public BVAlgebra(CharSetSolver solver, BDD[] minterms) :
base(new MintermClassifier(solver, minterms), solver.ComputeDomainSizes(minterms), minterms)
{
_mintermGenerator = new MintermGenerator <BV>(this);
False = BV.CreateFalse(_bits);
True = BV.CreateTrue(_bits);
var singleBitVectors = new BV[_bits];
for (int i = 0; i < singleBitVectors.Length; i++)
{
singleBitVectors[i] = BV.CreateSingleBit(_bits, i);
}
_minterms = singleBitVectors;
}
/// <summary>Initializes the factory.</summary>
public SymbolicRegexRunnerFactory(RegexTree regexTree, RegexOptions options, TimeSpan matchTimeout, CultureInfo culture)
{
// RightToLeft and ECMAScript are currently not supported in conjunction with NonBacktracking.
if ((options & (RegexOptions.RightToLeft | RegexOptions.ECMAScript)) != 0)
{
throw new NotSupportedException(
SR.Format(SR.NotSupported_NonBacktrackingConflictingOption,
(options & RegexOptions.RightToLeft) != 0 ? nameof(RegexOptions.RightToLeft) : nameof(RegexOptions.ECMAScript)));
}
var converter = new RegexNodeConverter(culture, regexTree.CaptureNumberSparseMapping);
CharSetSolver solver = CharSetSolver.Instance;
SymbolicRegexNode <BDD> root = converter.ConvertToSymbolicRegexNode(regexTree.Root, tryCreateFixedLengthMarker: true);
BDD[] minterms = root.ComputeMinterms();
if (minterms.Length > 64)
{
// Use BitVector to represent a predicate
var algebra = new BitVectorAlgebra(solver, minterms);
var builder = new SymbolicRegexBuilder <BitVector>(algebra)
{
// The default constructor sets the following predicates to False; this update happens after the fact.
// It depends on whether anchors where used in the regex whether the predicates are actually different from False.
_wordLetterPredicateForAnchors = algebra.ConvertFromCharSet(solver, converter._builder._wordLetterPredicateForAnchors),
_newLinePredicate = algebra.ConvertFromCharSet(solver, converter._builder._newLinePredicate)
};
// Convert the BDD-based AST to BitVector-based AST
SymbolicRegexNode <BitVector> rootNode = converter._builder.Transform(root, builder, bdd => builder._solver.ConvertFromCharSet(solver, bdd));
_matcher = new SymbolicRegexMatcher <BitVector>(rootNode, regexTree, minterms, matchTimeout);
}
else
{
// Use ulong to represent a predicate
var algebra = new BitVector64Algebra(solver, minterms);
var builder = new SymbolicRegexBuilder <ulong>(algebra)
{
// The default constructor sets the following predicates to False, this update happens after the fact
// It depends on whether anchors where used in the regex whether the predicates are actually different from False
_wordLetterPredicateForAnchors = algebra.ConvertFromCharSet(solver, converter._builder._wordLetterPredicateForAnchors),
_newLinePredicate = algebra.ConvertFromCharSet(solver, converter._builder._newLinePredicate)
};
// Convert the BDD-based AST to ulong-based AST
SymbolicRegexNode <ulong> rootNode = converter._builder.Transform(root, builder, bdd => builder._solver.ConvertFromCharSet(solver, bdd));
_matcher = new SymbolicRegexMatcher <ulong>(rootNode, regexTree, minterms, matchTimeout);
}
}
/// <summary>Initializes the factory.</summary>
public SymbolicRegexRunnerFactory(RegexTree regexTree, RegexOptions options, TimeSpan matchTimeout)
{
Debug.Assert((options & (RegexOptions.RightToLeft | RegexOptions.ECMAScript)) == 0);
var charSetSolver = new CharSetSolver();
var bddBuilder = new SymbolicRegexBuilder <BDD>(charSetSolver, charSetSolver);
var converter = new RegexNodeConverter(bddBuilder, regexTree.CaptureNumberSparseMapping);
SymbolicRegexNode <BDD> rootNode = converter.ConvertToSymbolicRegexNode(regexTree.Root);
BDD[] minterms = rootNode.ComputeMinterms();
_matcher = minterms.Length > 64 ?
