {
-    def || (q : => Parser[List[Token], List[Token]]) = new AltParser[List[Token], List[Token]](List(t), q)
-    def || (qs : List[Token]) = new AltParser[List[Token], List[Token]](List(t), qs)
-    def ==>[S] (f: => List[Token] => S) = new FunParser[List[Token], List[Token], S](List(t), f)
-    def ~[S](q : => Parser[List[Token], S]) =
-        new SeqParser[List[Token], List[Token], S](List(t), q)
-    def ~ (qs : List[Token]) =
-        new SeqParser[List[Token], List[Token], List[Token]](List(t), qs)
-}
-
-implicit def TokenListOps(ts: List[Token]) = new {
-    def || (q : => Parser[List[Token], List[Token]]) = new AltParser[List[Token], List[Token]](ts, q)
-    def || (qs : List[Token]) = new AltParser[List[Token], List[Token]](ts, qs)
-    def ==>[S] (f: => List[Token] => S) = new FunParser[List[Token], List[Token], S](ts, f)
-    def ~[S](q : => Parser[List[Token], S]) =
-        new SeqParser[List[Token], List[Token], S](ts, q)
-    def ~ (qs : List[Token]) =
-        new SeqParser[List[Token], List[Token], List[Token]](ts, qs)
-}
-
-// Abstract Syntax Trees
-abstract class Stmt
-abstract class AExp
-abstract class BExp
-
-type Block = List[Stmt]
-
-case object Skip extends Stmt
-case class If(a: BExp, bl1: Block, bl2: Block) extends Stmt
-case class While(b: BExp, bl: Block) extends Stmt
-case class Assign(s: String, a: AExp) extends Stmt
-case class Read(s: String) extends Stmt
-case class WriteId(s: String) extends Stmt  // for printing values of variables
-case class WriteString(s: String) extends Stmt  // for printing words
-
-case class Var(s: String) extends AExp
-case class Num(i: Int) extends AExp
-case class Aop(o: String, a1: AExp, a2: AExp) extends AExp
-
-case object True extends BExp
-case object False extends BExp
-case class Bop(o: String, a1: AExp, a2: AExp) extends BExp
-case class And(b1: BExp, b2: BExp) extends BExp
-case class Or(b1: BExp, b2: BExp) extends BExp
-
-case class IdParser() extends Parser[List[Token], String] {
-    def parse(tsb: List[Token]) = tsb match {
-        case T_ID(id) :: rest => Set((id, rest))
-        case _ => Set()
-    }
-}
-
-case class NumParser() extends Parser[List[Token], Int] {
-    def parse(tsb: List[Token]) = tsb match {
-        case T_NUM(n) :: rest => Set((n, rest))
-        case _ => Set()
-    }
-}
-
-case class StringParser() extends Parser[List[Token], String] {
-    def parse(tsb: List[Token]) = tsb match {
-        case T_STRING(s) :: rest => Set((s, rest))
-        case _ => Set()
-    }
-}
-
-case class TokParser(s: String) extends Parser[List[Token], String] {
-    def parse(tsb: List[Token]) = tsb match {
-        case T_OP(o) :: rest if s == o => Set((o, rest))
-        case T_KWD(k) :: rest if s == k => Set((k, rest))
-        case _ => Set()
-    }
-}
-
-implicit def parser_interpolation(sc: StringContext) = new {
-    def p(args: Any*) = TokParser(sc.s(args:_*))
-}    
-
-
-// WHILE Language Parsing
-lazy val AExp: Parser[List[Token], AExp] = 
-  (Te ~ T_OP("+") ~ AExp) ==> { case x ~ _ ~ z => Aop("+", x, z): AExp } ||
-  (Te ~ T_OP("-") ~ AExp) ==> { case x ~ _ ~ z => Aop("-", x, z): AExp } || Te
-lazy val Te: Parser[List[Token], AExp] = 
-  (Fa ~ T_OP("*") ~ Te) ==> { case x ~ _ ~ z => Aop("*", x, z): AExp } || 
-  (Fa ~ T_OP("/") ~ Te) ==> { case x ~ _ ~ z => Aop("/", x, z): AExp } || 
-  (Fa ~ T_OP("%") ~ Te) ==> { case x ~ _ ~ z => Aop("%", x, z): AExp } || Fa  
-lazy val Fa: Parser[List[Token], AExp] = 
-   (T_PAREN("(") ~ AExp ~ T_PAREN(")")) ==> { case _ ~ y ~ _ => y } || 
-   IdParser() ==> Var  || 
-   NumParser() ==> Num
-
-lazy val BExp: Parser[List[Token], BExp] = 
-   (AExp ~ T_OP("==") ~ AExp) ==> { case x ~ _ ~ z => Bop("==", x, z): BExp } || 
-   (AExp ~ T_OP("!=") ~ AExp) ==> { case x ~ _ ~ z => Bop("!=", x, z): BExp } || 
-   (AExp ~ T_OP("<") ~ AExp) ==> { case x ~ _ ~ z => Bop("<", x, z): BExp } || 
-   (AExp ~ T_OP(">") ~ AExp) ==> { case x ~ _ ~ z => Bop(">", x, z): BExp } ||
-   (T_PAREN("(") ~ BExp ~ List(T_PAREN(")"), T_OP("&&")) ~ BExp) ==> { case _ ~ y ~ _ ~ v => And(y, v): BExp } ||
