org.antlr.v4.semantics.SemanticPipeline Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of antlr4 Show documentation
Show all versions of antlr4 Show documentation
The ANTLR 4 grammar compiler.
/*
* Copyright (c) 2012 The ANTLR Project. All rights reserved.
* Use of this file is governed by the BSD-3-Clause license that
* can be found in the LICENSE.txt file in the project root.
*/
package org.antlr.v4.semantics;
import org.antlr.v4.analysis.LeftFactoringRuleTransformer;
import org.antlr.v4.analysis.LeftRecursiveRuleTransformer;
import org.antlr.v4.automata.LexerATNFactory;
import org.antlr.v4.parse.ANTLRParser;
import org.antlr.v4.runtime.Token;
import org.antlr.v4.runtime.misc.NotNull;
import org.antlr.v4.runtime.misc.Tuple2;
import org.antlr.v4.tool.ErrorType;
import org.antlr.v4.tool.Grammar;
import org.antlr.v4.tool.LexerGrammar;
import org.antlr.v4.tool.Rule;
import org.antlr.v4.tool.ast.GrammarAST;
import org.antlr.v4.tool.ast.RuleAST;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
/** Do as much semantic checking as we can and fill in grammar
* with rules, actions, and token definitions.
* The only side effects are in the grammar passed to process().
* We consume a bunch of memory here while we build up data structures
* to perform checking, but all of it goes away after this pipeline object
* gets garbage collected.
*
* After this pipeline finishes, we can be sure that the grammar
* is syntactically correct and that it's semantically correct enough for us
* to attempt grammar analysis. We have assigned all token types.
* Note that imported grammars bring in token and rule definitions
* but only the root grammar and any implicitly created lexer grammar
* get their token definitions filled up. We are treating the
* imported grammars like includes.
*
* The semantic pipeline works on root grammars (those that do the importing,
* if any). Upon entry to the semantic pipeline, all imported grammars
* should have been loaded into delegate grammar objects with their
* ASTs created. The pipeline does the BasicSemanticChecks on the
* imported grammar before collecting symbols. We cannot perform the
* simple checks such as undefined rule until we have collected all
* tokens and rules from the imported grammars into a single collection.
*/
public class SemanticPipeline {
public Grammar g;
public SemanticPipeline(Grammar g) {
this.g = g;
}
public void process() {
if ( g.ast==null ) return;
// COLLECT RULE OBJECTS
RuleCollector ruleCollector = new RuleCollector(g);
ruleCollector.process(g.ast);
// CLONE RULE ASTs FOR CONTEXT REFERENCE
for (Rule rule : ruleCollector.rules.values()) {
List list = g.contextASTs.get(rule.getBaseContext());
if (list == null) {
list = new ArrayList();
g.contextASTs.put(rule.getBaseContext(), list);
}
list.add((RuleAST)rule.ast.dupTree());
}
// DO BASIC / EASY SEMANTIC CHECKS
int prevErrors = g.tool.errMgr.getNumErrors();
BasicSemanticChecks basics = new BasicSemanticChecks(g, ruleCollector);
basics.process();
if ( g.tool.errMgr.getNumErrors()>prevErrors ) return;
// TRANSFORM LEFT-RECURSIVE RULES
prevErrors = g.tool.errMgr.getNumErrors();
LeftRecursiveRuleTransformer lrtrans =
new LeftRecursiveRuleTransformer(g.ast, ruleCollector.rules.values(), g);
lrtrans.translateLeftRecursiveRules();
// don't continue if we got errors during left-recursion elimination
if ( g.tool.errMgr.getNumErrors()>prevErrors ) {
return;
}
// AUTO LEFT FACTORING
LeftFactoringRuleTransformer lftrans = new LeftFactoringRuleTransformer(g.ast, ruleCollector.rules, g);
lftrans.translateLeftFactoredRules();
// STORE RULES IN GRAMMAR
for (Rule r : ruleCollector.rules.values()) {
g.defineRule(r);
}
// COLLECT SYMBOLS: RULES, ACTIONS, TERMINALS, ...
SymbolCollector collector = new SymbolCollector(g);
collector.process(g.ast);
// CHECK FOR SYMBOL COLLISIONS
SymbolChecks symcheck = new SymbolChecks(g, collector);
symcheck.process(); // side-effect: strip away redef'd rules.
for (GrammarAST a : collector.namedActions) {
g.defineAction(a);
}
// LINK (outermost) ALT NODES WITH Alternatives
for (Rule r : g.rules.values()) {
for (int i=1; i<=r.numberOfAlts; i++) {
r.alt[i].ast.alt = r.alt[i];
}
}
// ASSIGN TOKEN TYPES
g.importTokensFromTokensFile();
if ( g.isLexer() ) {
assignLexerTokenTypes(g, collector.tokensDefs);
}
else {
assignTokenTypes(g, collector.tokensDefs,
collector.tokenIDRefs, collector.terminals);
}
symcheck.checkForModeConflicts(g);
assignChannelTypes(g, collector.channelDefs);
// CHECK RULE REFS NOW (that we've defined rules in grammar)
symcheck.checkRuleArgs(g, collector.rulerefs);
identifyStartRules(collector);
symcheck.checkForQualifiedRuleIssues(g, collector.qualifiedRulerefs);
// don't continue if we got symbol errors
if ( g.tool.getNumErrors()>0 ) return;
// CHECK ATTRIBUTE EXPRESSIONS FOR SEMANTIC VALIDITY
AttributeChecks.checkAllAttributeExpressions(g);
UseDefAnalyzer.trackTokenRuleRefsInActions(g);
}
void identifyStartRules(SymbolCollector collector) {
for (GrammarAST ref : collector.rulerefs) {
String ruleName = ref.getText();
Rule r = g.getRule(ruleName);
if ( r!=null ) r.isStartRule = false;
}
}
void assignLexerTokenTypes(Grammar g, List tokensDefs) {
Grammar G = g.getOutermostGrammar(); // put in root, even if imported
for (GrammarAST def : tokensDefs) {
// tokens { id (',' id)* } so must check IDs not TOKEN_REF
if ( Grammar.isTokenName(def.getText()) ) {
G.defineTokenName(def.getText());
}
}
/* Define token types for nonfragment rules which do not include a 'type(...)'