SymbolicRegexMatcher <BitVector> .Create(regexTree.CaptureCount, regexTree.FindOptimizations, bddBuilder, rootNode, new BitVectorSolver(minterms, charSetSolver), matchTimeout) :
SymbolicRegexMatcher <ulong> .Create(regexTree.CaptureCount, regexTree.FindOptimizations, bddBuilder, rootNode, new UInt64Solver(minterms, charSetSolver), matchTimeout);
}
/// <summary>
/// Assumes that set is a union of some minterms (or empty).
/// If null then null is returned.
/// </summary>
public BitVector ConvertFromBDD(BDD set, CharSetSolver solver)
{
BDD[] partition = _minterms;
BitVector result = Empty;
for (int i = 0; i < partition.Length; i++)
{
if (!solver.IsEmpty(solver.And(partition[i], set)))
{
result = BitVector.Or(result, _mintermVectors[i]);
}
}
return(result);
}
/// <summary>Gets a <see cref="BDD"/> that represents the \w character class.</summary>
/// <remarks>\w is the union of the 8 categories: 0,1,2,3,4,5,8,18</remarks>
public static BDD WordLetter(CharSetSolver solver) =>
s_wordLetter ??
Interlocked.CompareExchange(ref s_wordLetter,
solver.Or(new[]
{
GetCategory(UnicodeCategory.UppercaseLetter),
GetCategory(UnicodeCategory.LowercaseLetter),
GetCategory(UnicodeCategory.TitlecaseLetter),
GetCategory(UnicodeCategory.ModifierLetter),
GetCategory(UnicodeCategory.OtherLetter),
GetCategory(UnicodeCategory.NonSpacingMark),
GetCategory(UnicodeCategory.DecimalDigitNumber),
GetCategory(UnicodeCategory.ConnectorPunctuation),
}),
null) ??
s_wordLetter;
/// <summary>
/// Assumes that set is a union of some minterms (or empty).
/// If null then 0 is returned.
/// </summary>
public ulong ConvertFromBDD(BDD set, CharSetSolver solver)
{
BDD[] partition = _minterms;
ulong result = 0;
for (int i = 0; i < partition.Length; i++)
{
// Set the i'th bit if the i'th minterm is in the set.
if (!solver.IsEmpty(solver.And(partition[i], set)))
{
result |= (ulong)1 << i;
}
}
return(result);
}
public BitVectorSolver(BDD[] minterms, CharSetSolver solver)
{
_minterms = minterms;
_classifier = new MintermClassifier(minterms, solver);
var singleBitVectors = new BitVector[minterms.Length];
for (int i = 0; i < singleBitVectors.Length; i++)
{
singleBitVectors[i] = BitVector.CreateSingleBit(minterms.Length, i);
}
_mintermVectors = singleBitVectors;
Empty = BitVector.CreateFalse(minterms.Length);
Full = BitVector.CreateTrue(minterms.Length);
}
public BitVectorAlgebra(CharSetSolver solver, BDD[] minterms)
{
_minterms = minterms;
_classifier = new MintermClassifier(solver, minterms);
_mintermGenerator = new MintermGenerator <BitVector>(this);
var singleBitVectors = new BitVector[minterms.Length];
for (int i = 0; i < singleBitVectors.Length; i++)
{
singleBitVectors[i] = BitVector.CreateSingleBit(minterms.Length, i);
}
_mintermVectors = singleBitVectors;
False = BitVector.CreateFalse(minterms.Length);
True = BitVector.CreateTrue(minterms.Length);
}
public SymbolicRegexSampler(SymbolicRegexNode <TSet> root, int randomseed, bool negative)
{
_root = negative ? root._builder.Not(root) : root;
// Treat 0 as no seed and instead choose a random seed randomly
RandomSeed = randomseed == 0 ? new Random().Next() : randomseed;
_random = new Random(RandomSeed);
_solver = root._builder._solver;
_charSetSolver = new CharSetSolver();
_asciiWordCharacters = _charSetSolver.Or(new BDD[] {
_charSetSolver.CreateSetFromRange('A', 'Z'),
_charSetSolver.CreateSetFromRange('a', 'z'),
_charSetSolver.CreateFromChar('_'),
_charSetSolver.CreateSetFromRange('0', '9')
