/** Remove any lexer rules from a COMBINED; already passed to lexer */
protected void TrimGrammar()
{
if (grammar.type != GrammarType.Combined)
{
return;
}
// form is (header ... ) ( grammar ID (scope ...) ... ( rule ... ) ( rule ... ) ... )
GrammarAST p = root;
// find the grammar spec
while (!p.Text.Equals("grammar"))
{
p = (GrammarAST)p.Parent.GetChild(p.ChildIndex + 1);
}
for (int i = 0; i < p.ChildCount; i++)
{
if (p.GetChild(i).Type != RULE)
{
continue;
}
string ruleName = p.GetChild(i).GetChild(0).Text;
//Console.Out.WriteLine( "rule " + ruleName + " prev=" + prev.getText() );
if (Rule.GetRuleType(ruleName) == RuleType.Lexer)
{
// remove lexer rule
p.DeleteChild(i);
i--;
}
}
//Console.Out.WriteLine( "root after removal is: " + root.ToStringList() );
}
// TODO: this strips the tree properly, but since text()
// uses the start of stop token index and gets text from that
// ineffectively ignores this routine.
public virtual GrammarAST StripLeftRecursion(GrammarAST altAST)
{
GrammarAST lrlabel = null;
GrammarAST first = (GrammarAST)altAST.GetChild(0);
int leftRecurRuleIndex = 0;
if (first.Type == ELEMENT_OPTIONS)
{
first = (GrammarAST)altAST.GetChild(1);
leftRecurRuleIndex = 1;
}
ITree rref = first.GetChild(1); // if label=rule
if ((first.Type == RULE_REF && first.Text.Equals(ruleName)) ||
(rref != null && rref.Type == RULE_REF && rref.Text.Equals(ruleName)))
{
if (first.Type == ASSIGN || first.Type == PLUS_ASSIGN)
{
lrlabel = (GrammarAST)first.GetChild(0);
}
// remove rule ref (first child unless options present)
altAST.DeleteChild(leftRecurRuleIndex);
// reset index so it prints properly (sets token range of
// ALT to start to right of left recur rule we deleted)
GrammarAST newFirstChild = (GrammarAST)altAST.GetChild(leftRecurRuleIndex);
altAST.TokenStartIndex = newFirstChild.TokenStartIndex;
}
return(lrlabel);
}
protected virtual GrammarAST ExpandOptionalQuantifiersForAlt(GrammarAST alt)
{
if (alt.ChildCount == 0)
{
return(null);
}
if (alt.GetChild(0).Type == ANTLRParser.OPTIONAL)
{
GrammarAST root = (GrammarAST)adaptor.Nil();
GrammarAST alt2 = alt.DupTree();
alt2.DeleteChild(0);
if (alt2.ChildCount == 0)
{
adaptor.AddChild(alt2, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
alt.SetChild(0, alt.GetChild(0).GetChild(0));
if (alt.GetChild(0).Type == ANTLRParser.BLOCK && alt.GetChild(0).ChildCount == 1 && alt.GetChild(0).GetChild(0).Type == ANTLRParser.ALT)
{
GrammarAST list = (GrammarAST)adaptor.Nil();
foreach (object tree in ((GrammarAST)alt.GetChild(0).GetChild(0)).Children)
{
adaptor.AddChild(list, tree);
}
adaptor.ReplaceChildren(alt, 0, 0, list);
}
adaptor.AddChild(root, alt);
adaptor.AddChild(root, alt2);
return(root);
}
else if (alt.GetChild(0).Type == ANTLRParser.CLOSURE)
{
GrammarAST root = (GrammarAST)adaptor.Nil();
GrammarAST alt2 = alt.DupTree();
alt2.DeleteChild(0);
if (alt2.ChildCount == 0)
{
adaptor.AddChild(alt2, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
PlusBlockAST plusBlockAST = new PlusBlockAST(ANTLRParser.POSITIVE_CLOSURE, adaptor.CreateToken(ANTLRParser.POSITIVE_CLOSURE, "+"), null);
for (int i = 0; i < alt.GetChild(0).ChildCount; i++)
{
plusBlockAST.AddChild(alt.GetChild(0).GetChild(i));
}
alt.SetChild(0, plusBlockAST);
adaptor.AddChild(root, alt);
adaptor.AddChild(root, alt2);
return(root);
}
return(alt);
}
/** Manually get option node from tree; return null if no defined. */
public static GrammarAST FindOptionValueAST(GrammarRootAST root, string option)
{
GrammarAST options = (GrammarAST)root.GetFirstChildWithType(ANTLRParser.OPTIONS);
if (options != null && options.ChildCount > 0)
{
foreach (object o in options.Children)
{
GrammarAST c = (GrammarAST)o;
if (c.Type == ANTLRParser.ASSIGN &&
c.GetChild(0).Text.Equals(option))
{
return (GrammarAST)c.GetChild(1);
}
}
}
return null;
}
protected override void EnterLabeledLexerElement(GrammarAST tree)
{
IToken label = ((GrammarAST)tree.GetChild(0)).Token;
g.tool.errMgr.GrammarError(ErrorType.V3_LEXER_LABEL,
g.fileName,
label,
label.Text);
}
internal virtual void CheckNumRules(GrammarAST rulesNode)
{
if (rulesNode.ChildCount == 0)
{
GrammarAST root = (GrammarAST)rulesNode.Parent;
GrammarAST IDNode = (GrammarAST)root.GetChild(0);
g.tool.errMgr.GrammarError(ErrorType.NO_RULES, g.fileName,
null, IDNode.Text, g);
}
}
/**
* Match (RULE RULE_REF (BLOCK (ALT .*) (ALT RULE_REF[self] .*) (ALT .*)))
* Match (RULE RULE_REF (BLOCK (ALT .*) (ALT (ASSIGN ID RULE_REF[self]) .*) (ALT .*)))
*/
public static bool HasImmediateRecursiveRuleRefs(GrammarAST t, string ruleName)
{
if (t == null)
{
return(false);
}
GrammarAST blk = (GrammarAST)t.GetFirstChildWithType(BLOCK);
if (blk == null)
{
return(false);
}
int n = blk.Children.Count;
for (int i = 0; i < n; i++)
{
GrammarAST alt = (GrammarAST)blk.Children[i];
ITree first = alt.GetChild(0);
if (first == null)
{
continue;
}
if (first.Type == ELEMENT_OPTIONS)
{
first = alt.GetChild(1);
if (first == null)
{
continue;
}
}
