various bugfixes for generics; added generic sort proc

lib/impure/re.nim

@@ -62,7 +62,7 @@ proc finalizeRegEx(x: TRegEx) =
 
 proc re*(s: string, flags = {reExtended}): TRegEx =
   ## Constructor of regular expressions. Note that Nimrod's
-  ## extended raw string literals supports this syntax ``re"[abc]"`` as
+  ## extended raw string literals support this syntax ``re"[abc]"`` as
   ## a short form for ``re(r"[abc]")``.
   new(result, finalizeRegEx)
   result.h = rawCompile(s, cast[cint](flags))
@@ -152,8 +152,12 @@ proc find*(s: string, pattern: TRegEx, matches: var openarray[string],
 proc find*(s: string, pattern: TRegEx, start = 0): int =
   ## returns the starting position of ``pattern`` in ``s``. If it does not
   ## match, -1 is returned.
-  var matches: array[0..maxSubpatterns-1, string]
+  var
-  result = find(s, pattern, matches, start)
+    rawMatches: array[0..3 - 1, cint]
+    res = pcre.Exec(pattern.h, nil, s, len(s), start, 0'i32,
+      cast[ptr cint](addr(rawMatches)), 3)
+  if res < 0'i32: return res
+  return rawMatches[0]
   
 iterator findAll*(s: string, pattern: TRegEx, start = 0): string = 
   ## yields all matching captures of pattern in `s`.

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 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

lib/system.nim

@@ -623,7 +623,7 @@ template `not_in` * (x, y: expr): expr = not contains(y, x)
 proc `is` *[T, S](x: T, y: S): bool {.magic: "Is", noSideEffect.}
 template `is_not` *(x, y: expr): expr = not (x is y)
 
-proc cmp*[T, S: typeDesc](x: T, y: S): int =
+proc cmp*[T, S: typeDesc](x: T, y: S): int {.procvar.} =
   ## Generic compare proc. Returns a value < 0 iff x < y, a value > 0 iff x > y
   ## and 0 iff x == y. This is useful for writing generic algorithms without
   ## performance loss. This generic implementation uses the `==` and `<`
@@ -632,7 +632,7 @@ proc cmp*[T, S: typeDesc](x: T, y: S): int =
   if x < y: return -1
   return 1
 
-proc cmp*(x, y: string): int {.noSideEffect.}
+proc cmp*(x, y: string): int {.noSideEffect, procvar.}
   ## Compare proc for strings. More efficient than the generic version.
 
 proc `@` * [IDX, T](a: array[IDX, T]): seq[T] {.
@@ -714,7 +714,8 @@ const
 
   cpuEndian* {.magic: "CpuEndian"}: TEndian = littleEndian
     ## is the endianness of the target CPU. This is a valuable piece of
-    ## information for low-level code only. This works thanks to compiler magic.
+    ## information for low-level code only. This works thanks to compiler
+    ## magic.
     
   hostOS* {.magic: "HostOS"}: string = ""
     ## a string that describes the host operating system. Possible values:
@@ -1307,7 +1308,7 @@ proc echo*[Ty](x: openarray[Ty]) {.magic: "Echo".}
   ## available for the ECMAScript target too!
 
 template newException*(exceptn, message: expr): expr = 
-  ## creates an exception object of type "exceptn" and sets its ``msg`` field
+  ## creates an exception object of type ``exceptn`` and sets its ``msg`` field
   ## to `message`. Returns the new exception object. 
   block: # open a new scope
     var
@@ -1365,7 +1366,7 @@ when not defined(EcmaScript) and not defined(NimrodVM):
   include "system/ansi_c"
 
   proc cmp(x, y: string): int =
-    return int(c_strcmp(x, y))
+    result = int(c_strcmp(x, y))
 
   const pccHack = if defined(pcc): "_" else: "" # Hack for PCC
   when defined(windows):
@@ -1419,8 +1420,9 @@ when not defined(EcmaScript) and not defined(NimrodVM):
       ## stream* is a good idea, but since it is named ``stderr`` there are few
       ## programs out there that distinguish properly between ``stdout`` and
       ## ``stderr``. So, that's what you get if you don't name your variables
-      ## appropriately. It also annoys people if redirection via ``>output.txt``
+      ## appropriately. It also annoys people if redirection
-      ## does not work because the program writes to ``stderr``.
+      ## via ``>output.txt`` does not work because the program writes
+      ## to ``stderr``.
 