});
// Visible ASCII range for input character generation
_ascii = _charSetSolver.CreateSetFromRange('\x20', '\x7E');
_asciiNonWordCharacters = _charSetSolver.And(_ascii, _charSetSolver.Not(_asciiWordCharacters));
}
public SymbolicRegexSampler(SymbolicRegexNode <S> root, int randomseed, bool negative)
{
_root = negative ? root._builder.Not(root) : root;
// Treat 0 as no seed and instead choose a random seed randomly
RandomSeed = randomseed == 0 ? new Random().Next() : randomseed;
_random = new Random(RandomSeed);
_solver = root._builder._solver;
CharSetSolver bddSolver = CharSetSolver.Instance;
_asciiWordCharacters = bddSolver.Or(new BDD[] {
bddSolver.RangeConstraint('A', 'Z'),
bddSolver.RangeConstraint('a', 'z'),
bddSolver.CharConstraint('_'),
bddSolver.RangeConstraint('0', '9')
});
// Visible ASCII range for input character generation
_ascii = bddSolver.RangeConstraint('\x20', '\x7E');
_asciiNonWordCharacters = bddSolver.And(_ascii, bddSolver.Not(_asciiWordCharacters));
}
public BDD ConvertToBDD(BitVector set, CharSetSolver solver)
{
BDD[] partition = _minterms;
// the result will be the union of all minterms in the set
BDD result = solver.Empty;
if (!set.Equals(Empty))
{
for (int i = 0; i < partition.Length; i++)
{
// include the i'th minterm in the union if the i'th bit is set
if (set[i])
{
result = solver.Or(result, partition[i]);
}
}
}
return(result);
}
public BDD ConvertToBDD(ulong set, CharSetSolver solver)
{
BDD[] partition = _minterms;
// the result will be the union of all minterms in the set
BDD result = BDD.False;
if (set != 0)
{
for (int i = 0; i < partition.Length; i++)
{
// include the i'th minterm in the union if the i'th bit is set
if ((set & ((ulong)1 << i)) != 0)
{
result = solver.Or(result, partition[i]);
}
}
}
return(result);
}
/// <summary>Constructs matcher for given symbolic regex.</summary>
internal SymbolicRegexMatcher(SymbolicRegexNode <TSetType> sr, RegexCode code, CharSetSolver css, BDD[] minterms, TimeSpan matchTimeout, CultureInfo culture)
{
Debug.Assert(sr._builder._solver is BV64Algebra or BVAlgebra or CharSetSolver, $"Unsupported algebra: {sr._builder._solver}");
_pattern = sr;
_builder = sr._builder;
_checkTimeout = Regex.InfiniteMatchTimeout != matchTimeout;
_timeout = (int)(matchTimeout.TotalMilliseconds + 0.5); // Round up, so it will be at least 1ms
_partitions = _builder._solver switch
{
BV64Algebra bv64 => bv64._classifier,
BVAlgebra bv => bv._classifier,
_ => new MintermClassifier((CharSetSolver)(object)_builder._solver, minterms),
};
if (code.FindOptimizations.FindMode != FindNextStartingPositionMode.NoSearch &&
code.FindOptimizations.LeadingAnchor == 0) // If there are any anchors, we're better off letting the DFA quickly do its job of determining whether there's a match.
{
_findOpts = code.FindOptimizations;
}
// Determine the number of initial states. If there's no anchor, only the default previous
// character kind 0 is ever going to be used for all initial states.
int statesCount = _pattern._info.ContainsSomeAnchor ? CharKind.CharKindCount : 1;
// Create the initial states for the original pattern.
var initialStates = new DfaMatchingState <TSetType> [statesCount];
for (uint i = 0; i < initialStates.Length; i++)
{
initialStates[i] = _builder.MkState(_pattern, i);
}
_initialStates = initialStates;
// Create the dot-star pattern (a concatenation of any* with the original pattern)
// and all of its initial states.