if (first.Type == RULE_REF && first.Text.Equals(ruleName))
{
return(true);
}
ITree rref = first.GetChild(1);
if (rref != null && rref.Type == RULE_REF && rref.Text.Equals(ruleName))
{
return(true);
}
}
return(false);
}
protected override void ExitMode(GrammarAST tree)
{
if (nonFragmentRuleCount == 0)
{
IToken token = tree.Token;
string name = "?";
if (tree.ChildCount > 0)
{
name = tree.GetChild(0).Text;
if (string.IsNullOrEmpty(name))
{
name = "?";
}
token = ((GrammarAST)tree.GetChild(0)).Token;
}
g.tool.errMgr.GrammarError(ErrorType.MODE_WITHOUT_RULES, g.fileName, token, name, g);
}
}
protected virtual GrammarAST TranslateLeftFactoredAlternative(GrammarAST alternative, string factoredRule, bool variant, DecisionFactorMode mode, bool includeFactoredElement)
{
if (mode == DecisionFactorMode.PARTIAL_UNFACTORED && includeFactoredElement)
{
throw new ArgumentException("Cannot include the factored element in unfactored alternatives.");
}
else if (mode == DecisionFactorMode.COMBINED_FACTOR && !includeFactoredElement)
{
throw new ArgumentException("Cannot return a combined answer without the factored element.");
}
Debug.Assert(alternative.ChildCount > 0);
if (alternative.GetChild(0).Type == ANTLRParser.EPSILON)
{
if (mode == DecisionFactorMode.PARTIAL_UNFACTORED)
{
return(alternative);
}
return(null);
}
GrammarAST translatedElement = TranslateLeftFactoredElement((GrammarAST)alternative.GetChild(0), factoredRule, variant, mode, includeFactoredElement);
if (translatedElement == null)
{
return(null);
}
alternative.ReplaceChildren(0, 0, translatedElement);
if (alternative.ChildCount == 0)
{
adaptor.AddChild(alternative, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
Debug.Assert(alternative.ChildCount > 0);
return(alternative);
}
public int CountAltsForBlock(GrammarAST t)
{
int n = 0;
for (int i = 0; i < t.ChildCount; i++)
{
if (t.GetChild(i).Type == ALT)
{
n++;
}
}
return(n);
}
private void HandleDotLoop(GrammarAST start)
{
GrammarAST block = (GrammarAST)start.GetChild(0);
IDictionary <string, object> opts = new Dictionary <string, object>();
opts["greedy"] = "false";
if (grammar.type != GrammarType.Lexer)
{
// parser grammars assume k=1 for .* loops otherwise they (analysis?) look til EOF!
opts["k"] = 1;
}
block.SetOptions(grammar, opts);
}
public PlusBlock(OutputModelFactory factory,
GrammarAST plusRoot,
IList <CodeBlockForAlt> alts)
: base(factory, plusRoot, alts)
{
BlockAST blkAST = (BlockAST)plusRoot.GetChild(0);
PlusBlockStartState blkStart = (PlusBlockStartState)blkAST.atnState;
PlusLoopbackState loop = blkStart.loopBackState;
stateNumber = blkStart.loopBackState.stateNumber;
blockStartStateNumber = blkStart.stateNumber;
loopBackStateNumber = loop.stateNumber;
this.error = GetThrowNoViableAlt(factory, plusRoot, null);
decision = loop.decision;
}
public LL1PlusBlockSingleAlt(OutputModelFactory factory, GrammarAST plusRoot, IList <CodeBlockForAlt> alts)
: base(factory, plusRoot, alts)
{
BlockAST blkAST = (BlockAST)plusRoot.GetChild(0);
PlusBlockStartState blkStart = (PlusBlockStartState)blkAST.atnState;
stateNumber = blkStart.loopBackState.stateNumber;
blockStartStateNumber = blkStart.stateNumber;
PlusBlockStartState plus = (PlusBlockStartState)blkAST.atnState;
this.decision = plus.loopBackState.decision;
IntervalSet[] altLookSets = factory.GetGrammar().decisionLOOK[decision];
IntervalSet loopBackLook = altLookSets[0];
loopExpr = AddCodeForLoopLookaheadTempVar(loopBackLook);
}
protected virtual bool ExpandOptionalQuantifiersForBlock(GrammarAST block, bool variant)
{
IList <GrammarAST> children = new List <GrammarAST>();
for (int i = 0; i < block.ChildCount; i++)
{
GrammarAST child = (GrammarAST)block.GetChild(i);
if (child.Type != ANTLRParser.ALT)
{
children.Add(child);
continue;
}
GrammarAST expandedAlt = ExpandOptionalQuantifiersForAlt(child);
if (expandedAlt == null)
{
return(false);
}
children.Add(expandedAlt);
}
GrammarAST newChildren = (GrammarAST)adaptor.Nil();
newChildren.AddChildren(children);
block.ReplaceChildren(0, block.ChildCount - 1, newChildren);
block.FreshenParentAndChildIndexesDeeply();
if (!variant && block.Parent is RuleAST)
{
RuleAST ruleAST = (RuleAST)block.Parent;
string ruleName = ruleAST.GetChild(0).Text;
Rule r = _rules[ruleName];
IList <GrammarAST> blockAlts = block.GetAllChildrenWithType(ANTLRParser.ALT);
r.numberOfAlts = blockAlts.Count;
r.alt = new Alternative[blockAlts.Count + 1];
for (int i = 0; i < blockAlts.Count; i++)
{
r.alt[i + 1] = new Alternative(r, i + 1);
r.alt[i + 1].ast = (AltAST)blockAlts[i];
}
}
return(true);
}
public virtual Handle Star([NotNull] GrammarAST starAST, [NotNull] Handle elem)
{
StarBlockStartState blkStart = (StarBlockStartState)elem.left;
BlockEndState blkEnd = (BlockEndState)elem.right;
preventEpsilonClosureBlocks.Add(Tuple.Create <Rule, ATNState, ATNState>(currentRule, blkStart, blkEnd));
StarLoopEntryState entry = NewState <StarLoopEntryState>(starAST);
entry.nonGreedy = !((QuantifierAST)starAST).GetGreedy();