   proc Open*(f: var TFile, filename: string,
              mode: TFileMode = fmRead, bufSize: int = -1): Bool

rod/ast.nim

@@ -750,6 +750,12 @@ proc newSymNode(sym: PSym): PNode =
   result.typ = sym.typ
   result.info = sym.info
 
+proc newSymNode*(sym: PSym, info: TLineInfo): PNode = 
+  result = newNode(nkSym)
+  result.sym = sym
+  result.typ = sym.typ
+  result.info = info
+
 proc newNodeI(kind: TNodeKind, info: TLineInfo): PNode = 
   result = newNode(kind)
   result.info = info
@@ -869,6 +875,11 @@ proc len*(n: PNode): int {.inline.} =
   if isNil(n.sons): result = 0
   else: result = len(n.sons)
   
+proc safeLen*(n: PNode): int {.inline.} =
+  ## works even for leaves.
+  if n.kind in {nkNone..nkNilLit} or isNil(n.sons): result = 0
+  else: result = len(n.sons)
+  
 proc add*(father, son: PNode) =
   assert son != nil
   if isNil(father.sons): father.sons = @[]

rod/semcall.nim

@@ -79,20 +79,42 @@ proc semDirectCall(c: PContext, n: PNode, filter: TSymKinds): PNode =
     initialBinding = nil
   result = semDirectCallWithBinding(c, n, f, filter, initialBinding)
 
-proc explictGenericInstantiation(c: PContext, n: PNode, s: PSym): PNode = 
+proc explicitGenericInstError(n: PNode): PNode =
+  LocalError(n.info, errCannotInstantiateX, renderTree(n))
+  result = n
+
+proc explicitGenericInstantiation(c: PContext, n: PNode, s: PSym): PNode = 
   assert n.kind == nkBracketExpr
   for i in 1..sonsLen(n)-1:
     n.sons[i].typ = semTypeNode(c, n.sons[i], nil)
-  # we cannot check for the proper number of type parameters because in
+  var s = s
-  # `f[a,b](x, y)` `f` is not resolved yet properly.
+  var a = n.sons[0]
-  # XXX: BUG this should be checked somehow!
+  if a.kind == nkSym:
-  assert n.sons[0].kind == nkSym
+    # common case; check the only candidate has the right
+    # number of generic type parameters:
+    if safeLen(s.ast.sons[genericParamsPos]) != n.len-1:
+      return explicitGenericInstError(n)
+  elif a.kind == nkSymChoice:
+    # choose the generic proc with the proper number of type parameters.
+    # XXX I think this could be improved by reusing sigmatch.ParamTypesMatch.
+    # It's good enough for now.
+    var candidateCount = 0
+    for i in countup(0, len(a)-1): 
+      var candidate = a.sons[i].sym
+      if candidate.kind in {skProc, skMethod, skConverter, skIterator}: 
+        # if suffices that the candidate has the proper number of generic 
+        # type parameters:
+        if safeLen(candidate.ast.sons[genericParamsPos]) == n.len-1:
+          s = candidate
+          inc(candidateCount)
+    if candidateCount != 1: return explicitGenericInstError(n)
+  else:
+    assert false
   
   var x: TCandidate
   initCandidate(x, s, n)
   var newInst = generateInstance(c, s, x.bindings, n.info)
   
   markUsed(n, s)
-  result = newSymNode(newInst)
+  result = newSymNode(newInst, n.info)
-  result.info = n.info
 

