implemented 'union' and 'unchecked' pragmas

compiler/ast.nim

@@ -488,6 +488,8 @@ const
   routineKinds* = {skProc, skMethod, skIterator, skConverter,
     skMacro, skTemplate}
   tfIncompleteStruct* = tfVarargs
+  tfUncheckedArray* = tfVarargs
+  tfUnion* = tfNoSideEffect
   skError* = skUnknown
   
   # type flags that are essential for type equality:

compiler/ccgexprs.nim

@@ -736,7 +736,7 @@ proc genArrayElem(p: BProc, e: PNode, d: var TLoc) =
   var ty = skipTypes(skipTypes(a.t, abstractVarRange), abstractPtrs)
   var first = intLiteral(firstOrd(ty))
   # emit range check:
-  if (optBoundsCheck in p.options):
+  if optBoundsCheck in p.options and tfUncheckedArray notin ty.flags:
     if not isConstExpr(e.sons[1]):
       # semantic pass has already checked for const index expressions
       if firstOrd(ty) == 0:

compiler/ccgtypes.nim

@@ -376,10 +376,13 @@ proc getTypePre(m: BModule, typ: PType): PRope =
   else: 
     result = getSimpleTypeDesc(m, typ)
     if result == nil: result = cacheGetType(m.typeCache, typ)
-  
+
+proc structOrUnion(t: PType): PRope =
+  (if tfUnion in t.flags: toRope("union") else: toRope("struct"))
+
 proc getForwardStructFormat(): string = 
-  if gCmd == cmdCompileToCpp: result = "struct $1;$n"
-  else: result = "typedef struct $1 $1;$n"
+  if gCmd == cmdCompileToCpp: result = "$1 $2;$n"
+  else: result = "typedef $1 $2 $2;$n"
   
 proc getTypeForward(m: BModule, typ: PType): PRope = 
   result = cacheGetType(m.forwTypeCache, typ)
@@ -390,7 +393,8 @@ proc getTypeForward(m: BModule, typ: PType): PRope =
   of tySequence, tyTuple, tyObject: 
     result = getTypeName(typ)
     if not isImportedType(typ): 
-      appf(m.s[cfsForwardTypes], getForwardStructFormat(), [result])
+      appf(m.s[cfsForwardTypes], getForwardStructFormat(),
+          [structOrUnion(typ), result])
     idTablePut(m.forwTypeCache, typ, result)
   else: internalError("getTypeForward(" & $typ.kind & ')')
   
@@ -445,7 +449,12 @@ proc genRecordFieldsAux(m: BModule, n: PNode,
     if accessExpr != nil: ae = ropef("$1.$2", [accessExpr, sname])
     else: ae = sname
     fillLoc(field.loc, locField, field.typ, ae, OnUnknown)
-    appf(result, "$1 $2;$n", [getTypeDescAux(m, field.loc.t, check), sname])
+    let fieldType = field.loc.t
+    if fieldType.kind == tyArray and tfUncheckedArray in fieldType.flags:
+      appf(result, "$1 $2[SEQ_DECL_SIZE];$n",
+          [getTypeDescAux(m, fieldType.elemType, check), sname])
+    else:
+      appf(result, "$1 $2;$n", [getTypeDescAux(m, fieldType, check), sname])
   else: internalError(n.info, "genRecordFieldsAux()")
   
 proc getRecordFields(m: BModule, typ: PType, check: var TIntSet): PRope = 
@@ -455,23 +464,24 @@ proc getRecordDesc(m: BModule, typ: PType, name: PRope,
                    check: var TIntSet): PRope = 
   # declare the record:
   var hasField = false
+  let aStruct = structOrUnion(typ)
   if typ.kind == tyObject: 
     if typ.sons[0] == nil: 
       if (typ.sym != nil and sfPure in typ.sym.flags) or tfFinal in typ.flags: 
-        result = ropecg(m, "struct $1 {$n", [name])
+        result = ropecg(m, "$1 $2 {$n", [aStruct, name])
       else: 
-        result = ropecg(m, "struct $1 {$n#TNimType* m_type;$n", [name])
+        result = ropecg(m, "$1 $2 {$n#TNimType* m_type;$n", [aStruct, name])
         hasField = true
     elif gCmd == cmdCompileToCpp: 
-      result = ropecg(m, "struct $1 : public $2 {$n", 
-                      [name, getTypeDescAux(m, typ.sons[0], check)])
+      result = ropecg(m, "$1 $2 : public $3 {$n", 
+                      [aStruct, name, getTypeDescAux(m, typ.sons[0], check)])
       hasField = true
     else: 
-      result = ropecg(m, "struct $1 {$n  $2 Sup;$n", 
-                      [name, getTypeDescAux(m, typ.sons[0], check)])
+      result = ropecg(m, "$1 $2 {$n  $3 Sup;$n", 
+                      [aStruct, name, getTypeDescAux(m, typ.sons[0], check)])
       hasField = true
   else: 
-    result = ropef("struct $1 {$n", [name])
+    result = ropef("$1 $2 {$n", [aStruct, name])
   var desc = getRecordFields(m, typ, check)
   if (desc == nil) and not hasField: 
     appf(result, "char dummy;$n", [])
@@ -480,8 +490,8 @@ proc getRecordDesc(m: BModule, typ: PType, name: PRope,
   app(result, "};" & tnl)
 
