various bugfixes for generics; added generic sort proc

lib/pure/algorithm.nim

@@ -0,0 +1,101 @@
+#
+#
+#            Nimrod's Runtime Library
+#        (c) Copyright 2010 Andreas Rumpf
+#
+#    See the file "copying.txt", included in this
+#    distribution, for details about the copyright.
+#
+
+## This module implements some common generic algorithms.
+
+type
+  TSortOrder* = enum   ## sort order
+    Descending, Ascending 
+
+proc `*`*(x: int, order: TSortOrder): int {.inline.} = 
+  ## flips `x` if ``order == Descending``;
+  ## if ``order == Ascending`` then `x` is returned.
+  ## `x` is supposed to be the result of a comparator.
+  var y = order.ord - 1
+  result = (x xor y) - y
+
+proc reverse*[T](a: var openArray[T], first, last: int) =
+  ## reverses the array ``a[first..last]``.
+  var x = first
+  var y = last
+  while x < y:
+    swap(a[x], a[y])
+    dec(y)
+    inc(x)
+
+proc reverse*[T](a: var openArray[T]) =
+  ## reverses the array `a`.
+  reverse(a, 0, a.high)
+
+const
+  onlySafeCode = false
+
+proc merge[T](a, b: var openArray[T], lo, m, hi: int, 
+              cmp: proc (x, y: T): int, order: TSortOrder) =
+  template `<-` (a, b: expr) = 
+    when onlySafeCode:
+      a = b
+    else:
+      copyMem(addr(a), addr(b), sizeof(T))
+  # optimization: If max(left) <= min(right) there is nothing to do!
+  # 1 2 3 4  ## 5 6 7 8
+  # -> O(n) for sorted arrays.
+  # On random data this safes up to 40% of merge calls
+  if cmp(a[m], a[m+1]) * order <= 0: return
+  var j = lo
+  # copy a[j..m] into b:
+  assert j <= m
+  when onlySafeCode:
+    var bb = 0
+    while j <= m:
+      b[bb] = a[j]
+      inc(bb)
+      inc(j)
+  else:
+    CopyMem(addr(b[0]), addr(a[j]), sizeof(T)*(m-j+1))
+    j = m+1
+  var i = 0
+  var k = lo
+  # copy proper element back:
+  while k < j and j <= hi:
+    if cmp(b[i], a[j]) * order <= 0:
+      a[k] <- b[i]
+      inc(i)
+    else:
+      a[k] <- a[j]
+      inc(j)
+    inc(k)
+  # copy rest of b:
+  when onlySafeCode:
+    while k < j:
+      a[k] = b[i]
+      inc(k)
+      inc(i)
+  else:
+    if k < j: copyMem(addr(a[k]), addr(b[i]), sizeof(T)*(j-k))
+
+proc sort*[T](a: var openArray[T], 
+              cmp: proc (x, y: T): int = cmp,
+              order = TSortOrder.Ascending) =
+  ## Default Nimrod sort. The sorting is guaranteed to be stable and 
+  ## the worst case is guaranteed to be O(n log n).
+  ## The current implementation uses an iterative
+  ## mergesort to achieve this. It uses a temporary sequence of 
+  ## length ``a.len div 2``.
+  var n = a.len
+  var b: seq[T]
+  newSeq(b, n div 2)
+  var s = 1
+  while s < n:
+    var m = n-1-s
+    while m >= 0:
+      merge(a, b, max(m-s+1, 0), m, m+s, cmp, order)
+      dec(m, s*2)
+    s = s*2
+

lib/pure/strutils.nim

@@ -97,14 +97,15 @@ proc normalize*(s: string): string {.noSideEffect, procvar,
       add result, s[i]
 
 proc cmpIgnoreCase*(a, b: string): int {.noSideEffect,
-  rtl, extern: "nsuCmpIgnoreCase".} =
+  rtl, extern: "nsuCmpIgnoreCase", procvar.} =
   ## Compares two strings in a case insensitive manner. Returns:
   ##
   ## | 0 iff a == b
   ## | < 0 iff a < b
   ## | > 0 iff a > b
-  var i = 0
-  while i < a.len and i < b.len:
+  var i = 0
+  var m = min(a.len, b.len)
+  while i < m:
     result = ord(toLower(a[i])) - ord(toLower(b[i]))
     if result != 0: return
     inc(i)
@@ -114,7 +115,7 @@ proc cmpIgnoreCase*(a, b: string): int {.noSideEffect,
                                        # thus we compile without checks here
 
 proc cmpIgnoreStyle*(a, b: string): int {.noSideEffect,
-  rtl, extern: "nsuCmpIgnoreStyle".} =
+  rtl, extern: "nsuCmpIgnoreStyle", procvar.} =
   ## Compares two strings normalized (i.e. case and
   ## underscores do not matter). Returns:
   ##

lib/pure/unicode.nim

@@ -1075,7 +1075,7 @@ proc binarySearch(c: irune, tab: openArray[iRune], len, stride: int): int =
     return t
   return -1
 