_dotStarredPattern = _builder.MkConcat(_builder._anyStar, _pattern);
var dotstarredInitialStates = new DfaMatchingState <TSetType> [statesCount];
for (uint i = 0; i < dotstarredInitialStates.Length; i++)
{
// Used to detect if initial state was reentered,
// but observe that the behavior from the state may ultimately depend on the previous
// input char e.g. possibly causing nullability of \b or \B or of a start-of-line anchor,
// in that sense there can be several "versions" (not more than StateCount) of the initial state.
DfaMatchingState <TSetType> state = _builder.MkState(_dotStarredPattern, i);
state.IsInitialState = true;
dotstarredInitialStates[i] = state;
}
_dotstarredInitialStates = dotstarredInitialStates;
// Create the reverse pattern (the original pattern in reverse order) and all of its
// initial states.
_reversePattern = _pattern.Reverse();
var reverseInitialStates = new DfaMatchingState <TSetType> [statesCount];
for (uint i = 0; i < reverseInitialStates.Length; i++)
{
reverseInitialStates[i] = _builder.MkState(_reversePattern, i);
}
_reverseInitialStates = reverseInitialStates;
// Initialize our fast-lookup for determining the character kind of ASCII characters.
// This is only required when the pattern contains anchors, as otherwise there's only
// ever a single kind used.
if (_pattern._info.ContainsSomeAnchor)
{
var asciiCharKinds = new uint[128];
for (int i = 0; i < asciiCharKinds.Length; i++)
{
TSetType predicate2;
uint charKind;
if (i == '\n')
{
predicate2 = _builder._newLinePredicate;
charKind = CharKind.Newline;
}
else
{
predicate2 = _builder._wordLetterPredicateForAnchors;
charKind = CharKind.WordLetter;
}
asciiCharKinds[i] = _builder._solver.And(GetMinterm(i), predicate2).Equals(_builder._solver.False) ? 0 : charKind;
}
_asciiCharKinds = asciiCharKinds;
}
}
/// <summary>Pretty print the bitvector bv as the character set it represents.</summary> public string PrettyPrint(BitVector bv, CharSetSolver solver) => solver.PrettyPrint(ConvertToBDD(bv, solver));
/// <summary>Generator for BDD Unicode category definitions.</summary>
/// <param name="namespacename">namespace for the class</param>
/// <param name="classname">name of the class</param>
/// <param name="path">path where the file classname.cs is written</param>
public static void Generate(string namespacename, string classname, string path)
{
Debug.Assert(namespacename != null);
Debug.Assert(classname != null);
Debug.Assert(path != null);
using StreamWriter sw = new StreamWriter($"{Path.Combine(path, classname)}.cs");
sw.WriteLine(
$@"// Licensed to the .NET Foundation under one or more agreements.
// The .NET Foundation licenses this file to you under the MIT license.
// This is a programmatically generated file from Regex.GenerateUnicodeTables.
// It provides serialized BDD Unicode category definitions for System.Environment.Version = {Environment.Version}
using System.Globalization;
namespace {namespacename}
{{
internal static class {classname}
{{");
var catMap = new Dictionary <UnicodeCategory, Ranges>();
foreach (UnicodeCategory c in Enum.GetValues <UnicodeCategory>())
{
catMap[c] = new Ranges();
}
Ranges whitespace = new Ranges();
Regex whitespaceRegex = new(@"\s");
for (int i = 0; i <= char.MaxValue; i++)
{
char ch = (char)i;
catMap[char.GetUnicodeCategory(ch)].Add(ch);
if (whitespaceRegex.IsMatch(ch.ToString()))
{
whitespace.Add(ch);
}
}
var charSetSolver = new CharSetSolver();
sw.WriteLine(" /// <summary>Serialized BDD representation of the set of all whitespace characters.</summary>");
sw.Write($" public static ReadOnlySpan<byte> SerializedWhitespaceBDD => ");
WriteByteArrayInitSyntax(sw, charSetSolver.CreateSetFromRanges(whitespace.ranges).SerializeToBytes());
sw.WriteLine(";");
// Generate a BDD representation of each UnicodeCategory.