* or 'more' lexer command.
*/
for (Rule r : g.rules.values()) {
if ( !r.isFragment() && !hasTypeOrMoreCommand(r) ) {
G.defineTokenName(r.name);
}
}
// FOR ALL X : 'xxx'; RULES, DEFINE 'xxx' AS TYPE X
List> litAliases = Grammar.getStringLiteralAliasesFromLexerRules(g.ast);
Set conflictingLiterals = new HashSet();
if ( litAliases!=null ) {
for (Tuple2 pair : litAliases) {
GrammarAST nameAST = pair.getItem1();
GrammarAST litAST = pair.getItem2();
if ( !G.stringLiteralToTypeMap.containsKey(litAST.getText()) ) {
G.defineTokenAlias(nameAST.getText(), litAST.getText());
}
else {
// oops two literal defs in two rules (within or across modes).
conflictingLiterals.add(litAST.getText());
}
}
for (String lit : conflictingLiterals) {
// Remove literal if repeated across rules so it's not
// found by parser grammar.
Integer value = G.stringLiteralToTypeMap.remove(lit);
if (value != null && value > 0 && value < G.typeToStringLiteralList.size() && lit.equals(G.typeToStringLiteralList.get(value))) {
G.typeToStringLiteralList.set(value, null);
}
}
}
}
boolean 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.getChildCount(); i++) {
GrammarAST node = (GrammarAST)altActionAst.getChild(i);
if (node.getType() == ANTLRParser.LEXER_ACTION_CALL) {
if ("type".equals(node.getChild(0).getText())) {
return true;
}
}
else if ("more".equals(node.getText())) {
return true;
}
}
return false;
}
void assignTokenTypes(Grammar g, List tokensDefs,
List tokenIDs, List terminals)
{
//Grammar G = g.getOutermostGrammar(); // put in root, even if imported
// create token types for tokens { A, B, C } ALIASES
for (GrammarAST alias : tokensDefs) {
if (g.getTokenType(alias.getText()) != Token.INVALID_TYPE) {
g.tool.errMgr.grammarError(ErrorType.TOKEN_NAME_REASSIGNMENT, g.fileName, alias.token, alias.getText());
}
g.defineTokenName(alias.getText());
}
// DEFINE TOKEN TYPES FOR TOKEN REFS LIKE ID, INT
for (GrammarAST idAST : tokenIDs) {
if (g.getTokenType(idAST.getText()) == Token.INVALID_TYPE) {
g.tool.errMgr.grammarError(ErrorType.IMPLICIT_TOKEN_DEFINITION, g.fileName, idAST.token, idAST.getText());
}
g.defineTokenName(idAST.getText());
}
// VERIFY TOKEN TYPES FOR STRING LITERAL REFS LIKE 'while', ';'
for (GrammarAST termAST : terminals) {
if (termAST.getType() != ANTLRParser.STRING_LITERAL) {
continue;
}
if (g.getTokenType(termAST.getText()) == Token.INVALID_TYPE) {
g.tool.errMgr.grammarError(ErrorType.IMPLICIT_STRING_DEFINITION, g.fileName, termAST.token, termAST.getText());
}
}
g.tool.log("semantics", "tokens="+g.tokenNameToTypeMap);
g.tool.log("semantics", "strings="+g.stringLiteralToTypeMap);
}
/**
* Assign constant values to custom channels defined in a grammar.
*
* @param g The grammar.
* @param channelDefs A collection of AST nodes defining individual channels
* within a {@code channels{}} block in the grammar.
*/
void assignChannelTypes(Grammar g, List channelDefs) {
Grammar outermost = g.getOutermostGrammar();
for (GrammarAST channel : channelDefs) {
String channelName = channel.getText();
// Channel names can't alias tokens or modes, because constant
// values are also assigned to them and the ->channel(NAME) lexer
// command does not distinguish between the various ways a constant
// can be declared. This method does not verify that channels do not
// alias rules, because rule names are not associated with constant
// values in ANTLR grammar semantics.
if (g.getTokenType(channelName) != Token.INVALID_TYPE) {
g.tool.errMgr.grammarError(ErrorType.CHANNEL_CONFLICTS_WITH_TOKEN, g.fileName, channel.token, channelName);
}
if (LexerATNFactory.COMMON_CONSTANTS.containsKey(channelName)) {
g.tool.errMgr.grammarError(ErrorType.CHANNEL_CONFLICTS_WITH_COMMON_CONSTANTS, g.fileName, channel.token, channelName);
}
if (outermost instanceof LexerGrammar) {
LexerGrammar lexerGrammar = (LexerGrammar)outermost;
if (lexerGrammar.modes.containsKey(channelName)) {
g.tool.errMgr.grammarError(ErrorType.CHANNEL_CONFLICTS_WITH_MODE, g.fileName, channel.token, channelName);
}
}
outermost.defineChannelName(channel.getText());
}
}
}