-proc toLower*(c: TRune): TRune {.rtl, extern: "nuc$1".} = 
+proc toLower*(c: TRune): TRune {.rtl, extern: "nuc$1", procvar.} = 
   ## Converts `c` into lower case. This works for any Unicode character.
   ## If possible, prefer `toLower` over `toUpper`. 
   var c = irune(c)
@@ -1087,7 +1087,7 @@ proc toLower*(c: TRune): TRune {.rtl, extern: "nuc$1".} =
     return TRune(c + toLowerSinglets[p+1] - 500)
   return TRune(c)
 
-proc toUpper*(c: TRune): TRune {.rtl, extern: "nuc$1".} = 
+proc toUpper*(c: TRune): TRune {.rtl, extern: "nuc$1", procvar.} = 
   ## Converts `c` into upper case. This works for any Unicode character.
   ## If possible, prefer `toLower` over `toUpper`. 
   var c = irune(c)
@@ -1099,14 +1099,14 @@ proc toUpper*(c: TRune): TRune {.rtl, extern: "nuc$1".} =
     return TRune(c + toUpperSinglets[p+1] - 500)
   return TRune(c)
 
-proc toTitle*(c: TRune): TRune {.rtl, extern: "nuc$1".} = 
+proc toTitle*(c: TRune): TRune {.rtl, extern: "nuc$1", procvar.} = 
   var c = irune(c)
   var p = binarySearch(c, toTitleSinglets, len(toTitleSinglets) div 2, 2)
   if p >= 0 and c == toTitleSinglets[p]:
     return TRune(c + toTitleSinglets[p+1] - 500)
   return TRune(c)
 
-proc isLower*(c: TRune): bool {.rtl, extern: "nuc$1".} = 
+proc isLower*(c: TRune): bool {.rtl, extern: "nuc$1", procvar.} = 
   ## returns true iff `c` is a lower case Unicode character
   ## If possible, prefer `isLower` over `isUpper`. 
   var c = irune(c)
@@ -1118,7 +1118,7 @@ proc isLower*(c: TRune): bool {.rtl, extern: "nuc$1".} =
   if p >= 0 and c == toUpperSinglets[p]:
     return true
 
-proc isUpper*(c: TRune): bool {.rtl, extern: "nuc$1".} = 
+proc isUpper*(c: TRune): bool {.rtl, extern: "nuc$1", procvar.} = 
   ## returns true iff `c` is a upper case Unicode character
   ## If possible, prefer `isLower` over `isUpper`. 
   var c = irune(c)
@@ -1130,7 +1130,7 @@ proc isUpper*(c: TRune): bool {.rtl, extern: "nuc$1".} =
   if p >= 0 and c == toLowerSinglets[p]:
     return true
 
-proc isAlpha*(c: TRune): bool {.rtl, extern: "nuc$1".} = 
+proc isAlpha*(c: TRune): bool {.rtl, extern: "nuc$1", procvar.} = 
   ## returns true iff `c` is an *alpha* Unicode character (i.e. a letter)
   if isUpper(c) or isLower(c): 
     return true
@@ -1142,10 +1142,10 @@ proc isAlpha*(c: TRune): bool {.rtl, extern: "nuc$1".} =
   if p >= 0 and c == alphaSinglets[p]:
     return true
   
-proc isTitle*(c: TRune): bool {.rtl, extern: "nuc$1".} = 
+proc isTitle*(c: TRune): bool {.rtl, extern: "nuc$1", procvar.} = 
   return isUpper(c) and isLower(c)
 
-proc isWhiteSpace*(c: TRune): bool {.rtl, extern: "nuc$1".} = 
+proc isWhiteSpace*(c: TRune): bool {.rtl, extern: "nuc$1", procvar.} = 
   ## returns true iff `c` is a Unicode whitespace character
   var c = irune(c)
   var p = binarySearch(c, spaceRanges, len(spaceRanges) div 2, 2)
@@ -1161,7 +1161,7 @@ iterator runes*(s: string): TRune =
     fastRuneAt(s, i, result, true)
     yield result
 
-proc cmpRunesIgnoreCase*(a, b: string): int {.rtl, extern: "nuc$1".} = 
+proc cmpRunesIgnoreCase*(a, b: string): int {.rtl, extern: "nuc$1", procvar.} = 
   ## compares two UTF8 strings and ignores the case. Returns:
   ##
   ## | 0 iff a == b