BDD[] catBDDs = new BDD[catMap.Count];
for (int c = 0; c < catBDDs.Length; c++)
{
catBDDs[c] = charSetSolver.CreateSetFromRanges(catMap[(UnicodeCategory)c].ranges);
}
sw.WriteLine();
sw.WriteLine(" /// <summary>Gets the serialized BDD representations of any defined UnicodeCategory.</summary>");
sw.WriteLine(" public static ReadOnlySpan<byte> GetSerializedCategory(UnicodeCategory category) =>");
sw.WriteLine(" (int)category switch");
sw.WriteLine(" {");
for (int i = 0; i < catBDDs.Length; i++)
{
sw.WriteLine($" {i} => SerializedCategory{i}_{(UnicodeCategory)i},");
}
sw.WriteLine($" _ => default,");
sw.WriteLine(" };");
for (int i = 0; i < catBDDs.Length; i++)
{
sw.WriteLine();
sw.WriteLine($" /// <summary>Serialized BDD representation of the set of all characters in UnicodeCategory.{(UnicodeCategory)i}.</summary>");
sw.Write($" private static ReadOnlySpan<byte> SerializedCategory{i}_{(UnicodeCategory)i} => ");
WriteByteArrayInitSyntax(sw, catBDDs[i].SerializeToBytes());
sw.WriteLine(";");
}
sw.WriteLine($@" }}
}}");
/// <summary>Write the DFA or NFA in DGML format into the TextWriter.</summary>
/// <param name="matcher">The <see cref="SymbolicRegexMatcher"/> for the regular expression.</param>
/// <param name="writer">Writer to which the DGML is written.</param>
/// <param name="nfa">True to create an NFA instead of a DFA.</param>
/// <param name="addDotStar">True to prepend .*? onto the pattern (outside of the implicit root capture).</param>
/// <param name="reverse">If true, then unwind the regex backwards (and <paramref name="addDotStar"/> is ignored).</param>
/// <param name="maxStates">The approximate maximum number of states to include; less than or equal to 0 for no maximum.</param>
/// <param name="maxLabelLength">maximum length of labels in nodes anything over that length is indicated with .. </param>
public static void Write(
TextWriter writer, SymbolicRegexMatcher <TSet> matcher,
bool nfa = false, bool addDotStar = true, bool reverse = false, int maxStates = -1, int maxLabelLength = -1)
{
var charSetSolver = new CharSetSolver();
var explorer = new DfaExplorer(matcher, nfa, addDotStar, reverse, maxStates);
var nonEpsilonTransitions = new Dictionary <(int SourceState, int TargetState), List <(SymbolicRegexNode <TSet>?, TSet)> >();
var epsilonTransitions = new List <Transition>();
foreach (Transition transition in explorer.GetTransitions())
{
if (transition.IsEpsilon)
{
epsilonTransitions.Add(transition);
}
else
{
(int SourceState, int TargetState)p = (transition.SourceState, transition.TargetState);
if (!nonEpsilonTransitions.TryGetValue(p, out List <(SymbolicRegexNode <TSet>?, TSet)>?rules))
{
nonEpsilonTransitions[p] = rules = new List <(SymbolicRegexNode <TSet>?, TSet)>();
}
rules.Add(transition.Label);
}
}
writer.WriteLine("<?xml version=\"1.0\" encoding=\"utf-8\"?>");
writer.WriteLine("<DirectedGraph xmlns=\"http://schemas.microsoft.com/vs/2009/dgml\" ZoomLevel=\"1.5\" GraphDirection=\"TopToBottom\" >");
writer.WriteLine(" <Nodes>");
writer.WriteLine(" <Node Id=\"dfa\" Label=\" \" Group=\"Collapsed\" Category=\"DFA\" DFAInfo=\"{0}\" />", GetDFAInfo(explorer, charSetSolver));
writer.WriteLine(" <Node Id=\"dfainfo\" Category=\"DFAInfo\" Label=\"{0}\"/>", GetDFAInfo(explorer, charSetSolver));
foreach (int state in explorer.GetStates())
{
writer.WriteLine(" <Node Id=\"{0}\" Label=\"{0}\" Category=\"State\" Group=\"Collapsed\" StateInfo=\"{1}\">", state, explorer.DescribeState(state));
if (state == explorer.InitialState)
{
writer.WriteLine(" <Category Ref=\"InitialState\" />");
}
if (explorer.IsFinalState(state))
{
writer.WriteLine(" <Category Ref=\"FinalState\" />");
}
writer.WriteLine(" </Node>");