-   (T_PAREN("(") ~ BExp ~ List(T_PAREN(")"), T_OP("||")) ~ BExp) ==> { case _ ~ y ~ _ ~ v => Or(y, v): BExp } ||
-   (T_KEYWORD("true") ==> (_ => True: BExp )) || 
-   (T_KEYWORD("false") ==> (_ => False: BExp )) ||
-   (T_PAREN("(") ~ BExp ~ T_PAREN(")")) ==> { case _ ~ x ~ _ => x }
-
-lazy val Stmt: Parser[List[Token], Stmt] =
-    T_KEYWORD("skip") ==> (_ => Skip: Stmt) ||
-    (IdParser() ~ T_OP(":=") ~ AExp) ==> { case id ~ _ ~ z => Assign(id, z): Stmt } ||
-    (T_KEYWORD("if") ~ BExp ~ T_KEYWORD("then") ~ Block ~ T_KEYWORD("else") ~ Block) ==> { case _ ~ y ~ _ ~ u ~ _ ~ w => If(y, u, w): Stmt } ||
-    (T_KEYWORD("while") ~ BExp ~ T_KEYWORD("do") ~ Block) ==> { case _ ~ y ~ _ ~ w => While(y, w) : Stmt } ||
-    (T_KEYWORD("read") ~ IdParser()) ==> { case _ ~ id => Read(id): Stmt} ||
-    (T_KEYWORD("write") ~ IdParser()) ==> { case _ ~ id => WriteId(id): Stmt} ||
-    (T_KEYWORD("write") ~ StringParser()) ==> { case _ ~ s => WriteString(s): Stmt}
-
-lazy val Stmts: Parser[List[Token], Block] =
-    (Stmt ~ T_SEMI ~ Stmts) ==> { case x ~ _ ~ z => x :: z : Block } ||
-    (Stmt ==> (s => List(s) : Block))
-
-lazy val Block: Parser[List[Token], Block] =
-    (T_PAREN("{") ~ Stmts ~ T_PAREN("}")) ==> { case x ~ y ~ z => y} ||
-    (Stmt ==> (s => List(s)))
-
-// Testing with programs 2 & 3
-
-println("Fibonacci")
-println(Stmts.parse_all(tokenise(os.read(os.pwd / "fib.while"))))
-
-println("Loops")
-println(Stmts.parse_all(tokenise(os.read(os.pwd / "loops.while"))))
-
-println("Collatz")
-println(Stmts.parse_all(tokenise(os.read(os.pwd / "collatz2.while"))))
-
-
-// Interpreter
-
-// Environment to store values of variables
-type Env = Map[String, Int]
-
-def eval_aexp(a: AExp, env: Env) : Int = a match {
-  case Num(i) => i
-  case Var(s) => env(s)
-  case Aop("+", a1, a2) => eval_aexp(a1, env) + eval_aexp(a2, env)
-  case Aop("-", a1, a2) => eval_aexp(a1, env) - eval_aexp(a2, env)
-  case Aop("*", a1, a2) => eval_aexp(a1, env) * eval_aexp(a2, env)
-  case Aop("/", a1, a2) => eval_aexp(a1, env) / eval_aexp(a2, env)
-  case Aop("%", a1, a2) => eval_aexp(a1, env) % eval_aexp(a2, env)
-}
-
-def eval_bexp(b: BExp, env: Env) : Boolean = b match {
-  case True => true
-  case False => false
-  case Bop("==", a1, a2) => eval_aexp(a1, env) == eval_aexp(a2, env)
-  case Bop("!=", a1, a2) => !(eval_aexp(a1, env) == eval_aexp(a2, env))
-  case Bop(">", a1, a2) => eval_aexp(a1, env) > eval_aexp(a2, env)
-  case Bop("<", a1, a2) => eval_aexp(a1, env) < eval_aexp(a2, env)
-  case And(b1, b2) => eval_bexp(b1, env) && eval_bexp(b2, env)
-  case Or(b1, b2) => eval_bexp(b1, env) || eval_bexp(b2, env)
-}
-
-// Import needed to take int as input from the user
-import scala.io.StdIn.readInt
-
-def eval_stmt(s: Stmt, env: Env) : Env = s match {
-  case Skip => env
-  case Assign(x, a) => env + (x -> eval_aexp(a, env))
-  case If(b, bl1, bl2) => if (eval_bexp(b, env)) eval_bl(bl1, env) else eval_bl(bl2, env) 
-  case While(b, bl) => 
-    if (eval_bexp(b, env)) eval_stmt(While(b, bl), eval_bl(bl, env))
-    else env
-
-  case WriteId(x) => { print(env(x)) ; env }
-  case WriteString(x) => {
-        print(x.replaceAll("\"", "").replaceAll("""\\n""", "\n")) ;
-        env
-       }
-
-  case Read(x) => { 
-      println("Enter an integer and press ENTER:") ; 
-      val n = readInt() ; // Note: Does not work when using the REPL
-      eval_stmt(Assign(x, Num(n)), env) 
-  }
-}
-
-def eval_bl(bl: Block, env: Env) : Env = bl match {
-  case Nil => env
-  case s::bl => eval_bl(bl, eval_stmt(s, env))
-}
-
-def eval(bl: Block) : Env = eval_bl(bl, Map())
-
-println("Factors eval")
-println(eval(Stmts.parse_all(tokenise(os.read(os.pwd / "factors.while"))).head))
-
-println("Primes eval")
-println(eval(Stmts.parse_all(tokenise(os.read(os.pwd / "primes.while"))).head))
-
-
-println("Collatz2 eval")
-println(eval(Stmts.parse_all(tokenise(os.read(os.pwd / "collatz2.while"))).head))
-
-println("Loops eval")
-val start = System.nanoTime()
-println(eval(Stmts.parse_all(tokenise(os.read(os.pwd / "loops.while"))).head))
-val end = System.nanoTime()
-println("Time taken in seconds: ")
-println((end - start)/(1.0e9))