rod/semexprs.nim

@@ -61,24 +61,22 @@ proc semSym(c: PContext, n: PNode, s: PSym, flags: TExprFlags): PNode =
     if s.typ.kind in ConstAbstractTypes: 
       result = copyTree(s.ast)
       result.typ = s.typ
+      result.info = n.info
     else: 
-      result = newSymNode(s)
+      result = newSymNode(s, n.info)
-    result.info = n.info
   of skMacro: result = semMacroExpr(c, n, s)
   of skTemplate: result = semTemplateExpr(c, n, s)
   of skVar: 
     markUsed(n, s)
     # if a proc accesses a global variable, it is not side effect free:
     if sfGlobal in s.flags: incl(c.p.owner.flags, sfSideEffect)
-    result = newSymNode(s)
+    result = newSymNode(s, n.info)
-    result.info = n.info
   of skGenericParam: 
     if s.ast == nil: InternalError(n.info, "no default for")
     result = semExpr(c, s.ast)
   else: 
     markUsed(n, s)
-    result = newSymNode(s)
+    result = newSymNode(s, n.info)
-    result.info = n.info
 
 proc checkConversionBetweenObjects(info: TLineInfo, castDest, src: PType) = 
   var diff = inheritanceDiff(castDest, src)
@@ -645,23 +643,22 @@ proc builtinFieldAccess(c: PContext, n: PNode, flags: TExprFlags): PNode =
   var ty = n.sons[0].Typ
   var f: PSym = nil
   result = nil
-  if ty.kind == tyEnum: 
+  if isTypeExpr(n.sons[0]):
-    # look up if the identifier belongs to the enum:
+    if ty.kind == tyEnum: 
-    while ty != nil: 
+      # look up if the identifier belongs to the enum:
-      f = getSymFromList(ty.n, i)
+      while ty != nil: 
-      if f != nil: break 
+        f = getSymFromList(ty.n, i)
-      ty = ty.sons[0]         # enum inheritance
+        if f != nil: break 
-    if f != nil: 
+        ty = ty.sons[0]         # enum inheritance
-      result = newSymNode(f)
+      if f != nil: 
-      result.info = n.info
+        result = newSymNode(f)
-      result.typ = ty
+        result.info = n.info
-      markUsed(n, f)
+        result.typ = ty
-    else: 
+        markUsed(n, f)
-      GlobalError(n.sons[1].info, errEnumHasNoValueX, i.s)
+        return 
-    return 
+    elif efAllowType notin flags: 
-  elif not (efAllowType in flags) and isTypeExpr(n.sons[0]): 
+      GlobalError(n.sons[0].info, errATypeHasNoValue)
-    GlobalError(n.sons[0].info, errATypeHasNoValue)
+      return
-    return 
   ty = skipTypes(ty, {tyGenericInst, tyVar, tyPtr, tyRef})
   var check: PNode = nil
   if ty.kind == tyObject: 
@@ -1027,7 +1024,8 @@ proc semExpr(c: PContext, n: PNode, flags: TExprFlags = {}): PNode =
     var s = qualifiedLookup(c, n.sons[0], {checkUndeclared})
     if s != nil and s.kind in {skProc, skMethod, skConverter, skIterator}: 
       # type parameters: partial generic specialization
-      result = explictGenericInstantiation(c, n, s)
+      n.sons[0] = semSym(c, n.sons[0], s, flags)
+      result = explicitGenericInstantiation(c, n, s)
     else: 
       result = semArrayAccess(c, n, flags)
   of nkPragmaExpr: 

rod/semgnrc.nim

@@ -40,13 +40,15 @@ proc semGenericStmtSymbol(c: PContext, n: PNode, s: PSym): PNode =
   of skMacro: 
     result = semMacroExpr(c, n, s, false)
   of skGenericParam: 
-    result = newSymNode(s)
+    result = newSymNode(s, n.info)
   of skParam: 
     result = n
   of skType: 
-    if (s.typ != nil) and (s.typ.kind != tyGenericParam): result = newSymNode(s)
+    if (s.typ != nil) and (s.typ.kind != tyGenericParam): 
-    else: result = n
+      result = newSymNode(s, n.info)
-  else: result = newSymNode(s)
+    else: 
+      result = n
+  else: result = newSymNode(s, n.info)
   