writer.WriteLine(" <Node Id=\"{0}info\" Label=\"{1}\" Category=\"StateInfo\"/>", state, explorer.DescribeState(state));
}
writer.WriteLine(" </Nodes>");
writer.WriteLine(" <Links>");
writer.WriteLine(" <Link Source=\"dfa\" Target=\"{0}\" Label=\"\" Category=\"StartTransition\" />", explorer.InitialState);
writer.WriteLine(" <Link Source=\"dfa\" Target=\"dfainfo\" Label=\"\" Category=\"Contains\" />");
foreach (Transition transition in epsilonTransitions)
{
writer.WriteLine(" <Link Source=\"{0}\" Target=\"{1}\" Category=\"EpsilonTransition\" />", transition.SourceState, transition.TargetState);
}
foreach (KeyValuePair <(int, int), List <(SymbolicRegexNode <TSet>?, TSet)> > transition in nonEpsilonTransitions)
{
string label = string.Join($",{Environment.NewLine} ", DescribeLabels(explorer, transition.Value, charSetSolver));
string info = "";
if (label.Length > (uint)maxLabelLength)
{
info = $"FullLabel = \"{label}\" ";
label = string.Concat(label.AsSpan(0, maxLabelLength), "..");
}
writer.WriteLine($" <Link Source=\"{transition.Key.Item1}\" Target=\"{transition.Key.Item2}\" Label=\"{label}\" Category=\"NonEpsilonTransition\" {info}/>");
}
foreach (int state in explorer.GetStates())
{
writer.WriteLine(" <Link Source=\"{0}\" Target=\"{0}info\" Category=\"Contains\" />", state);
}
writer.WriteLine(" </Links>");
writer.WriteLine(" <Categories>");
writer.WriteLine(" <Category Id=\"DFA\" Label=\"DFA\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"EpsilonTransition\" Label=\"Epsilon transition\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"StartTransition\" Label=\"Initial transition\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"FinalLabel\" Label=\"Final transition\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"FinalState\" Label=\"Final\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"SinkState\" Label=\"Sink state\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"EpsilonState\" Label=\"Epsilon state\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"InitialState\" Label=\"Initial\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"NonEpsilonTransition\" Label=\"Nonepsilon transition\" IsTag=\"True\" />");
writer.WriteLine(" <Category Id=\"State\" Label=\"State\" IsTag=\"True\" />");
writer.WriteLine(" </Categories>");
writer.WriteLine(" <Styles>");
writer.WriteLine(" <Style TargetType=\"Node\" GroupLabel=\"InitialState\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('InitialState')\" />");
writer.WriteLine(" <Setter Property=\"Background\" Value=\"lightgray\" />");
writer.WriteLine(" <Setter Property=\"MinWidth\" Value=\"0\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Node\" GroupLabel=\"FinalState\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('FinalState')\" />");
writer.WriteLine(" <Setter Property=\"Background\" Value=\"lightgreen\" />");
writer.WriteLine(" <Setter Property=\"StrokeThickness\" Value=\"4\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Node\" GroupLabel=\"State\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('State')\" />");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"black\" />");
writer.WriteLine(" <Setter Property=\"Background\" Value=\"white\" />");
writer.WriteLine(" <Setter Property=\"MinWidth\" Value=\"0\" />");
writer.WriteLine(" <Setter Property=\"FontSize\" Value=\"12\" />");
writer.WriteLine(" <Setter Property=\"FontFamily\" Value=\"Arial\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Link\" GroupLabel=\"NonEpsilonTransition\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('NonEpsilonTransition')\" />");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"black\" />");
writer.WriteLine(" <Setter Property=\"FontSize\" Value=\"18\" />");
