private IEnumerable <SYMBOL_ENUM> recursiveTrackBySet(SYMBOL_ENUM nonTerm)
{
var track = new LinkedList <SYMBOL_ENUM>();
SYMBOL_ENUM current = nonTerm;
while (true)
{
SYMBOL_ENUM sym;
if (!covers[nonTerm].TryGetValue(SymbolChunk.Create(current), out sym))
{
if (current.Equals(nonTerm))
{
return(track);
}
else
{
throw new NotImplementedException("Internal error");
}
}
track.AddFirst(sym);
if (sym.Equals(nonTerm))
{
return(track);
}
current = sym;
}
}
private IEnumerable <SymbolChunk <SYMBOL_ENUM> > enumerate(int width)
{
if (width == 1)
{
return(symbols.Select(it => SymbolChunk.Create(it)));
}
else
{
return(enumerate(width - 1).Select(chunk => symbols.Select(s => SymbolChunk.Create(chunk.Symbols.Concat(s)))).Flatten());
}
}
private void bootstrapFromFirstSets(SYMBOL_ENUM symbol)
{
SymbolChunkSet <SYMBOL_ENUM> follow_set = followSets[symbol];
// in all RHS of productions find our symbol
foreach (Production <SYMBOL_ENUM, TREE_NODE> prod in productions.ProductionsWithNoErrorSymbol())
{
for (int i = 0; i < prod.RhsSymbols.Count(); ++i)
{
// we are above our symbol on RHS of the production
if (prod.RhsSymbols[i].Equals(symbol))
{
// get first set of what can happen AFTER our symbol -- this will be follow set by definition
SymbolChunkSet <SYMBOL_ENUM> chunk_set
= firstSets.GetFirstsOf(prod.RhsSymbols.Skip(i + 1), terminals, syntaxErrorSymbol, lookaheadWidth);
{
SymbolChunkSet <SYMBOL_ENUM> tmp = new SymbolChunkSet <SYMBOL_ENUM>();
// add LHS non-terminal at the end of too short chunks as generator
// because what follows LHS also follows given symbol
foreach (SymbolChunk <SYMBOL_ENUM> chunk in chunk_set.Chunks)
{
if (chunk.Count < lookaheadWidth)
{
tmp.Add(chunk.Append(prod.LhsNonTerminal));
}
else
{
tmp.Add(chunk);
}
}
chunk_set = tmp;
}
if (chunk_set.IsEmpty)
{
chunk_set.Add(SymbolChunk.Create(prod.LhsNonTerminal));
}
follow_set.Add(chunk_set);
}
}
}
}
/// <summary>
/// Add from lower to highest, symbols in single call are equal
/// </summary>
public void AddOperator(AssociativityEnum assoc, params SYMBOL_ENUM[] opSymbols)
{
if (priorityGroupCounter == 0)
{
++runningPriority;
}
foreach (SYMBOL_ENUM op_symbol in opSymbols)
{
addEntry(this.operators, new SymbolPrecedence <SYMBOL_ENUM>(SymbolPrecedence.ModeEnum.BasicOperatorSearch)
{
Symbols = SymbolChunk.Create(op_symbol),
Associativity = assoc,
Priority = runningPriority
}, symbolsRep);
}
}
private void bootstrapFromFirstSets(SYMBOL_ENUM symbol)
{
SymbolChunkSet <SYMBOL_ENUM> current_set = horizonSets[symbol];
// in all RHS of productions find our symbol
foreach (Production <SYMBOL_ENUM, TREE_NODE> prod in productions.Entries)
{
if (symbol.Equals(prod.LhsNonTerminal) || coverSets[symbol].Contains(SymbolChunk.Create(prod.LhsNonTerminal)))
{
continue;
}
for (int i = 0; i < prod.RhsSymbols.Count(); ++i)
{
// we are on our symbol on RHS of the production
if (prod.RhsSymbols[i].Equals(symbol))
{
// get first set of what can happen AFTER our symbol -- this will be our horizon
SymbolChunkSet <SYMBOL_ENUM> src_set = firstSets.GetFirstsOf(prod.RhsSymbols.Skip(i + 1), terminals, syntaxErrorSymbol, lookaheadWidth);
SymbolChunkSet <SYMBOL_ENUM> chunk_set = new SymbolChunkSet <SYMBOL_ENUM>();
// add follow set at the end of too short chunks
foreach (SymbolChunk <SYMBOL_ENUM> chunk in src_set.Chunks)
{
if (chunk.Count < lookaheadWidth)
{
chunk_set.Add(SymbolChunkSet.MultiConcat(chunk, followSets[prod.LhsNonTerminal], lookaheadWidth));
}
else
{
chunk_set.Add(chunk);
