/// <summary>
/// Build the NFSA from the abstract syntax tree.
/// There is an NfsaInstance for each start state.
/// Each rule starts with a new nfsa state, which
/// is the target of a new epsilon transition from
/// the real start state, nInst.Entry.
/// </summary>
/// <param name="ast"></param>
public void Build(AAST ast)
{
int index = 0;
DateTime time0 = DateTime.Now;
nfas = new NfsaInstance[ast.StartStateCount];
foreach (KeyValuePair<string, StartState> p in ast.startStates)
{
StartState s = p.Value;
string name = p.Key;
if (!s.IsAll)
{
NfsaInstance nInst = new NfsaInstance(s, this);
nfas[index++] = nInst;
nInst.key = name;
// for each pattern do ...
for (int i = 0; i < s.rules.Count; i++)
{
RuleDesc rule = s.rules[i];
RegExTree tree = rule.Tree;
if (rule.isPredDummyRule)
{
NState entry = nInst.Entry;
nInst.MakePath(tree, entry, entry);
}
else
{
NState start = nInst.MkState();
NState endSt = nInst.MkState();
if (tree.op == RegOp.leftAnchor) // this is a left anchored pattern
{
nInst.AnchorState.AddEpsTrns(start);
tree = ((Unary)tree).kid;
}
else // this is not a left anchored pattern
nInst.Entry.AddEpsTrns(start);
//
// Now check for right anchors, and add states as necessary.
//
if (tree.op == RegOp.eof)
{
//
// <<EOF>> rules are always emitted outside
// of the usual subset construction framework.
// We ensure that we do not get spurious warnings.
//
rule.useCount = 1;
nInst.eofAction = rule.aSpan;
nInst.MakePath(tree, start, endSt);
nInst.MarkAccept(endSt, rule);
}
else if (tree.op == RegOp.rightAnchor)
{
tree = ((Unary)tree).kid;
nInst.MakePath(tree, start, endSt);
AddAnchorContext(nInst, endSt, rule);
}
else
{
nInst.MakePath(tree, start, endSt);
nInst.MarkAccept(endSt, rule);
}
}
}
}
}
if (task.Verbose)
{
Console.Write("GPLEX: NFSA built");
Console.Write((task.Errors ? ", errors detected" : " without error"));
Console.Write((task.Warnings ? "; warnings issued. " : ". "));
Console.WriteLine(TaskState.ElapsedTime(time0));
}
if (task.Summary)
WriteSummary(time0);
}
internal void Process(string fileArg)
{
GetNames(fileArg);
// check for file exists
OpenSource();
// parse source file
if (inputFile != null)
{
DateTime start = DateTime.Now;
try
{
handler = new ErrorHandler();
scanner = new QUT.Gplex.Lexer.Scanner(inputFile);
parser = new QUT.Gplex.Parser.Parser(scanner);
scanner.yyhdlr = handler;
parser.Initialize(this, scanner, handler, new OptionParser2(ParseOption));
aast = parser.Aast;
parser.Parse();
// aast.DiagnosticDump();
if (verbose)
Status(start);
CheckOptions();
if (!Errors && !ParseOnly)
{ // build NFSA
if (ChrClasses) {
DateTime t0 = DateTime.Now;
partition = new Partition( TargetSymCardinality, this );
partition.FindClasses( aast );
partition.FixMap();
if (verbose)
ClassStatus( t0, partition.Length );
}
else
CharRange.Init( TargetSymCardinality );
nfsa = new NFSA(this);
nfsa.Build(aast);
if (!Errors)
{ // convert to DFSA
dfsa = new DFSA(this);
dfsa.Convert(nfsa);
if (!Errors)
{ // minimize automaton
if (minimize)
dfsa.Minimize();
if (!Errors && !checkOnly)
{ // emit the scanner to output file
TextReader frameRdr = FrameReader();
TextWriter outputWrtr = OutputWriter();
dfsa.EmitScanner(frameRdr, outputWrtr);
if (!embedBuffers)
CopyBufferCode();
// Clean up!
if (frameRdr != null)
frameRdr.Close();
if (outputWrtr != null)
outputWrtr.Close();
}
}
}
}
}
catch (Exception ex)
{
string str = ex.Message;
handler.AddError(str, aast.AtStart);
throw;
}
}
}
/// <summary>
/// This method constructs a RangeLiteral holding
/// all of the codepoints from all planes for which
/// the Test delegate returns true.
/// </summary>
/// <param name="name"></param>
/// <param name="aast"></param>
/// <param name="max"></param>
internal void Populate( string name, AAST aast ) {
DateTime begin = DateTime.Now;
int max = aast.Task.TargetSymCardinality;
this.rangeLit = new RangeLiteral( false );
//
// Run the delegate over all the values
// between '\0' and (max-1). Find contiguous
// true values and add to the CharRange list.
//
int j = 0;
int codepage = aast.CodePage;
if (max > 256 ||
codepage == Automaton.TaskState.rawCP ||
codepage == Automaton.TaskState.guessCP) {
if (max <= 256 && codepage == Automaton.TaskState.guessCP)
aast.hdlr.ListError( aast.AtStart, 93 );
//
// We are generating a set of numeric code points with
// the named property. No interpretation is needed, either
// (1) because this is for a unicode scanner that has
// already decoded its input element stream, OR
// (2) the user has commanded /codepoint:raw to indicate
// that no interpretation is to be used.