 proc getIdentNode(n: PNode): PNode = 
   case n.kind
@@ -101,12 +103,12 @@ proc semGenericStmt(c: PContext, n: PNode, flags: TSemGenericFlags = {}): PNode
       of skProc, skMethod, skIterator, skConverter: 
         n.sons[0] = symChoice(c, n.sons[0], s)
       of skGenericParam: 
-        n.sons[0] = newSymNode(s)
+        n.sons[0] = newSymNode(s, n.sons[0].info)
       of skType: 
         # bad hack for generics:
         if (s.typ != nil) and (s.typ.kind != tyGenericParam): 
-          n.sons[0] = newSymNode(s)
+          n.sons[0] = newSymNode(s, n.sons[0].info)
-      else: n.sons[0] = newSymNode(s)
+      else: n.sons[0] = newSymNode(s, n.sons[0].info)
     for i in countup(1, sonsLen(n) - 1): 
       n.sons[i] = semGenericStmt(c, n.sons[i], flags)
   of nkMacroStmt: 
@@ -221,9 +223,9 @@ proc semGenericStmt(c: PContext, n: PNode, flags: TSemGenericFlags = {}): PNode
       if (a.kind != nkIdentDefs): IllFormedAst(a)
       checkMinSonsLen(a, 3)
       L = sonsLen(a)
-      a.sons[L - 1] = semGenericStmt(c, a.sons[L - 2], {withinTypeDesc})
+      a.sons[L-2] = semGenericStmt(c, a.sons[L-2], {withinTypeDesc})
-      a.sons[L - 1] = semGenericStmt(c, a.sons[L - 1])
+      a.sons[L-1] = semGenericStmt(c, a.sons[L-1])
-      for j in countup(0, L - 3): 
+      for j in countup(0, L-3): 
         addDecl(c, newSymS(skUnknown, getIdentNode(a.sons[j]), c))
   of nkProcDef, nkMethodDef, nkConverterDef, nkMacroDef, nkTemplateDef, 
      nkIteratorDef, nkLambda: 

rod/semstmts.nim

@@ -522,11 +522,15 @@ proc SemTypeSection(c: PContext, n: PNode): PNode =
         InternalError(a.info, "semTypeSection: containerID")
       s.typ.containerID = getID()
       a.sons[1] = semGenericParamList(c, a.sons[1], s.typ)
-      addSon(s.typ, nil)      # to be filled out later
+      
+      # we fill it out later. For magic generics like 'seq', it won't be filled
+      # so we use tyEmpty instead of nil to not crash for strange conversions
+      # like: mydata.seq 
+      addSon(s.typ, newTypeS(tyEmpty, c))
       s.ast = a
       body = semTypeNode(c, a.sons[2], nil)
       if body != nil: body.sym = s
-      s.typ.sons[sonsLen(s.typ) - 1] = body #debug(s.typ);
+      s.typ.sons[sonsLen(s.typ) - 1] = body
       popOwner()
       closeScope(c.tab)
     elif a.sons[2].kind != nkEmpty: 

rod/semtypes.nim

@@ -488,7 +488,8 @@ proc paramType(c: PContext, n, genericParams: PNode, cl: var TIntSet): PType =
     result = addTypeVarsOfGenericBody(c, result, genericParams, cl)
     #if result.kind == tyGenericInvokation: debug(result)
   
-proc semProcTypeNode(c: PContext, n, genericParams: PNode, prev: PType): PType = 
+proc semProcTypeNode(c: PContext, n, genericParams: PNode, 
+                     prev: PType): PType = 
   var 
     def, res: PNode
     typ: PType
@@ -515,12 +516,12 @@ proc semProcTypeNode(c: PContext, n, genericParams: PNode, prev: PType): PType =
     if a.sons[length - 1].kind != nkEmpty: 
       def = semExprWithType(c, a.sons[length - 1]) 
       # check type compability between def.typ and typ:
-      if typ != nil: 
+      if typ == nil: 
-        if cmpTypes(typ, def.typ) < isConvertible: 
-          typeMismatch(a.sons[length - 1], typ, def.typ)
-        def = fitNode(c, typ, def)
-      else: 
         typ = def.typ
+      elif def != nil and def.typ != nil and def.typ.kind != tyNone:
+        # example code that triggers it:
+        # proc sort[T](cmp: proc(a, b: T): int = cmp)
+        def = fitNode(c, typ, def)
     else: 
       def = ast.emptyNode
     for j in countup(0, length - 3): 