}
}
if (chunk_set.IsEmpty)
{
chunk_set.Add(followSets[prod.LhsNonTerminal]);
}
current_set.Add(chunk_set);
}
}
}
}
private CoverSets <SYMBOL_ENUM> initCoverSets()
{
var cover_sets = new CoverSets <SYMBOL_ENUM>();
foreach (SYMBOL_ENUM sym in terminals)
{
var set = new SymbolChunkSet <SYMBOL_ENUM>();
set.Add(SymbolChunk.Create(sym));
cover_sets.Add(sym, set, sym);
}
// fill non terminals with everything found in the productions, terminals will stay as they are added here
// non terminals will serve as generators
foreach (SYMBOL_ENUM non_term in nonTerminals)
{
var set = new SymbolChunkSet <SYMBOL_ENUM>();
foreach (Production <SYMBOL_ENUM, TREE_NODE> prod in productions.FilterByLhs(non_term))
{
foreach (SYMBOL_ENUM sym in prod.RhsSymbols)
{
// if we have error symbol inside production it means LHS cover all non-terminals, because
// error symbol for that LHS covers everything except what stands after error symbol (stop marker)
// thus, everything except stop marker plus stop marker gives --> everything
if (sym.Equals(syntaxErrorSymbol))
{
set.Add(terminals.Select(it => SymbolChunk.Create(it)));
}
else
{
set.Add(SymbolChunk.Create(sym));
}
}
}
cover_sets.Add(non_term, set, non_term);
}
return(cover_sets);
}
private void initFirstSets()
{
firstSets = new FirstSets <SYMBOL_ENUM>(lookaheadWidth);
foreach (SYMBOL_ENUM sym in terminals)
{
var set = new SymbolChunkSet <SYMBOL_ENUM>();
set.Add(SymbolChunk.Create(sym));
firstSets.Add(sym, set);
}
foreach (SYMBOL_ENUM sym in nonTerminals)
{
var set = new SymbolChunkSet <SYMBOL_ENUM>();
// iterate over all productions for this symbol except for error recovery productions
foreach (Production <SYMBOL_ENUM, TREE_NODE> prod in productions
.FilterByLhs(sym)
.Where(it => !it.RhsSymbols.Contains(syntaxErrorSymbol)))
{
// we take only terminals from front
IEnumerable <SYMBOL_ENUM> first_chunk = prod.RhsSymbols.TakeWhile(it => terminals.Contains(it));
int count = first_chunk.Count();
if (count >= lookaheadWidth) // we have more than enough
{
set.Add(SymbolChunk.Create(first_chunk.Take(lookaheadWidth)));
}
// we have less, but there was no more here
else if (count == prod.RhsSymbols.Count)
{
set.Add(SymbolChunk.Create(first_chunk));
}
}
firstSets.Add(sym, set);
}
}
// if we have clear action to do (shift/reduce) pass it forward
// if not, here we try to recover from syntax error
private ActionRecoveryEnum getActionOrRecover(int nodeId,
out IEnumerable <ParseAction <SYMBOL_ENUM, TREE_NODE> > parseActions,
bool startWithRecovering)
{
ActionRecoveryEnum success_result = ActionRecoveryEnum.Success;
while (true)
{
if (startWithRecovering)
{
parseActions = null;
}
else
{
// in normal run we ignore grammar conflicts, user should get conflicts just once, at validating stage
parseActions = actionTable.Get(nodeId, stackMaster.InputView);
}
startWithRecovering = false;
if (parseActions != null)
{
++consecutiveCorrectActionsCount;
// it could be success after naive recovery
return(success_result);
}
// trying to recover from old recovery point
if (stackMaster.IsForked)
{
return(ActionRecoveryEnum.SyntaxError);
}
else
{
// make a lazy message
if (!callUserErrorHandler(stackMaster.InputHead, () => "No action defined at node " + nodeId
+ " for input \"" + SymbolChunk.Create(stackMaster.InputTokens.Take(lookaheadWidth)).ToString(symbolsRep) + "\" with stack \""
+ String.Join(" ", stackMaster.Stack.TakeTail(historyHorizon)
.Select(it => symbolsRep.Get(it.Symbol))) + "\"."))