writer.WriteLine(" <Setter Property=\"FontFamily\" Value=\"Arial\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Link\" GroupLabel=\"StartTransition\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('StartTransition')\" />");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"black\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Link\" GroupLabel=\"EpsilonTransition\" ValueLabel=\"True\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('EpsilonTransition')\" />");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"black\" />");
writer.WriteLine(" <Setter Property=\"StrokeDashArray\" Value=\"8 8\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Link\" GroupLabel=\"FinalLabel\" ValueLabel=\"False\">");
writer.WriteLine(" <Condition Expression=\"HasCategory('FinalLabel')\" />");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"black\" />");
writer.WriteLine(" <Setter Property=\"StrokeDashArray\" Value=\"8 8\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Node\" GroupLabel=\"StateInfo\" ValueLabel=\"True\">");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"white\" />");
writer.WriteLine(" <Setter Property=\"FontSize\" Value=\"18\" />");
writer.WriteLine(" <Setter Property=\"FontFamily\" Value=\"Arial\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" <Style TargetType=\"Node\" GroupLabel=\"DFAInfo\" ValueLabel=\"True\">");
writer.WriteLine(" <Setter Property=\"Stroke\" Value=\"white\" />");
writer.WriteLine(" <Setter Property=\"FontSize\" Value=\"18\" />");
writer.WriteLine(" <Setter Property=\"FontFamily\" Value=\"Arial\" />");
writer.WriteLine(" </Style>");
writer.WriteLine(" </Styles>");
writer.WriteLine("</DirectedGraph>");
}
/// <summary>Pretty print the bitvector bv as the character set it represents.</summary>
public string PrettyPrint(BV bv)
{
CharSetSolver solver = CharSetSolver.Instance;
return(solver.PrettyPrint(ConvertToCharSet(solver, bv)));
}
public override IEnumerable <string> SampleMatches(int k, int randomseed)
{
// Zero is treated as no seed, instead using a system provided one
Random random = randomseed != 0 ? new Random(randomseed) : new Random();
ISolver <TSet> solver = _builder._solver;
CharSetSolver charSetSolver = _builder._charSetSolver;
// Create helper BDDs for handling anchors and preferentially generating ASCII inputs
BDD asciiWordCharacters = charSetSolver.Or(new BDD[] {
charSetSolver.CreateBDDFromRange('A', 'Z'),
charSetSolver.CreateBDDFromRange('a', 'z'),
charSetSolver.CreateBDDFromChar('_'),
charSetSolver.CreateBDDFromRange('0', '9')
});
// Visible ASCII range for input character generation
BDD ascii = charSetSolver.CreateBDDFromRange('\x20', '\x7E');
BDD asciiNonWordCharacters = charSetSolver.And(ascii, charSetSolver.Not(asciiWordCharacters));
// Set up two sets of minterms, one with the additional special minterm for the last end-of-line
Debug.Assert(_builder._minterms is not null);
int[] mintermIdsWithoutZ = new int[_builder._minterms.Length];
int[] mintermIdsWithZ = new int[_builder._minterms.Length + 1];
for (int i = 0; i < _builder._minterms.Length; ++i)
{
mintermIdsWithoutZ[i] = i;
mintermIdsWithZ[i] = i;
}
mintermIdsWithZ[_builder._minterms.Length] = _builder._minterms.Length;
for (int i = 0; i < k; i++)
{
// Holds the generated input so far
StringBuilder inputSoFar = new();
StringBuilder?latestCandidate = null;
// Current set of states reached initially contains just the root
NfaMatchingState states = new(_builder);
// Here one could also consider previous characters for example for \b, \B, and ^ anchors
// and initialize inputSoFar accordingly
states.InitializeFrom(_initialStates[GetCharKind(ReadOnlySpan <char> .Empty, -1)]);
