/// <summary>
/// Nested repeaters just get multiplied with each other if they're not
/// too lumpy
/// </summary>
private RegexNode ReduceRep()
{
RegexNode u = this;
RegexNode child;
int type = Type();
int min = M;
int max = N;
for (; ;)
{
if (u.ChildCount() == 0)
{
break;
}
child = u.Child(0);
// multiply reps of the same type only
if (child.Type() != type)
{
int childType = child.Type();
if (!(childType >= Oneloop && childType <= Setloop && type == Loop ||
childType >= Onelazy && childType <= Setlazy && type == Lazyloop))
{
break;
}
}
// child can be too lumpy to blur, e.g., (a {100,105}) {3} or (a {2,})?
// [but things like (a {2,})+ are not too lumpy...]
if (u.M == 0 && child.M > 1 || child.N < child.M * 2)
{
break;
}
u = child;
if (u.M > 0)
{
u.M = min = ((int.MaxValue - 1) / u.M < min) ? int.MaxValue : u.M * min;
}
if (u.N > 0)
{
u.N = max = ((int.MaxValue - 1) / u.N < max) ? int.MaxValue : u.N * max;
}
}
return(min == int.MaxValue ? new RegexNode(Nothing, Options) : u);
}
/// <summary>
/// This is a related computation: it takes a RegexTree and computes the
/// leading substring if it see one. It's quite trivial and gives up easily.
/// </summary>
public static RegexPrefix Prefix(RegexTree tree)
{
RegexNode curNode = tree.Root;
RegexNode concatNode = null;
int nextChild = 0;
for (; ;)
{
switch (curNode.NType)
{
case RegexNode.Concatenate:
if (curNode.ChildCount() > 0)
{
concatNode = curNode;
nextChild = 0;
}
break;
case RegexNode.Greedy:
case RegexNode.Capture:
curNode = curNode.Child(0);
concatNode = null;
continue;
case RegexNode.Oneloop:
case RegexNode.Onelazy:
// In release, cutoff at a length to which we can still reasonably construct a string
// In debug, use a smaller cutoff to exercise the cutoff path in tests
const int Cutoff =
#if DEBUG
50;
#else
1_000_000;
#endif
if (curNode.M > 0 && curNode.M < Cutoff)
{
string pref = string.Empty.PadRight(curNode.M, curNode.Ch);
return(new RegexPrefix(pref, 0 != (curNode.Options & RegexOptions.IgnoreCase)));
}
else
{
return(RegexPrefix.Empty);
}
case RegexNode.One:
return(new RegexPrefix(curNode.Ch.ToString(), 0 != (curNode.Options & RegexOptions.IgnoreCase)));
case RegexNode.Multi:
return(new RegexPrefix(curNode.Str, 0 != (curNode.Options & RegexOptions.IgnoreCase)));
case RegexNode.Bol:
case RegexNode.Eol:
case RegexNode.Boundary:
case RegexNode.ECMABoundary:
case RegexNode.Beginning:
case RegexNode.Start:
case RegexNode.EndZ:
case RegexNode.End:
case RegexNode.Empty:
case RegexNode.Require:
case RegexNode.Prevent:
break;
default:
return(RegexPrefix.Empty);
}
if (concatNode == null || nextChild >= concatNode.ChildCount())
{
return(RegexPrefix.Empty);
}
curNode = concatNode.Child(nextChild++);
}
}
/// <summary>
/// The top level RegexCode generator. It does a depth-first walk
/// through the tree and calls EmitFragment to emits code before
/// and after each child of an interior node, and at each leaf.