entry.sll = false; // no way to express SLL restriction
atn.DefineDecisionState(entry);
LoopEndState end = NewState <LoopEndState>(starAST);
StarLoopbackState loop = NewState <StarLoopbackState>(starAST);
entry.loopBackState = loop;
end.loopBackState = loop;
BlockAST blkAST = (BlockAST)starAST.GetChild(0);
if (((QuantifierAST)starAST).GetGreedy())
{
if (ExpectNonGreedy(blkAST))
{
g.tool.errMgr.GrammarError(ErrorType.EXPECTED_NON_GREEDY_WILDCARD_BLOCK, g.fileName, starAST.Token, starAST.Token.Text);
}
Epsilon(entry, blkStart); // loop enter edge (alt 1)
Epsilon(entry, end); // bypass loop edge (alt 2)
}
else
{
// if not greedy, priority to exit branch; make it first
Epsilon(entry, end); // bypass loop edge (alt 1)
Epsilon(entry, blkStart); // loop enter edge (alt 2)
}
Epsilon(blkEnd, loop); // block end hits loop back
Epsilon(loop, entry); // loop back to entry/exit decision
starAST.atnState = entry; // decision is to enter/exit; blk is its own decision
return(new Handle(entry, end));
}
public virtual Handle Plus([NotNull] GrammarAST plusAST, [NotNull] Handle blk)
{
PlusBlockStartState blkStart = (PlusBlockStartState)blk.left;
BlockEndState blkEnd = (BlockEndState)blk.right;
preventEpsilonClosureBlocks.Add(Tuple.Create <Rule, ATNState, ATNState>(currentRule, blkStart, blkEnd));
PlusLoopbackState loop = NewState <PlusLoopbackState>(plusAST);
loop.nonGreedy = !((QuantifierAST)plusAST).GetGreedy();
loop.sll = false; // no way to express SLL restriction
atn.DefineDecisionState(loop);
LoopEndState end = NewState <LoopEndState>(plusAST);
blkStart.loopBackState = loop;
end.loopBackState = loop;
plusAST.atnState = loop;
Epsilon(blkEnd, loop); // blk can see loop back
BlockAST blkAST = (BlockAST)plusAST.GetChild(0);
if (((QuantifierAST)plusAST).GetGreedy())
{
if (ExpectNonGreedy(blkAST))
{
g.tool.errMgr.GrammarError(ErrorType.EXPECTED_NON_GREEDY_WILDCARD_BLOCK, g.fileName, plusAST.Token, plusAST.Token.Text);
}
Epsilon(loop, blkStart); // loop back to start
Epsilon(loop, end); // or exit
}
else
{
// if not greedy, priority to exit branch; make it first
Epsilon(loop, end); // exit
Epsilon(loop, blkStart); // loop back to start
}
return(new Handle(blkStart, end));
}
private void HandleElementPlusAssign(GrammarAST id, GrammarAST elementStart)
{
if (elementStart.Type == ANTLRParser.ROOT || elementStart.Type == ANTLRParser.BANG)
{
elementStart = (GrammarAST)elementStart.GetChild(0);
}
if (elementStart.Type == RULE_REF)
{
grammar.DefineRuleListLabel(currentRuleName, id.Token, elementStart);
}
else if (elementStart.Type == WILDCARD && grammar.type == GrammarType.TreeParser)
{
grammar.DefineWildcardTreeListLabel(currentRuleName, id.Token, elementStart);
}
else
{
grammar.DefineTokenListLabel(currentRuleName, id.Token, elementStart);
}
}
internal virtual bool HasTypeOrMoreCommand([NotNull] Rule r)
{
GrammarAST ast = r.ast;
if (ast == null)
{
return(false);
}
GrammarAST altActionAst = (GrammarAST)ast.GetFirstDescendantWithType(ANTLRParser.LEXER_ALT_ACTION);
if (altActionAst == null)
{
// the rule isn't followed by any commands
return(false);
}
// first child is the alt itself, subsequent are the actions
for (int i = 1; i < altActionAst.ChildCount; i++)
{
GrammarAST node = (GrammarAST)altActionAst.GetChild(i);
if (node.Type == ANTLRParser.LEXER_ACTION_CALL)
{
if ("type".Equals(node.GetChild(0).Text))
{
return(true);
}
}
else if ("more".Equals(node.Text))
{
return(true);
}
}
return(false);
}
/** Return true if successful */
public virtual bool TranslateLeftRecursiveRule(GrammarRootAST ast,
LeftRecursiveRule r,
string language)
{
//tool.log("grammar", ruleAST.toStringTree());
GrammarAST prevRuleAST = r.ast;
string ruleName = prevRuleAST.GetChild(0).Text;
LeftRecursiveRuleAnalyzer leftRecursiveRuleWalker =
new LeftRecursiveRuleAnalyzer(prevRuleAST, tool, ruleName, language);
bool isLeftRec;
try
{
//System.Console.WriteLine("TESTING ---------------\n" +
// leftRecursiveRuleWalker.Text(ruleAST));
isLeftRec = leftRecursiveRuleWalker.rec_rule();
}
catch (RecognitionException)
{
isLeftRec = false; // didn't match; oh well
}
if (!isLeftRec)
{
return(false);
}
// replace old rule's AST; first create text of altered rule
GrammarAST RULES = (GrammarAST)ast.GetFirstChildWithType(ANTLRParser.RULES);
string newRuleText = leftRecursiveRuleWalker.GetArtificialOpPrecRule();
//System.Console.WriteLine("created: " + newRuleText);
// now parse within the context of the grammar that originally created
// the AST we are transforming. This could be an imported grammar so
// we cannot just reference this.g because the role might come from
// the imported grammar and not the root grammar (this.g)
RuleAST t = ParseArtificialRule(prevRuleAST.g, newRuleText);
// reuse the name token from the original AST since it refers to the proper source location in the original grammar
((GrammarAST)t.GetChild(0)).Token = ((GrammarAST)prevRuleAST.GetChild(0)).Token;
// update grammar AST and set rule's AST.