rod/sigmatch.nim

@@ -130,7 +130,10 @@ proc concreteType(mapping: TIdTable, t: PType): PType =
     result = t
     while true: 
       result = PType(idTableGet(mapping, t))
-      if result == nil: InternalError("lookup failed")
+      if result == nil:
+        break # it's ok, no match
+        # example code that triggers it:
+        # proc sort[T](cmp: proc(a, b: T): int = cmp)
       if result.kind != tyGenericParam: break
   else: 
     result = t                # Note: empty is valid here

rod/suggest.nim

@@ -123,7 +123,7 @@ proc suggestFieldAccess(c: PContext, n: PNode) =
     else:
       # fallback:
       suggestEverything(c, n)
-  elif typ.kind == tyEnum: 
+  elif typ.kind == tyEnum and n.kind == nkSym and n.sym.kind == skType: 
     # look up if the identifier belongs to the enum:
     var t = typ
     while t != nil: 

tests/accept/compile/tsortdev.nim

@@ -0,0 +1,69 @@
+
+import math, algorithm
+
+proc sorted[T](a: openArray[T], order: TSortOrder): bool = 
+  result = true
+  for i in 0 .. < a.high:
+    if cmp(a[i], a[i+1]) * order > 0: 
+      echo "Out of order: ", a[i], " ", a[i+1]
+      result = false
+
+proc bubbleSort[T](a: var openArray[T], 
+                   cmp: proc (x, y: T): int = cmp,
+                   order = TSortOrder.Ascending) =
+  while true:
+    var sorted = true
+    for i in 0 .. a.len-2:
+      if cmp(a[i], a[i+1]) * order > 0:
+        swap(a[i], a[i+1])
+        sorted = false
+    if sorted: break
+
+when isMainModule:
+  proc main() =
+    const order = Ascending
+    var data: seq[string] = @[]
+
+    for i in 0..10_000: 
+      var L = random(59)
+      setLen(data, L)
+      for j in 0 .. L-1: 
+        data[j] = $(math.random(90) - 10)
+      var copy = data
+      sort(data, cmp, order)
+      if not sorted(data, order):
+        #for x in items(data): echo x
+        break
+      else:
+        echo "SUCCESS!"
+      bubblesort(copy, cmp, order)
+      if copy.len != data.len: 
+        quit "lengths differ!"
+      for i in 0 .. copy.high:
+        if copy[i] != data[i]:
+          quit "algorithms differ!"
+
+    for i in 0..10_000: 
+      var data: seq[int] = @[]
+      var L = random(59)
+      setLen(data, L)
+      for j in 0 .. L-1: 
+        data[j] = (math.random(90) - 10)
+      var copy = data
+      sort(data, cmp[int, int], order)
+      if not sorted(data, order):
+        #for x in items(data): echo x
+        break
+      else:
+        echo "SUCCESS!"
+      bubblesort(copy)
+      if copy.len != data.len: 
+        quit "lengths differ!"
+      for i in 0 .. copy.high:
+        if copy[i] != data[i]:
+          quit "algorithms differ!"
+
+  main()
+
+echo "done"
+

tests/accept/run/tgenerics1.nim

@@ -4,51 +4,47 @@ discard """
 
 # A min-heap.
 type
-    TNode[T] = tuple[priority: int, data: T]
+  TNode[T] = tuple[priority: int, data: T]
 
-    TBinHeap[T] = object
+  TBinHeap[T] = object
-        heap: seq[TNode[T]]
+    heap: seq[TNode[T]]
-        last: int
+    last: int
   
-    PBinHeap[T] = ref TBinHeap[T]
+  PBinHeap[T] = ref TBinHeap[T]
 
 proc newBinHeap*[T](heap: var PBinHeap[T], size: int) =
-    new(heap)
+  new(heap)
-    heap.last = 0
+  heap.last = 0
-    newSeq(heap.seq, size)
+  newSeq(heap.heap, size)
+  #newSeq(heap.seq, size)
  