/// </summary>
public RegexCode RegexCodeFromRegexTree(RegexTree tree)
{
// construct sparse capnum mapping if some numbers are unused
int capsize;
if (tree.CapNumList == null || tree.CapTop == tree.CapNumList.Length)
{
capsize = tree.CapTop;
_caps = null;
}
else
{
capsize = tree.CapNumList.Length;
_caps = tree.Caps;
for (int i = 0; i < tree.CapNumList.Length; i++)
{
_caps[tree.CapNumList[i]] = i;
}
}
RegexNode curNode = tree.Root;
int curChild = 0;
Emit(RegexCode.Lazybranch, 0);
for (; ;)
{
if (curNode.Children == null)
{
EmitFragment(curNode.NType, curNode, 0);
}
else if (curChild < curNode.Children.Count)
{
EmitFragment(curNode.NType | BeforeChild, curNode, curChild);
curNode = curNode.Children[curChild];
_intStack.Push(curChild);
curChild = 0;
continue;
}
if (_intStack.Count == 0)
{
break;
}
curChild = _intStack.Pop();
curNode = curNode.Next;
EmitFragment(curNode.NType | AfterChild, curNode, curChild);
curChild++;
}
PatchJump(0, _emitted.Count);
Emit(RegexCode.Stop);
RegexPrefix?fcPrefix = RegexFCD.FirstChars(tree);
RegexPrefix prefix = RegexFCD.Prefix(tree);
bool rtl = ((tree.Options & RegexOptions.RightToLeft) != 0);
CultureInfo culture = (tree.Options & RegexOptions.CultureInvariant) != 0 ? CultureInfo.InvariantCulture : CultureInfo.CurrentCulture;
RegexBoyerMoore bmPrefix;
if (prefix.Prefix.Length > 0)
{
bmPrefix = new RegexBoyerMoore(prefix.Prefix, prefix.CaseInsensitive, rtl, culture);
}
else
{
bmPrefix = null;
}
int anchors = RegexFCD.Anchors(tree);
int[] emitted = _emitted.ToArray();
return(new RegexCode(emitted, _stringTable, _trackCount, _caps, capsize, bmPrefix, fcPrefix, anchors, rtl));
}
/// <summary>
/// FC computation and shortcut cases for each node type
/// </summary>
private void CalculateFC(int NodeType, RegexNode node, int CurIndex)
{
bool ci = false;
bool rtl = false;
if (NodeType <= RegexNode.Ref)
{
if ((node.Options & RegexOptions.IgnoreCase) != 0)
{
ci = true;
}
if ((node.Options & RegexOptions.RightToLeft) != 0)
{
rtl = true;
}
}
switch (NodeType)
{
case RegexNode.Concatenate | BeforeChild:
case RegexNode.Alternate | BeforeChild:
case RegexNode.Testref | BeforeChild:
case RegexNode.Loop | BeforeChild:
case RegexNode.Lazyloop | BeforeChild:
break;
case RegexNode.Testgroup | BeforeChild:
if (CurIndex == 0)
{
SkipChild();
}
break;
case RegexNode.Empty:
PushFC(new RegexFC(true));
break;
case RegexNode.Concatenate | AfterChild:
if (CurIndex != 0)
{
RegexFC child = PopFC();
RegexFC cumul = TopFC();
_failed = !cumul.AddFC(child, true);
}
if (!TopFC()._nullable)
{
_skipAllChildren = true;
}
break;
case RegexNode.Testgroup | AfterChild:
if (CurIndex > 1)
{
RegexFC child = PopFC();
RegexFC cumul = TopFC();
_failed = !cumul.AddFC(child, false);
}
break;
case RegexNode.Alternate | AfterChild:
case RegexNode.Testref | AfterChild:
if (CurIndex != 0)
{
RegexFC child = PopFC();
RegexFC cumul = TopFC();
_failed = !cumul.AddFC(child, false);
}
break;
case RegexNode.Loop | AfterChild:
case RegexNode.Lazyloop | AfterChild:
if (node.M == 0)
{
TopFC()._nullable = true;
}
break;
case RegexNode.Group | BeforeChild:
case RegexNode.Group | AfterChild:
case RegexNode.Capture | BeforeChild:
case RegexNode.Capture | AfterChild:
case RegexNode.Greedy | BeforeChild:
case RegexNode.Greedy | AfterChild:
break;
case RegexNode.Require | BeforeChild:
case RegexNode.Prevent | BeforeChild:
SkipChild();
PushFC(new RegexFC(true));
break;
case RegexNode.Require | AfterChild:
case RegexNode.Prevent | AfterChild:
break;
case RegexNode.One:
case RegexNode.Notone:
PushFC(new RegexFC(node.Ch, NodeType == RegexNode.Notone, false, ci));
break;
case RegexNode.Oneloop:
case RegexNode.Onelazy:
PushFC(new RegexFC(node.Ch, false, node.M == 0, ci));
break;
case RegexNode.Notoneloop:
case RegexNode.Notonelazy:
PushFC(new RegexFC(node.Ch, true, node.M == 0, ci));
break;
case RegexNode.Multi:
if (node.Str.Length == 0)
{
PushFC(new RegexFC(true));
}
else if (!rtl)
{
PushFC(new RegexFC(node.Str[0], false, false, ci));
}
else
{
PushFC(new RegexFC(node.Str[node.Str.Length - 1], false, false, ci));
}
break;
case RegexNode.Set:
PushFC(new RegexFC(node.Str, false, ci));
break;
case RegexNode.Setloop:
case RegexNode.Setlazy:
PushFC(new RegexFC(node.Str, node.M == 0, ci));
break;
case RegexNode.Ref:
PushFC(new RegexFC(RegexCharClass.AnyClass, true, false));
break;
case RegexNode.Nothing:
case RegexNode.Bol:
case RegexNode.Eol:
case RegexNode.Boundary:
case RegexNode.Nonboundary:
case RegexNode.ECMABoundary:
case RegexNode.NonECMABoundary:
case RegexNode.Beginning:
case RegexNode.Start:
case RegexNode.EndZ:
case RegexNode.End:
PushFC(new RegexFC(true));
break;
default:
throw new ArgumentException($"Unexpected opcode in regular expression generation: {NodeType.ToString(CultureInfo.CurrentCulture)}.");
}
}
/// <summary>
/// The main RegexCode generator. It does a depth-first walk
/// through the tree and calls EmitFragment to emits code before
/// and after each child of an interior node, and at each leaf.