//
while (j < max) {
int start;
while (j < max && !Test( j ))
j++;
if (j == max)
break;
start = j;
while (j < max && Test( j ))
j++;
this.rangeLit.list.Add( new CharRange( start, (j - 1) ) );
}
}
else {
// We are generating a set of byte values from the
// 0x00 to 0xFF "alphabet" that correspond to unicode
// characters with the named property. The meaning of
// "corresponds" is defined by the nominated codepage.
//
// Check codepage for single byte property.
//
Encoding enc = Encoding.GetEncoding( codepage );
Decoder decoder = enc.GetDecoder();
if (!enc.IsSingleByte)
aast.hdlr.ListError( aast.AtStart, 92 );
//
// Construct character map for bytes.
//
int bNum, cNum;
bool done;
char[] cArray = new char[256];
byte[] bArray = new byte[256];
for (int b = 0; b < 256; b++) {
bArray[b] = (byte)b;
cArray[b] = '?';
}
decoder.Convert( bArray, 0, 256, cArray, 0, 256, true, out bNum, out cNum, out done );
//
// Now construct the CharRange literal
//
while (j < max) {
int start;
while (j < max && !Test( cArray[j] ))
j++;
if (j == max)
break;
start = j;
while (j < max && Test( cArray[j] ))
j++;
this.rangeLit.list.Add( new CharRange( start, (j - 1) ) );
}
}
if (aast.IsVerbose) {
Console.WriteLine( "GPLEX: Generating [:{0}:], {1}", name, Gplex.Automaton.TaskState.ElapsedTime( begin ) );
}
}
internal ReParser( string str, LexSpan spn, AAST parent ) {
if (parent.task.Unicode)
CharacterUtilities.SetUnicode();
symCard = parent.task.HostSymCardinality;
pat = str;
span = spn;
InitReParser();
this.parent = parent;
//
// This is ugly, but we cannot manipulate
// RangeLists unless the alphabet upper bound
// is known to the code of class Partition.
//
CharRange.Init( parent.task.TargetSymCardinality );
}
internal void ParseRE( AAST aast ) { regX = new AAST.ReParser( verb, vrbSpan, aast ).Parse(); }
/// <summary>
/// This method detects the presence of code *between* rules. Such code has
/// no unambiguous meaning, and is skipped, with a warning message.
/// </summary>
/// <param name="aast"></param>
internal void FinalizeCode( AAST aast ) {
for (int i = 0; i < locs.Count; i++) {
LexSpan loc = locs[i];
if (loc.startLine < FLine)
aast.AddCodeSpan( AAST.Destination.scanProlog, loc );
else if (loc.startLine > LLine)
aast.AddCodeSpan( AAST.Destination.scanEpilog, loc );
else // code is between rules
aast.hdlr.ListError( loc, 110 );
}
}
void Check( AAST aast, RegExTree tree ) {
Binary bnryTree;
Unary unryTree;
if (tree == null) return;
switch (tree.op) {
case RegOp.charClass:
case RegOp.primitive:
case RegOp.litStr:
case RegOp.eof:
break;
case RegOp.context:
case RegOp.concat:
case RegOp.alt:
bnryTree = (Binary)tree;
Check( aast, bnryTree.lKid );
Check( aast, bnryTree.rKid );
if (tree.op == RegOp.context &&
bnryTree.lKid.contextLength() == 0 &&
bnryTree.rKid.contextLength() == 0) aast.hdlr.ListError( pSpan, 75 );
break;
case RegOp.closure:
case RegOp.finiteRep:
unryTree = (Unary)tree;
Check( aast, unryTree.kid );
break;
case RegOp.leftAnchor:
case RegOp.rightAnchor:
aast.hdlr.ListError( pSpan, 69 );
break;
}
}
/// <summary>
/// This is the place to perform any semantic checks on the
/// trees corresponding to a rule of the LEX grammar,
/// during a recursive traversal of the tree. It is hard
/// to do these on the fly during AST construction, because
/// of the tree-grafting that happens for lexical categories.
///
/// First check is that '^' and '$' can only appear
/// (logically) at the ends of the pattern.
/// Later need to check ban on multiple right contexts ...
/// </summary>
/// <param name="aast"></param>
void SemanticCheck( AAST aast ) {
RegExTree tree = reAST;
if (tree != null && tree.op == RegOp.leftAnchor) tree = ((Unary)tree).kid;
if (tree != null && tree.op == RegOp.rightAnchor) {
tree = ((Unary)tree).kid;
if (tree.op == RegOp.context)
aast.hdlr.ListError( pSpan, 100 );
}
Check( aast, tree );
if (tree != null)
minPatternLength = tree.minimumLength();
}
// internal void Dump() { Console.WriteLine(pattern); }
internal void ParseRE( AAST aast ) {
reAST = new AAST.ReParser( pattern, pSpan, aast ).Parse();
SemanticCheck( aast );
}
internal static RuleDesc MkDummyRuleDesc( LexCategory cat, AAST aast ) {
RuleDesc result = new RuleDesc();
result.pSpan = null;
result.aSpan = aast.AtStart;
result.isBarAction = false;
result.isPredDummyRule = true;
result.pattern = String.Format( CultureInfo.InvariantCulture, "{{{0}}}", cat.Name );
result.list = new List<StartState>();
result.ParseRE( aast );
result.list.Add( aast.StartStateValue( cat.PredDummyName ) );
return result;
}
internal ReParser(string str, LexSpan spn, AAST parent)
{
if (parent.task.Unicode)
CharacterUtilities.SetUnicode();
symCard = parent.task.HostSymCardinality;
pat = str;
span = spn;
InitReParser();
this.parent = parent;
}