 proc getTupleDesc(m: BModule, typ: PType, name: PRope, 
-                  check: var TIntSet): PRope = 
-  result = ropef("struct $1 {$n", [name])
+                  check: var TIntSet): PRope =
+  result = ropef("$1 $2 {$n", [structOrUnion(typ), name])
   var desc: PRope = nil
   for i in countup(0, sonsLen(typ) - 1): 
     appf(desc, "$1 Field$2;$n", 
@@ -557,7 +567,8 @@ proc getTypeDescAux(m: BModule, typ: PType, check: var TIntSet): PRope =
     if result == nil: 
       result = getTypeName(t)
       if not isImportedType(t): 
-        appf(m.s[cfsForwardTypes], getForwardStructFormat(), [result])
+        appf(m.s[cfsForwardTypes], getForwardStructFormat(),
+            [structOrUnion(t), result])
       idTablePut(m.forwTypeCache, t, result)
     assert(cacheGetType(m.typeCache, t) == nil)
     idTablePut(m.typeCache, t, con(result, "*"))
@@ -588,7 +599,8 @@ proc getTypeDescAux(m: BModule, typ: PType, check: var TIntSet): PRope =
     if result == nil: 
       result = getTypeName(t)
       if not isImportedType(t): 
-        appf(m.s[cfsForwardTypes], getForwardStructFormat(), [result])
+        appf(m.s[cfsForwardTypes], getForwardStructFormat(),
+           [structOrUnion(t), result])
       idTablePut(m.forwTypeCache, t, result)
     idTablePut(m.typeCache, t, result) # always call for sideeffects:
     if t.kind != tyTuple: recdesc = getRecordDesc(m, t, result, check)

compiler/condsyms.nim

@@ -47,6 +47,7 @@ proc initDefines*() =
   defineSymbol("nimeffects")
   defineSymbol("nimbabel")
   defineSymbol("nimcomputedgoto")
+  defineSymbol("nimunion")
   
   # add platform specific symbols:
   case targetCPU

compiler/pragmas.nim

@@ -50,9 +50,9 @@ const
     wDeprecated, wExtern, wThread, wImportCpp, wImportObjC, wAsmNoStackFrame,
     wRaises, wTags}
   typePragmas* = {wImportc, wExportc, wDeprecated, wMagic, wAcyclic, wNodecl, 
-    wPure, wHeader, wCompilerproc, wFinal, wSize, wExtern, wShallow, 
+    wPure, wHeader, wCompilerproc, wFinal, wSize, wExtern, wShallow,
     wImportCpp, wImportObjC, wError, wIncompleteStruct, wByCopy, wByRef,
-    wInheritable, wGensym, wInject, wRequiresInit}
+    wInheritable, wGensym, wInject, wRequiresInit, wUnchecked, wUnion}
   fieldPragmas* = {wImportc, wExportc, wDeprecated, wExtern, 
     wImportCpp, wImportObjC, wError}
   varPragmas* = {wImportc, wExportc, wVolatile, wRegister, wThreadVar, wNodecl, 
@@ -718,6 +718,14 @@ proc singlePragma(c: PContext, sym: PSym, n: PNode, i: int,
           noVal(it)
           if sym.typ == nil: invalidPragma(it)
           else: incl(sym.typ.flags, tfIncompleteStruct)
+        of wUnchecked:
+          noVal(it)
+          if sym.typ == nil: invalidPragma(it)
+          else: incl(sym.typ.flags, tfUncheckedArray)
+        of wUnion:
+          noVal(it)
+          if sym.typ == nil: invalidPragma(it)
+          else: incl(sym.typ.flags, tfUnion)
         of wRequiresInit:
           noVal(it)
           if sym.typ == nil: invalidPragma(it)