RULES.SetChild(prevRuleAST.ChildIndex, t);
r.ast = t;
// Reduce sets in newly created rule tree
GrammarTransformPipeline transform = new GrammarTransformPipeline(g, g.tool);
transform.ReduceBlocksToSets(r.ast);
transform.ExpandParameterizedLoops(r.ast);
// Rerun semantic checks on the new rule
RuleCollector ruleCollector = new RuleCollector(g);
ruleCollector.Visit(t, "rule");
BasicSemanticChecks basics = new BasicSemanticChecks(g, ruleCollector);
// disable the assoc element option checks because they are already
// handled for the pre-transformed rule.
basics.checkAssocElementOption = false;
basics.Visit(t, "rule");
// track recursive alt info for codegen
r.recPrimaryAlts = new List <LeftRecursiveRuleAltInfo>();
foreach (var altInfo in leftRecursiveRuleWalker.prefixAndOtherAlts)
{
r.recPrimaryAlts.Add(altInfo);
}
if (r.recPrimaryAlts.Count == 0)
{
tool.errMgr.GrammarError(ErrorType.NO_NON_LR_ALTS, g.fileName, ((GrammarAST)r.ast.GetChild(0)).Token, r.name);
}
r.recOpAlts = new OrderedHashMap <int, LeftRecursiveRuleAltInfo>();
foreach (var pair in leftRecursiveRuleWalker.binaryAlts)
{
r.recOpAlts[pair.Key] = pair.Value;
}
foreach (var pair in leftRecursiveRuleWalker.ternaryAlts)
{
r.recOpAlts[pair.Key] = pair.Value;
}
foreach (var pair in leftRecursiveRuleWalker.suffixAlts)
{
r.recOpAlts[pair.Key] = pair.Value;
}
// walk alt info records and set their altAST to point to appropriate ALT subtree
// from freshly created AST
SetAltASTPointers(r, t);
// update Rule to just one alt and add prec alt
ActionAST arg = (ActionAST)r.ast.GetFirstChildWithType(ANTLRParser.ARG_ACTION);
if (arg != null)
{
r.args = ScopeParser.ParseTypedArgList(arg, arg.Text, g);
r.args.type = AttributeDict.DictType.ARG;
r.args.ast = arg;
arg.resolver = r.alt[1]; // todo: isn't this Rule or something?
}
// define labels on recursive rule refs we delete; they don't point to nodes of course
// these are so $label in action translation works
foreach (System.Tuple <GrammarAST, string> pair in leftRecursiveRuleWalker.leftRecursiveRuleRefLabels)
{
GrammarAST labelNode = pair.Item1;
GrammarAST labelOpNode = (GrammarAST)labelNode.Parent;
GrammarAST elementNode = (GrammarAST)labelOpNode.GetChild(1);
LabelElementPair lp = new LabelElementPair(g, labelNode, elementNode, labelOpNode.Type);
r.alt[1].labelDefs.Map(labelNode.Text, lp);
}
// copy to rule from walker
r.leftRecursiveRuleRefLabels = leftRecursiveRuleWalker.leftRecursiveRuleRefLabels;
tool.Log("grammar", "added: " + t.ToStringTree());
return(true);
}
/** Merge all the rules, token definitions, and named actions from
* imported grammars into the root grammar tree. Perform:
*
* (tokens { X (= Y 'y')) + (tokens { Z ) -> (tokens { X (= Y 'y') Z)
*
* (@ members {foo}) + (@ members {bar}) -> (@ members {foobar})
*
* (RULES (RULE x y)) + (RULES (RULE z)) -> (RULES (RULE x y z))
*
* Rules in root prevent same rule from being appended to RULES node.
*
* The goal is a complete combined grammar so we can ignore subordinate
* grammars.
*/
public virtual void IntegrateImportedGrammars(Grammar rootGrammar)
{
IList <Grammar> imports = rootGrammar.GetAllImportedGrammars();
if (imports == null)
{
return;
}
GrammarAST root = rootGrammar.ast;
GrammarAST id = (GrammarAST)root.GetChild(0);
GrammarASTAdaptor adaptor = new GrammarASTAdaptor(id.Token.InputStream);
GrammarAST tokensRoot = (GrammarAST)root.GetFirstChildWithType(ANTLRParser.TOKENS_SPEC);
IList <GrammarAST> actionRoots = root.GetNodesWithType(ANTLRParser.AT);
// Compute list of rules in root grammar and ensure we have a RULES node
GrammarAST RULES = (GrammarAST)root.GetFirstChildWithType(ANTLRParser.RULES);
ISet <string> rootRuleNames = new HashSet <string>();
// make list of rules we have in root grammar
IList <GrammarAST> rootRules = RULES.GetNodesWithType(ANTLRParser.RULE);
foreach (GrammarAST r in rootRules)
{
rootRuleNames.Add(r.GetChild(0).Text);
}
foreach (Grammar imp in imports)
{
// COPY TOKENS
GrammarAST imp_tokensRoot = (GrammarAST)imp.ast.GetFirstChildWithType(ANTLRParser.TOKENS_SPEC);
if (imp_tokensRoot != null)
{
rootGrammar.tool.Log("grammar", "imported tokens: " + imp_tokensRoot.Children);
if (tokensRoot == null)
{
tokensRoot = (GrammarAST)adaptor.Create(ANTLRParser.TOKENS_SPEC, "TOKENS");
tokensRoot.g = rootGrammar;
root.InsertChild(1, tokensRoot); // ^(GRAMMAR ID TOKENS...)