/// </summary>
private void EmitFragment(int nodetype, RegexNode node, int curIndex)
{
int bits = 0;
if (nodetype <= RegexNode.Ref)
{
if (node.UseOptionR())
{
bits |= RegexCode.Rtl;
}
if ((node.Options & RegexOptions.IgnoreCase) != 0)
{
bits |= RegexCode.Ci;
}
}
switch (nodetype)
{
case RegexNode.Concatenate | BeforeChild:
case RegexNode.Concatenate | AfterChild:
case RegexNode.Empty:
break;
case RegexNode.Alternate | BeforeChild:
if (curIndex < node.Children.Count - 1)
{
_intStack.Push(_emitted.Count);
Emit(RegexCode.Lazybranch, 0);
}
break;
case RegexNode.Alternate | AfterChild:
{
if (curIndex < node.Children.Count - 1)
{
int LBPos = _intStack.Pop();
_intStack.Push(_emitted.Count);
Emit(RegexCode.Goto, 0);
PatchJump(LBPos, _emitted.Count);
}
else
{
int I;
for (I = 0; I < curIndex; I++)
{
PatchJump(_intStack.Pop(), _emitted.Count);
}
}
break;
}
case RegexNode.Testref | BeforeChild:
switch (curIndex)
{
case 0:
Emit(RegexCode.Setjump);
_intStack.Push(_emitted.Count);
Emit(RegexCode.Lazybranch, 0);
Emit(RegexCode.Testref, MapCapnum(node.M));
Emit(RegexCode.Forejump);
break;
}
break;
case RegexNode.Testref | AfterChild:
switch (curIndex)
{
case 0:
{
int Branchpos = _intStack.Pop();
_intStack.Push(_emitted.Count);
Emit(RegexCode.Goto, 0);
PatchJump(Branchpos, _emitted.Count);
Emit(RegexCode.Forejump);
if (node.Children.Count > 1)
{
break;
}
// else fallthrough
goto case 1;
}
case 1:
PatchJump(_intStack.Pop(), _emitted.Count);
break;
}
break;
case RegexNode.Testgroup | BeforeChild:
switch (curIndex)
{
case 0:
Emit(RegexCode.Setjump);
Emit(RegexCode.Setmark);
_intStack.Push(_emitted.Count);
Emit(RegexCode.Lazybranch, 0);
break;
}
break;
case RegexNode.Testgroup | AfterChild:
switch (curIndex)
{
case 0:
Emit(RegexCode.Getmark);
Emit(RegexCode.Forejump);
break;
case 1:
int Branchpos = _intStack.Pop();
_intStack.Push(_emitted.Count);
Emit(RegexCode.Goto, 0);
PatchJump(Branchpos, _emitted.Count);
Emit(RegexCode.Getmark);
Emit(RegexCode.Forejump);
if (node.Children.Count > 2)
{
break;
}
// else fallthrough
goto case 2;
case 2:
PatchJump(_intStack.Pop(), _emitted.Count);
break;
}
break;
case RegexNode.Loop | BeforeChild:
case RegexNode.Lazyloop | BeforeChild:
if (node.N < int.MaxValue || node.M > 1)
{
Emit(node.M == 0 ? RegexCode.Nullcount : RegexCode.Setcount, node.M == 0 ? 0 : 1 - node.M);
}
else
{
Emit(node.M == 0 ? RegexCode.Nullmark : RegexCode.Setmark);
}
if (node.M == 0)
{
_intStack.Push(_emitted.Count);
Emit(RegexCode.Goto, 0);
}
_intStack.Push(_emitted.Count);
break;
case RegexNode.Loop | AfterChild:
case RegexNode.Lazyloop | AfterChild:
{
int StartJumpPos = _emitted.Count;
int Lazy = (nodetype - (RegexNode.Loop | AfterChild));
if (node.N < int.MaxValue || node.M > 1)