{
return(ActionRecoveryEnum.StopParsing);
}
consecutiveCorrectActionsCount = 0;
IEnumerable <NfaCell <SYMBOL_ENUM, TREE_NODE> > recovery_items;
if (stack.FindLastWhere(it => it.IsRecoverable, it => it.RecoveryItems, out recovery_items))
{
if (options.Trace)
{
parseLog.Last.Value.Recovered = true;
}
// we would like to get minimal recovery item
// i.e. the one which wastes the minimum of the input in order to recover
NfaCell <SYMBOL_ENUM, TREE_NODE> min_recovery_item = recovery_items
.ArgMin(rec => stackMaster.Input
// for each recovery item compute the count of required tokens from input
.TakeWhile(it => !rec.MatchesRecoveryTerminal(it.Token) && !it.Token.Equals(EofSymbol)).Count())
// not single, because we could hit EOF in several cases
.First();
parseActions = new[] { new ParseAction <SYMBOL_ENUM, TREE_NODE>(false, ReductionAction.Create(min_recovery_item)) };
stackMaster.AdvanceInputWhile(it => !min_recovery_item.MatchesRecoveryTerminal(it.Token) && !it.Token.Equals(EofSymbol));
// we hit the wall
if (IsEndOfInput)
{
return(ActionRecoveryEnum.SyntaxError);
}
stackMaster.AdvanceInput(); // advance past the marker
// setting stack as if we were the old recovery point
// (sometimes we really are, because the last element on the stack can be recovery point)
stackMaster.RemoveLastWhile(it => !it.IsRecoverable);
return(ActionRecoveryEnum.Recovered);
}
else if (IsEndOfInput)
{
return(ActionRecoveryEnum.SyntaxError);
}
else
{
if (options.Trace)
{
parseLog.Last.Value.Recovered = true;
}
// there is no recovery rule defined by the user so try to
// "fix" the errors step by step
stackMaster.AdvanceInput();
// further success will be in fact the result of recovery
success_result = ActionRecoveryEnum.Recovered;
}
}
}
}
public void AddReduceShiftPattern(AssociativityEnum assoc, SYMBOL_ENUM inputSymbol, SYMBOL_ENUM reduceProduction,
IEnumerable <SYMBOL_ENUM> stackOperators, SYMBOL_ENUM shiftProduction, params SYMBOL_ENUM[] restShiftProductions)
{
AddReduceShiftPattern(assoc, SymbolChunk.Create(inputSymbol), reduceProduction, stackOperators, shiftProduction, restShiftProductions);
}
public void AddReduceReducePattern(AssociativityEnum assoc, SYMBOL_ENUM inputSymbol, SYMBOL_ENUM production, params SYMBOL_ENUM[] restProductions)
{
AddReduceReducePattern(assoc, SymbolChunk.Create(inputSymbol), production, restProductions);
}
private SymbolPrecedence <SYMBOL_ENUM> findOperator(SYMBOL_ENUM opSymbol)
{
return(findOperator(SymbolChunk.Create(opSymbol)));
}