}
tokensRoot.AddChildren(imp_tokensRoot.Children);
}
IList <GrammarAST> all_actionRoots = new List <GrammarAST>();
IList <GrammarAST> imp_actionRoots = imp.ast.GetAllChildrenWithType(ANTLRParser.AT);
if (actionRoots != null)
{
foreach (var actionRoot in actionRoots)
{
all_actionRoots.Add(actionRoot);
}
}
foreach (var actionRoot in imp_actionRoots)
{
all_actionRoots.Add(actionRoot);
}
// COPY ACTIONS
if (imp_actionRoots != null)
{
IDictionary <System.Tuple <string, string>, GrammarAST> namedActions =
new Dictionary <System.Tuple <string, string>, GrammarAST>();
rootGrammar.tool.Log("grammar", "imported actions: " + imp_actionRoots);
foreach (GrammarAST at in all_actionRoots)
{
string scopeName = rootGrammar.GetDefaultActionScope();
GrammarAST scope, name, action;
if (at.ChildCount > 2)
{ // must have a scope
scope = (GrammarAST)at.GetChild(0);
scopeName = scope.Text;
name = (GrammarAST)at.GetChild(1);
action = (GrammarAST)at.GetChild(2);
}
else
{
name = (GrammarAST)at.GetChild(0);
action = (GrammarAST)at.GetChild(1);
}
GrammarAST prevAction;
if (!namedActions.TryGetValue(Tuple.Create(scopeName, name.Text), out prevAction) || prevAction == null)
{
namedActions[Tuple.Create(scopeName, name.Text)] = action;
}
else
{
if (prevAction.g == at.g)
{
rootGrammar.tool.errMgr.GrammarError(ErrorType.ACTION_REDEFINITION,
at.g.fileName, name.Token, name.Text);
}
else
{
string s1 = prevAction.Text;
s1 = s1.Substring(1, s1.Length - 2);
string s2 = action.Text;
s2 = s2.Substring(1, s2.Length - 2);
string combinedAction = "{" + s1 + '\n' + s2 + "}";
prevAction.Token.Text = combinedAction;
}
}
}
// at this point, we have complete list of combined actions,
// some of which are already living in root grammar.
// Merge in any actions not in root grammar into root's tree.
foreach (string scopeName in namedActions.Keys.Select(i => i.Item1).Distinct())
{
foreach (string name in namedActions.Keys.Where(i => i.Item1 == scopeName).Select(i => i.Item2))
{
GrammarAST action = namedActions[Tuple.Create(scopeName, name)];
rootGrammar.tool.Log("grammar", action.g.name + " " + scopeName + ":" + name + "=" + action.Text);
if (action.g != rootGrammar)
{
root.InsertChild(1, action.Parent);
}
}
}
}
// COPY RULES
IList <GrammarAST> rules = imp.ast.GetNodesWithType(ANTLRParser.RULE);
if (rules != null)
{
foreach (GrammarAST r in rules)
{
rootGrammar.tool.Log("grammar", "imported rule: " + r.ToStringTree());
string name = r.GetChild(0).Text;
bool rootAlreadyHasRule = rootRuleNames.Contains(name);
if (!rootAlreadyHasRule)
{
RULES.AddChild(r); // merge in if not overridden
rootRuleNames.Add(name);
}
}
}
GrammarAST optionsRoot = (GrammarAST)imp.ast.GetFirstChildWithType(ANTLRParser.OPTIONS);
if (optionsRoot != null)
{
// suppress the warning if the options match the options specified
// in the root grammar
// https://github.com/antlr/antlr4/issues/707
bool hasNewOption = false;
foreach (KeyValuePair <string, GrammarAST> option in imp.ast.GetOptions())
{
string importOption = imp.ast.GetOptionString(option.Key);
if (importOption == null)
{
continue;
}
string rootOption = rootGrammar.ast.GetOptionString(option.Key);
if (!importOption.Equals(rootOption))
{
hasNewOption = true;
break;
}
}
if (hasNewOption)
{
rootGrammar.tool.errMgr.GrammarError(ErrorType.OPTIONS_IN_DELEGATE,
optionsRoot.g.fileName, optionsRoot.Token, imp.name);
}
}
}
rootGrammar.tool.Log("grammar", "Grammar: " + rootGrammar.ast.ToStringTree());
}
/** Build lexer grammar from combined grammar that looks like:
*
* (COMBINED_GRAMMAR A
* (tokens { X (= Y 'y'))
* (OPTIONS (= x 'y'))
* (@ members {foo})
* (@ lexer header {package jj;})
* (RULES (RULE .+)))
*
* Move rules and actions to new tree, don't dup. Split AST apart.
* We'll have this Grammar share token symbols later; don't generate
* tokenVocab or tokens{} section. Copy over named actions.
*
* Side-effects: it removes children from GRAMMAR & RULES nodes
* in combined AST. Anything cut out is dup'd before
* adding to lexer to avoid "who's ur daddy" issues
*/
public virtual GrammarRootAST ExtractImplicitLexer(Grammar combinedGrammar)
{
GrammarRootAST combinedAST = combinedGrammar.ast;
//tool.log("grammar", "before="+combinedAST.toStringTree());
GrammarASTAdaptor adaptor = new GrammarASTAdaptor(combinedAST.Token.InputStream);
GrammarAST[] elements = combinedAST.GetChildrenAsArray();
// MAKE A GRAMMAR ROOT and ID
string lexerName = combinedAST.GetChild(0).Text + "Lexer";
GrammarRootAST lexerAST =
new GrammarRootAST(new CommonToken(ANTLRParser.GRAMMAR, "LEXER_GRAMMAR"), combinedGrammar.ast.tokenStream);
lexerAST.grammarType = ANTLRParser.LEXER;
lexerAST.Token.InputStream = combinedAST.Token.InputStream;
lexerAST.AddChild((ITree)adaptor.Create(ANTLRParser.ID, lexerName));
// COPY OPTIONS
GrammarAST optionsRoot =
(GrammarAST)combinedAST.GetFirstChildWithType(ANTLRParser.OPTIONS);
if (optionsRoot != null && optionsRoot.ChildCount != 0)
{
GrammarAST lexerOptionsRoot = (GrammarAST)adaptor.DupNode(optionsRoot);
lexerAST.AddChild(lexerOptionsRoot);
GrammarAST[] options = optionsRoot.GetChildrenAsArray();
foreach (GrammarAST o in options)
{
string optionName = o.GetChild(0).Text;