CurrentState statesWrapper = new(states);
// Used for end suffixes
List <string> possibleEndings = new();
while (true)
{
Debug.Assert(states.NfaStateSet.Count > 0);
// Gather the possible endings for satisfying nullability
possibleEndings.Clear();
if (NfaStateHandler.CanBeNullable(ref statesWrapper))
{
// Unconditionally final state or end of the input due to \Z anchor for example
if (NfaStateHandler.IsNullable(ref statesWrapper) ||
NfaStateHandler.IsNullableFor(_builder, ref statesWrapper, CharKind.BeginningEnd))
{
possibleEndings.Add("");
}
// End of line due to end-of-line anchor
if (NfaStateHandler.IsNullableFor(_builder, ref statesWrapper, CharKind.Newline))
{
possibleEndings.Add("\n");
}
// Related to wordborder due to \b or \B
if (NfaStateHandler.IsNullableFor(_builder, ref statesWrapper, CharKind.WordLetter))
{
possibleEndings.Add(ChooseChar(random, asciiWordCharacters, ascii, charSetSolver).ToString());
}
// Related to wordborder due to \b or \B
if (NfaStateHandler.IsNullableFor(_builder, ref statesWrapper, CharKind.General))
{
possibleEndings.Add(ChooseChar(random, asciiNonWordCharacters, ascii, charSetSolver).ToString());
}
}
// If we have a possible ending, then store a candidate input
if (possibleEndings.Count > 0)
{
latestCandidate ??= new();
latestCandidate.Clear();
latestCandidate.Append(inputSoFar);
//Choose some suffix that allows some anchor (if any) to be nullable
latestCandidate.Append(Choose(random, possibleEndings));
// Choose to stop here based on a coin-toss
if (FlipBiasedCoin(random, SampleMatchesStoppingProbability))
{
yield return(latestCandidate.ToString());
break;
}
}
// Shuffle the minterms, including the last end-of-line marker if appropriate
int[] mintermIds = NfaStateHandler.StartsWithLineAnchor(_builder, ref statesWrapper) ?
Shuffle(random, mintermIdsWithZ) :
Shuffle(random, mintermIdsWithoutZ);
foreach (int mintermId in mintermIds)
{
bool success = NfaStateHandler.TakeTransition(_builder, ref statesWrapper, mintermId);
Debug.Assert(success);
if (states.NfaStateSet.Count > 0)
{
TSet minterm = _builder.GetMinterm(mintermId);
// Append a random member of the minterm
inputSoFar.Append(ChooseChar(random, ToBDD(minterm, solver, charSetSolver), ascii, charSetSolver));
break;
}
else
{
// The transition was a dead end, undo and continue to try another minterm
NfaStateHandler.UndoTransition(ref statesWrapper);
}
}
// In the case that there are no next states or input has become too large: stop here
if (states.NfaStateSet.Count == 0 || inputSoFar.Length > SampleMatchesMaxInputLength)
{
// Ending up here without an ending is unlikely but possible for example for infeasible patterns
// such as @"no\bway" or due to poor choice of c -- no anchor is enabled -- so this is a deadend.
if (latestCandidate != null)
{
yield return(latestCandidate.ToString());
}
break;
}
}
}
static BDD ToBDD(TSet set, ISolver <TSet> solver, CharSetSolver charSetSolver) => solver.ConvertToBDD(set, charSetSolver);
/// <summary>
/// Gets a <see cref="BDD"/> that represents <see cref="WordLetter"/> together with the characters
/// \u200C (zero width non joiner) and \u200D (zero width joiner) that are treated as if they were
/// word characters in the context of the anchors \b and \B.
/// </summary>
public static BDD WordLetterForAnchors(CharSetSolver solver) =>
s_wordLetterForAnchors ??
Interlocked.CompareExchange(ref s_wordLetterForAnchors, solver.Or(WordLetter(solver), solver.CreateBDDFromRange('\u200C', '\u200D')), null) ??
s_wordLetterForAnchors;
/// <summary>Pretty print the bitvector bv as the character set it represents.</summary> public string PrettyPrint(ulong bv, CharSetSolver solver) => solver.PrettyPrint(ConvertToBDD(bv, solver));