-when false:
+proc parent(elem: int): int {.inline.} =
+  return (elem-1) div 2
 
-  proc parent(elem: int): int =
+proc siftUp[T](heap: PBinHeap[T], elem: int) =
-      return (elem-1) div 2
+  var idx = elem
+  while idx != 0:
+    var p = parent(idx)
+    if heap.heap[idx].priority < heap.heap[p].priority:
+      swap(heap.heap[idx], heap.heap[p])
+      idx = p
+    else:
+      break
 
-  proc siftUp[T](heap: PBinHeap[T], elem: int) =
+proc add*[T](heap: PBinHeap[T], priority: int, data: T) =
-      var idx = elem
+  var node: TNode[T]
-      while idx != 0:
+  node.priority = priority
-          var p = parent(idx)
+  node.data = data
-          if heap.heap[idx] < heap.heap[p]:
+  heap.heap[heap.last] = node
-              var tmp = heap.heap[idx]
+  siftUp(heap, heap.last)
-              heap.heap[idx] = heap.heap[p]
+  inc(heap.last)
-              heap.heap[p] = tmp
-              idx = p
-          else:
-              break
 
-  proc add*[T](heap: PBinHeap[T], priority: int, data: T) =
+proc print*[T](heap: PBinHeap[T]) =
-      var node: TNode
+  for i in countup(0, heap.last):
-      new(node)
+    echo heap.heap[i].data
-      node.priority = int
-      node.data = data
-      heap.heap[heap.last] = node
-      siftUp(heap, heap.last)
-      inc(heap.last)
-
-  proc print*[T](heap: PBinHeap[T]) =
-      for i in countup(0, heap.last):
-          echo($heap.heap[i])
 
 var
-    heap: PBinHeap[int]
+  heap: PBinHeap[int]
 
 newBinHeap(heap, 256)
 add(heap, 1, 100)

tests/gc/gcleak.nim

@@ -0,0 +1,11 @@
+type
+  TTestObj = object of TObject
+    x: string
+
+proc MakeObj(): TTestObj =
+  result.x = "Hello"
+
+while true:
+  var obj = MakeObj()
+
+

tests/reject/ttypenoval.nim

@@ -0,0 +1,55 @@
+discard """
+  file: "tambsym.nim"
+  line: 36
+  errormsg: "a type has no value"
+"""
+
+# A min-heap.
+type
+  TNode[T] = tuple[priority: int, data: T]
+
+  TBinHeap[T] = object
+    heap: seq[TNode[T]]
+    last: int
+  
+  PBinHeap[T] = ref TBinHeap[T]
+
+proc newBinHeap*[T](heap: var PBinHeap[T], size: int) =
+  new(heap)
+  heap.last = 0
+  newSeq(heap.heap, size)
+  #newSeq(heap.seq, size)
+ 
+proc parent(elem: int): int {.inline.} =
+  return (elem-1) div 2
+
+proc siftUp[T](heap: PBinHeap[T], elem: int) =
+  var idx = elem
+  while idx != 0:
+    var p = parent(idx)
+    if heap.heap[idx] < heap.heap[p]:
+      swap(heap.heap[idx], heap.heap[p])
+      idx = p
+    else:
+      break
+
+proc add*[T](heap: PBinHeap[T], priority: int, data: T) =
+  var node: TNode[T]
+  node.priority = int
+  node.data = data
+  heap.heap[heap.last] = node
+  siftUp(heap, heap.last)
+  inc(heap.last)
+
+proc print*[T](heap: PBinHeap[T]) =
+  for i in countup(0, heap.last):
+    echo($heap.heap[i])
+
+var
+  heap: PBinHeap[int]
+
+newBinHeap(heap, 256)
+add(heap, 1, 100)
+print(heap)
+
+

todo.txt

@@ -35,7 +35,6 @@ Bugs
 To implement
 ------------
 
-* sort routine
 * hash tables and sets
 * distinct types for array/seq indexes
 * constant sequences