if (Grammar.lexerOptions.Contains(optionName) &&
!Grammar.doNotCopyOptionsToLexer.Contains(optionName))
{
GrammarAST optionTree = (GrammarAST)adaptor.DupTree(o);
lexerOptionsRoot.AddChild(optionTree);
lexerAST.SetOption(optionName, (GrammarAST)optionTree.GetChild(1));
}
}
}
// COPY all named actions, but only move those with lexer:: scope
IList <GrammarAST> actionsWeMoved = new List <GrammarAST>();
foreach (GrammarAST e in elements)
{
if (e.Type == ANTLRParser.AT)
{
lexerAST.AddChild((ITree)adaptor.DupTree(e));
if (e.GetChild(0).Text.Equals("lexer"))
{
actionsWeMoved.Add(e);
}
}
}
foreach (GrammarAST r in actionsWeMoved)
{
combinedAST.DeleteChild(r);
}
GrammarAST combinedRulesRoot =
(GrammarAST)combinedAST.GetFirstChildWithType(ANTLRParser.RULES);
if (combinedRulesRoot == null)
{
return(lexerAST);
}
// MOVE lexer rules
GrammarAST lexerRulesRoot = (GrammarAST)adaptor.Create(ANTLRParser.RULES, "RULES");
lexerAST.AddChild(lexerRulesRoot);
IList <GrammarAST> rulesWeMoved = new List <GrammarAST>();
GrammarASTWithOptions[] rules;
if (combinedRulesRoot.ChildCount > 0)
{
rules = combinedRulesRoot.Children.Cast <GrammarASTWithOptions>().ToArray();
}
else
{
rules = new GrammarASTWithOptions[0];
}
foreach (GrammarASTWithOptions r in rules)
{
string ruleName = r.GetChild(0).Text;
if (Grammar.IsTokenName(ruleName))
{
lexerRulesRoot.AddChild((ITree)adaptor.DupTree(r));
rulesWeMoved.Add(r);
}
}
foreach (GrammarAST r in rulesWeMoved)
{
combinedRulesRoot.DeleteChild(r);
}
// Will track 'if' from IF : 'if' ; rules to avoid defining new token for 'if'
IList <System.Tuple <GrammarAST, GrammarAST> > litAliases =
Grammar.GetStringLiteralAliasesFromLexerRules(lexerAST);
ISet <string> stringLiterals = combinedGrammar.GetStringLiterals();
// add strings from combined grammar (and imported grammars) into lexer
// put them first as they are keywords; must resolve ambigs to these rules
// tool.log("grammar", "strings from parser: "+stringLiterals);
int insertIndex = 0;
foreach (string lit in stringLiterals)
{
// if lexer already has a rule for literal, continue
if (litAliases != null)
{
foreach (System.Tuple <GrammarAST, GrammarAST> pair in litAliases)
{
GrammarAST litAST = pair.Item2;
if (lit.Equals(litAST.Text))
{
goto continueNextLit;
}
}
}
// create for each literal: (RULE <uniquename> (BLOCK (ALT <lit>))
string rname = combinedGrammar.GetStringLiteralLexerRuleName(lit);
// can't use wizard; need special node types
GrammarAST litRule = new RuleAST(ANTLRParser.RULE);
BlockAST blk = new BlockAST(ANTLRParser.BLOCK);
AltAST alt = new AltAST(ANTLRParser.ALT);
TerminalAST slit = new TerminalAST(new CommonToken(ANTLRParser.STRING_LITERAL, lit));
alt.AddChild(slit);
blk.AddChild(alt);
CommonToken idToken = new CommonToken(ANTLRParser.TOKEN_REF, rname);
litRule.AddChild(new TerminalAST(idToken));
litRule.AddChild(blk);
lexerRulesRoot.InsertChild(insertIndex, litRule);
// lexerRulesRoot.getChildren().add(0, litRule);
lexerRulesRoot.FreshenParentAndChildIndexes(); // reset indexes and set litRule parent
// next literal will be added after the one just added
insertIndex++;
continueNextLit:
;
}
// TODO: take out after stable if slow
lexerAST.SanityCheckParentAndChildIndexes();
combinedAST.SanityCheckParentAndChildIndexes();
// tool.log("grammar", combinedAST.toTokenString());
combinedGrammar.tool.Log("grammar", "after extract implicit lexer =" + combinedAST.ToStringTree());
combinedGrammar.tool.Log("grammar", "lexer =" + lexerAST.ToStringTree());
if (lexerRulesRoot.ChildCount == 0)
{
return(null);
}
return(lexerAST);
}
protected virtual GrammarAST TranslateLeftFactoredElement(GrammarAST element, string factoredRule, bool variant, DecisionFactorMode mode, bool includeFactoredElement)
{
if (mode == DecisionFactorMode.PARTIAL_UNFACTORED && includeFactoredElement)
{
throw new ArgumentException("Cannot include the factored element in unfactored alternatives.");
}
if (mode == DecisionFactorMode.COMBINED_FACTOR)
{
throw new InvalidOperationException("Cannot return a combined answer.");
}
Debug.Assert(!mode.IncludeFactoredAlts() || !mode.IncludeUnfactoredAlts());
switch (element.Type)
{
case ANTLRParser.ASSIGN:
case ANTLRParser.PLUS_ASSIGN:
{
/* label=a
*
* ==>
*
* factoredElement label=a_factored
*/
GrammarAST translatedChildElement = TranslateLeftFactoredElement((GrammarAST)element.GetChild(1), factoredRule, variant, mode, includeFactoredElement);
if (translatedChildElement == null)
{
return(null);
}
RuleAST ruleAST = (RuleAST)element.GetAncestor(ANTLRParser.RULE);
#if false
LOGGER.log(Level.WARNING, "Could not left factor ''{0}'' out of decision in rule ''{1}'': labeled rule references are not yet supported.",
new object[] { factoredRule, ruleAST.GetChild(0).Text });
#endif
return(null);
//if (!translatedChildElement.IsNil)
//{
// GrammarAST root = (GrammarAST)adaptor.Nil();
// object factoredElement = translatedChildElement;
// if (outerRule)
// {
// adaptor.AddChild(root, factoredElement);
// }
// string action = string.Format("_localctx.{0} = (ContextType)_localctx.getParent().getChild(_localctx.getParent().getChildCount() - 1);", element.GetChild(0).Text);