compiler/wordrecg.nim

@@ -63,7 +63,7 @@ type
     wAcyclic, wShallow, wUnroll, wLinearScanEnd, wComputedGoto, wInjectStmt,
     wWrite, wGensym, wInject, wDirty, wInheritable, wThreadVar, wEmit, 
     wAsmNoStackFrame,
-    wImplicitStatic, wGlobal, wCodegenDecl,
+    wImplicitStatic, wGlobal, wCodegenDecl, wUnchecked,
 
     wAuto, wBool, wCatch, wChar, wClass,
     wConst_cast, wDefault, wDelete, wDouble, wDynamic_cast,
@@ -145,7 +145,7 @@ const
     "subschar", "acyclic", "shallow", "unroll", "linearscanend",
     "computedgoto", "injectstmt",
     "write", "gensym", "inject", "dirty", "inheritable", "threadvar", "emit",
-    "asmnostackframe", "implicitstatic", "global", "codegendecl",
+    "asmnostackframe", "implicitstatic", "global", "codegendecl", "unchecked",
     
     "auto", "bool", "catch", "char", "class",
     "const_cast", "default", "delete", "double",

doc/manual.txt

@@ -5320,6 +5320,53 @@ strings automatically:
   printf("hallo %s", "world") # "world" will be passed as C string
 
 
+Union pragma
+------------
+The `union`:idx: pragma can be applied to any ``object`` type. It means all
+of the object's fields are overlaid in memory. This produces a ``union``
+instead of a ``struct`` in the generated C/C++ code. The object declaration
+then must not use inheritance or any GC'ed memory but this is currently not
+checked.
+
+**Future directions**: GC'ed memory should be allowed in unions and the GC
+should scan unions conservatively.
+
+
+Unchecked pragma
+----------------
+The `unchecked`:idx: pragma can be used to mark a named array as ``unchecked``
+meaning its bounds are not checked. This is often useful when one wishes to
+implement his own flexibly sized arrays. Additionally an unchecked array is
+translated into a C array of undetermined size:
+
+.. code-block:: nimrod
+  type
+    ArrayPart{.unchecked.} = array[0..0, int]
+    MySeq = object
+      len, cap: int
+      data: ArrayPart
+
+Produces roughly this C code:
+
+.. code-block:: C
+  typedef struct {
+    NI len;
+    NI cap;
+    NI data[];
+  } MySeq;
+
+The bounds checking done at compile time is not disabled for now, so to access
+``s.data[C]`` (where ``C`` is a constant) the array's index needs needs to
+include ``C``.
+
+The base type of the unchecked array may not contain any GC'ed memory but this
+is currently not checked.
+
+**Future directions**: GC'ed memory should be allowed in unchecked arrays and
+there should be an explicit annotation of how the GC is to determine the
+runtime size of the array.
+
+
 Dynlib pragma for import
 ------------------------
 With the `dynlib`:idx: pragma a procedure or a variable can be imported from

lib/nimbase.h

@@ -285,8 +285,8 @@ static N_INLINE(NI32, float32ToInt32)(float x) {
 
 typedef struct TStringDesc* string;
 
-/* declared size of a sequence: */
-#if defined(__GNUC__)
+/* declared size of a sequence/variable length array: */
+#if defined(__GNUC__) || defined(__clang__) || defined(_MSC_VER)
 #  define SEQ_DECL_SIZE /* empty is correct! */
 #else
 #  define SEQ_DECL_SIZE 1000000

todo.txt

@@ -1,7 +1,6 @@
 version 0.9.4
 =============
 
-- implement unchecked arrays
 - make testament produce full JSON information
 - fix gensym capture bug
 - vm
@@ -28,7 +27,7 @@ version 0.9.x
 =============
 
 - memory manager: add a measure of fragmentation
-- implement 'union' and 'bits' pragmas
+- implement 'bits' pragmas
 - fix closures/lambdalifting
 - ensure (ref T)(a, b) works as a type conversion and type constructor
 - optimize 'genericReset'; 'newException' leads to code bloat