{
Emit(RegexCode.Branchcount + Lazy, _intStack.Pop(), node.N == int.MaxValue ? int.MaxValue : node.N - node.M);
}
else
{
Emit(RegexCode.Branchmark + Lazy, _intStack.Pop());
}
if (node.M == 0)
{
PatchJump(_intStack.Pop(), StartJumpPos);
}
}
break;
case RegexNode.Group | BeforeChild:
case RegexNode.Group | AfterChild:
break;
case RegexNode.Capture | BeforeChild:
Emit(RegexCode.Setmark);
break;
case RegexNode.Capture | AfterChild:
Emit(RegexCode.Capturemark, MapCapnum(node.M), MapCapnum(node.N));
break;
case RegexNode.Require | BeforeChild:
// NOTE: the following line causes lookahead/lookbehind to be
// NON-BACKTRACKING. It can be commented out with (*)
Emit(RegexCode.Setjump);
Emit(RegexCode.Setmark);
break;
case RegexNode.Require | AfterChild:
Emit(RegexCode.Getmark);
// NOTE: the following line causes lookahead/lookbehind to be
// NON-BACKTRACKING. It can be commented out with (*)
Emit(RegexCode.Forejump);
break;
case RegexNode.Prevent | BeforeChild:
Emit(RegexCode.Setjump);
_intStack.Push(_emitted.Count);
Emit(RegexCode.Lazybranch, 0);
break;
case RegexNode.Prevent | AfterChild:
Emit(RegexCode.Backjump);
PatchJump(_intStack.Pop(), _emitted.Count);
Emit(RegexCode.Forejump);
break;
case RegexNode.Greedy | BeforeChild:
Emit(RegexCode.Setjump);
break;
case RegexNode.Greedy | AfterChild:
Emit(RegexCode.Forejump);
break;
case RegexNode.One:
case RegexNode.Notone:
Emit(node.NType | bits, node.Ch);
break;
case RegexNode.Notoneloop:
case RegexNode.Notonelazy:
case RegexNode.Oneloop:
case RegexNode.Onelazy:
if (node.M > 0)
{
Emit(((node.NType == RegexNode.Oneloop || node.NType == RegexNode.Onelazy) ?
RegexCode.Onerep : RegexCode.Notonerep) | bits, node.Ch, node.M);
}
if (node.N > node.M)
{
Emit(node.NType | bits, node.Ch, node.N == int.MaxValue ?
int.MaxValue : node.N - node.M);
}
break;
case RegexNode.Setloop:
case RegexNode.Setlazy:
if (node.M > 0)
{
Emit(RegexCode.Setrep | bits, StringCode(node.Str), node.M);
}
if (node.N > node.M)
{
Emit(node.NType | bits, StringCode(node.Str),
(node.N == int.MaxValue) ? int.MaxValue : node.N - node.M);
}
break;
case RegexNode.Multi:
Emit(node.NType | bits, StringCode(node.Str));
break;
case RegexNode.Set:
Emit(node.NType | bits, StringCode(node.Str));
break;
case RegexNode.Ref:
Emit(node.NType | bits, MapCapnum(node.M));
break;
case RegexNode.Nothing:
case RegexNode.Bol:
case RegexNode.Eol:
case RegexNode.Boundary:
case RegexNode.Nonboundary:
case RegexNode.ECMABoundary:
case RegexNode.NonECMABoundary:
case RegexNode.Beginning:
case RegexNode.Start:
case RegexNode.EndZ:
case RegexNode.End:
Emit(node.NType);
break;
default:
throw new ArgumentException($"Unexpected opcode in regular expression generation: {nodetype.ToString()}.");
}
}