// adaptor.AddChild(root, new ActionAST(adaptor.CreateToken(ANTLRParser.ACTION, action)));
// return root;
//}
//else
//{
// GrammarAST root = (GrammarAST)adaptor.Nil();
// object factoredElement = adaptor.DeleteChild(translatedChildElement, 0);
// if (outerRule)
// {
// adaptor.AddChild(root, factoredElement);
// }
// adaptor.AddChild(root, element);
// adaptor.ReplaceChildren(element, 1, 1, translatedChildElement);
// return root;
//}
}
case ANTLRParser.RULE_REF:
{
if (factoredRule.Equals(element.Token.Text))
{
if (!mode.IncludeFactoredAlts())
{
return(null);
}
if (includeFactoredElement)
{
// this element is already left factored
return(element);
}
GrammarAST root1 = (GrammarAST)adaptor.Nil();
root1.AddChild((ITree)adaptor.Create(TokenConstants.Epsilon, "EPSILON"));
root1.DeleteChild(0);
return(root1);
}
Rule targetRule;
if (!_rules.TryGetValue(element.Token.Text, out targetRule))
{
return(null);
}
RuleVariants ruleVariants = CreateLeftFactoredRuleVariant(targetRule, factoredRule);
switch (ruleVariants)
{
case RuleVariants.NONE:
if (!mode.IncludeUnfactoredAlts())
{
return(null);
}
// just call the original rule (leave the element unchanged)
return(element);
case RuleVariants.FULLY_FACTORED:
if (!mode.IncludeFactoredAlts())
{
return(null);
}
break;
case RuleVariants.PARTIALLY_FACTORED:
break;
default:
throw new InvalidOperationException();
}
string marker = mode.IncludeFactoredAlts() ? ATNSimulator.RuleLfVariantMarker : ATNSimulator.RuleNolfVariantMarker;
element.SetText(element.Text + marker + factoredRule);
GrammarAST root = (GrammarAST)adaptor.Nil();
if (includeFactoredElement)
{
Debug.Assert(mode.IncludeFactoredAlts());
RuleRefAST factoredRuleRef = new RuleRefAST(adaptor.CreateToken(ANTLRParser.RULE_REF, factoredRule));
factoredRuleRef.SetOption(SUPPRESS_ACCESSOR, (GrammarAST)adaptor.Create(ANTLRParser.ID, "true"));
Rule factoredRuleDef = _rules[factoredRule];
if (factoredRuleDef is LeftRecursiveRule)
{
factoredRuleRef.SetOption(LeftRecursiveRuleTransformer.PRECEDENCE_OPTION_NAME, (GrammarAST)adaptor.Create(ANTLRParser.INT, "0"));
}
if (factoredRuleDef.args != null && factoredRuleDef.args.Size() > 0)
{
throw new NotImplementedException("Cannot left-factor rules with arguments yet.");
}
adaptor.AddChild(root, factoredRuleRef);
}
adaptor.AddChild(root, element);
return(root);
}
case ANTLRParser.BLOCK:
{
GrammarAST cloned = element.DupTree();
if (!TranslateLeftFactoredDecision(cloned, factoredRule, variant, mode, includeFactoredElement))
{
return(null);
}
if (cloned.ChildCount != 1)
{
return(null);
}
GrammarAST root = (GrammarAST)adaptor.Nil();
for (int i = 0; i < cloned.GetChild(0).ChildCount; i++)
{
adaptor.AddChild(root, cloned.GetChild(0).GetChild(i));
}
return(root);
}
case ANTLRParser.POSITIVE_CLOSURE:
{
/* a+
*
* =>
*
* factoredElement a_factored a*
*/
GrammarAST originalChildElement = (GrammarAST)element.GetChild(0);
GrammarAST translatedElement = TranslateLeftFactoredElement(originalChildElement.DupTree(), factoredRule, variant, mode, includeFactoredElement);
if (translatedElement == null)
{
return(null);
}
GrammarAST closure = new StarBlockAST(ANTLRParser.CLOSURE, adaptor.CreateToken(ANTLRParser.CLOSURE, "CLOSURE"), null);
adaptor.AddChild(closure, originalChildElement);
GrammarAST root = (GrammarAST)adaptor.Nil();
if (mode.IncludeFactoredAlts())
{
if (includeFactoredElement)
{
object factoredElement = adaptor.DeleteChild(translatedElement, 0);
adaptor.AddChild(root, factoredElement);
}
}
adaptor.AddChild(root, translatedElement);
adaptor.AddChild(root, closure);
return(root);
}
case ANTLRParser.CLOSURE:
case ANTLRParser.OPTIONAL:
// not yet supported
if (mode.IncludeUnfactoredAlts())
{
return(element);
}
return(null);
case ANTLRParser.DOT:
// ref to imported grammar, not yet supported
if (mode.IncludeUnfactoredAlts())
{
return(element);
}
return(null);
case ANTLRParser.ACTION:
case ANTLRParser.SEMPRED:
if (mode.IncludeUnfactoredAlts())
{
return(element);
}
return(null);
case ANTLRParser.WILDCARD:
case ANTLRParser.STRING_LITERAL:
case ANTLRParser.TOKEN_REF:
case ANTLRParser.NOT:
// terminals
if (mode.IncludeUnfactoredAlts())
{
return(element);
}
return(null);
case ANTLRParser.EPSILON:
// empty tree
if (mode.IncludeUnfactoredAlts())
{
return(element);
}
return(null);
default:
// unknown
return(null);
}
}
protected virtual bool TranslateLeftFactoredDecision(GrammarAST block, string factoredRule, bool variant, DecisionFactorMode mode, bool includeFactoredElement)
{
if (mode == DecisionFactorMode.PARTIAL_UNFACTORED && includeFactoredElement)
{
throw new ArgumentException("Cannot include the factored element in unfactored alternatives.");
}
else if (mode == DecisionFactorMode.COMBINED_FACTOR && !includeFactoredElement)
{
throw new ArgumentException("Cannot return a combined answer without the factored element.");
}
if (!ExpandOptionalQuantifiersForBlock(block, variant))
{
return(false);
}
IList <GrammarAST> alternatives = block.GetAllChildrenWithType(ANTLRParser.ALT);
GrammarAST[] factoredAlternatives = new GrammarAST[alternatives.Count];
GrammarAST[] unfactoredAlternatives = new GrammarAST[alternatives.Count];
IntervalSet factoredIntervals = new IntervalSet();
IntervalSet unfactoredIntervals = new IntervalSet();
for (int i = 0; i < alternatives.Count; i++)
{
GrammarAST alternative = alternatives[i];
if (mode.IncludeUnfactoredAlts())
{
GrammarAST unfactoredAlt = TranslateLeftFactoredAlternative(alternative.DupTree(), factoredRule, variant, DecisionFactorMode.PARTIAL_UNFACTORED, false);
unfactoredAlternatives[i] = unfactoredAlt;
if (unfactoredAlt != null)
{
unfactoredIntervals.Add(i);
}
}
if (mode.IncludeFactoredAlts())
{
GrammarAST factoredAlt = TranslateLeftFactoredAlternative(alternative, factoredRule, variant, mode == DecisionFactorMode.COMBINED_FACTOR ? DecisionFactorMode.PARTIAL_FACTORED : DecisionFactorMode.FULL_FACTOR, includeFactoredElement);
factoredAlternatives[i] = factoredAlt;
if (factoredAlt != null)
{
factoredIntervals.Add(alternative.ChildIndex);
}
}
}
if (factoredIntervals.IsNil && !mode.IncludeUnfactoredAlts())
{
return(false);
}
else if (unfactoredIntervals.IsNil && !mode.IncludeFactoredAlts())
{
return(false);
}
if (unfactoredIntervals.IsNil && factoredIntervals.Count == alternatives.Count && mode.IncludeFactoredAlts() && !includeFactoredElement)
{
for (int i = 0; i < factoredAlternatives.Length; i++)
{
GrammarAST translatedAlt = factoredAlternatives[i];
if (translatedAlt.ChildCount == 0)
{
adaptor.AddChild(translatedAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.SetChild(block, i, translatedAlt);
}
return(true);
}
else if (factoredIntervals.IsNil && unfactoredIntervals.Count == alternatives.Count && mode.IncludeUnfactoredAlts())
{
for (int i = 0; i < unfactoredAlternatives.Length; i++)
{
GrammarAST translatedAlt = unfactoredAlternatives[i];
if (translatedAlt.ChildCount == 0)
{
adaptor.AddChild(translatedAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.SetChild(block, i, translatedAlt);
}
return(true);
}
if (mode == DecisionFactorMode.FULL_FACTOR)
{
return(false);
}
/* for a, b, c being arbitrary `element` trees, this block performs
* this transformation:
*
* factoredElement a
* | factoredElement b
* | factoredElement c
* | ...
*
* ==>
*
* factoredElement (a | b | c | ...)
*/
GrammarAST newChildren = (GrammarAST)adaptor.Nil();
for (int i = 0; i < alternatives.Count; i++)
{
if (mode.IncludeFactoredAlts() && factoredIntervals.Contains(i))
{
bool combineWithPrevious = i > 0 && factoredIntervals.Contains(i - 1) && (!mode.IncludeUnfactoredAlts() || !unfactoredIntervals.Contains(i - 1));
if (combineWithPrevious)
{
GrammarAST translatedAlt = factoredAlternatives[i];
if (translatedAlt.ChildCount == 0)
{
adaptor.AddChild(translatedAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
GrammarAST previous = (GrammarAST)newChildren.GetChild(newChildren.ChildCount - 1);
#if false
if (LOGGER.isLoggable(Level.FINE))
{
LOGGER.log(Level.FINE, previous.ToStringTree());
LOGGER.log(Level.FINE, translatedAlt.ToStringTree());
}
#endif
if (previous.ChildCount == 1 || previous.GetChild(1).Type != ANTLRParser.BLOCK)
{
GrammarAST newBlock = new BlockAST(adaptor.CreateToken(ANTLRParser.BLOCK, "BLOCK"));
GrammarAST newAlt = new AltAST(adaptor.CreateToken(ANTLRParser.ALT, "ALT"));
adaptor.AddChild(newBlock, newAlt);
while (previous.ChildCount > 1)
{
adaptor.AddChild(newAlt, previous.DeleteChild(1));
}
if (newAlt.ChildCount == 0)
{
adaptor.AddChild(newAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.AddChild(previous, newBlock);
}
if (translatedAlt.ChildCount == 1 || translatedAlt.GetChild(1).Type != ANTLRParser.BLOCK)
{
GrammarAST newBlock = new BlockAST(adaptor.CreateToken(ANTLRParser.BLOCK, "BLOCK"));
GrammarAST newAlt = new AltAST(adaptor.CreateToken(ANTLRParser.ALT, "ALT"));
adaptor.AddChild(newBlock, newAlt);
while (translatedAlt.ChildCount > 1)
{
adaptor.AddChild(newAlt, translatedAlt.DeleteChild(1));
}
if (newAlt.ChildCount == 0)
{
adaptor.AddChild(newAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.AddChild(translatedAlt, newBlock);
}
GrammarAST combinedBlock = (GrammarAST)previous.GetChild(1);
adaptor.AddChild(combinedBlock, translatedAlt.GetChild(1).GetChild(0));
#if false
if (LOGGER.isLoggable(Level.FINE))
{
LOGGER.log(Level.FINE, previous.ToStringTree());
}
#endif
}
else
{
GrammarAST translatedAlt = factoredAlternatives[i];
if (translatedAlt.ChildCount == 0)
{
adaptor.AddChild(translatedAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.AddChild(newChildren, translatedAlt);
}
}
if (mode.IncludeUnfactoredAlts() && unfactoredIntervals.Contains(i))
{
GrammarAST translatedAlt = unfactoredAlternatives[i];
if (translatedAlt.ChildCount == 0)
{
adaptor.AddChild(translatedAlt, adaptor.Create(ANTLRParser.EPSILON, "EPSILON"));
}
adaptor.AddChild(newChildren, translatedAlt);
}
}
adaptor.ReplaceChildren(block, 0, block.ChildCount - 1, newChildren);
if (!variant && block.Parent is RuleAST)
{
RuleAST ruleAST = (RuleAST)block.Parent;
string ruleName = ruleAST.GetChild(0).Text;
Rule r = _rules[ruleName];
IList <GrammarAST> blockAlts = block.GetAllChildrenWithType(ANTLRParser.ALT);
r.numberOfAlts = blockAlts.Count;
r.alt = new Alternative[blockAlts.Count + 1];
for (int i = 0; i < blockAlts.Count; i++)
{
r.alt[i + 1] = new Alternative(r, i + 1);
r.alt[i + 1].ast = (AltAST)blockAlts[i];
}
}
return(true);
}