Using ANTLR4 for C++ target

For my research, I often need to produce recognizers of languages using different kinds of parsers. ANTLR4 produces Adaptive LL(*) parsers (Parr 2011), and here is a bare-bones example of how to produce a c++ implementation from an ANTLR4 grammar.

First, the grammar, to be placed in a file Expr.g4. This is a simple expression grammar. Note that the filename is important and should match the grammar. Secondly, the EOF at the end of main rule is important. Otherwise the parser will not signal an error if there are unparsed characters.

grammar Expr;

main: expr EOF;
expr: expr ('*'|'/') expr
    | expr ('+'|'-') expr
    | INT
    | '(' expr ')'
INT     : [0-9]+ ;

You need to install antlr4 or download the antlr4 jar file to compile this grammar. antlr4 can be installed in OSX using brew.

$ brew install antlr4

Note: The brew target may not exist. You may be better off downloading this and putting it in the correct place.

You will also need to download the corresponding runtime for your platform. OSX is here. Next, we need to implement a simple driver. That is done as follows

#include <iostream>
#include <strstream>
#include <string>
#include "antlr4-runtime.h"
#include "ExprLexer.h"
#include "ExprParser.h"

class MyParserErrorListener: public antlr4::BaseErrorListener {
  virtual void syntaxError(
      antlr4::Recognizer *recognizer,
      antlr4::Token *offendingSymbol,
      size_t line,
      size_t charPositionInLine,
      const std::string &msg,
      std::exception_ptr e) override {
    std::ostrstream s;
    s << "Line(" << line << ":" << charPositionInLine << ") Error(" << msg << ")";
    throw std::invalid_argument(s.str());

int main(int argc, char *argv[]) {
  antlr4::ANTLRInputStream input(argv[1]);
  ExprLexer lexer(&input);
  antlr4::CommonTokenStream tokens(&lexer);

  MyParserErrorListener errorListner;

  // Only if you want to list the tokens
  // for (auto token : tokens.getTokens()) {
  //  std::cout << token->toString() << std::endl;
  // }
  ExprParser parser(&tokens);
  try {
    antlr4::tree::ParseTree* tree = parser.main();
    std::cout << tree->toStringTree() << std::endl;
    return 0;
  } catch (std::invalid_argument &e) {
    std::cout << e.what() << std::endl;
    return 10;

Compilation can be accomplished by using the following Makefile

# runtime is where you downloaded and extracted
CCARGS=-c -I $(RUNTIME)/antlr4-runtime/ -I $(GENERATED) -std=c++11	
# This assumes you have antlr-4.8-complete.jar in the current directory.
ANTLR4=$(JAVA) -jar antlr-4.8-complete.jar

ANTLRGEN=BaseListener Lexer Listener Parser 
OBJS=$(addsuffix .o,$(addprefix $(OUTPUT)/$(GRAMMAR),$(ANTLRGEN)))
GSOURCES=$(addsuffix .cpp,$(addprefix $(GENERATED)/$(GRAMMAR),$(ANTLRGEN)))

.precious: $(GSOURCES)

all: parser

parser: dirs antlr4 parser.cpp $(OBJS)
	$(CC) $(CCARGS) parser.cpp  -o $(OUTPUT)/parser.o 
	$(CC) $(LDARGS) $(OUTPUT)/parser.o $(OBJS) $(LIBS) -o parser

antlr4: $(GENERATED)/.generated;
$(GENERATED)/.generated: $(GRAMMAR).g4
	$(ANTLR4) -Dlanguage=Cpp -o $(GENERATED) $(GRAMMAR).g4
	@touch $(GENERATED)/.generated

$(OUTPUT)/%.o : $(GENERATED)/%.cpp
	$(CC) $(CCARGS) $< -o $@

$(GENERATED)/%.cpp: $(GENERATED)/.generated;

dirs:; mkdir -p $(OUTPUT) $(GENERATED) 
clean:; rm -rf $(OUTPUT) $(GENERATED)

You can use it as follows

$ make
$ ./parser '1+(2*3)'

Using ANTLR3 for C target

If you are looking to produce C rather than C++, ANTLR4 no longer fits the bill, and we have to use ANTLR3. Further, the grammar accepted by ANTLR3 is slightly more restrictive than that accepted by ANTLR4 (It is LL(*) rather than ALL(*)). Here is the grammar for ANTLR3, with its embedded C code to accept the same language as before.

grammar Expr;


 #include "antlr3defs.h"
 #include "ExprLexer.h"

 int main(int argc, char * argv[]) {

    pExprLexer  lex;
    pExprParser  parser;

    input  = antlr3StringStreamNew((pANTLR3_UINT8)argv[1], ANTLR3_ENC_8BIT, strlen(argv[1]), "_");
    lex    = ExprLexerNew(input);
    tokens = antlr3CommonTokenStreamSourceNew(ANTLR3_SIZE_HINT, TOKENSOURCE(lex));
    parser = ExprParserNew(tokens);



    return 0;

program: expr EOF;
expr: term  ( (PLUS|MINUS) term)*;
term: factor ( (MULT|DIV) factor)*;
factor: INT
    | OP expr CP;
INT  : (DIGIT)+;
OP: '(';
CP: ')';
PLUS: '+';
MINUS: '-';
MULT: '*';
DIV: '/';

WHITESPACE  : ( '\t' | ' ' | '\r' | '\n'| '\u000C' )+
    $channel = HIDDEN;

DIGIT: '0'..'9';

As before, compiling requires the runtime. Once you have that, you can generate and compile as below:

$ java -cp $ANTLR3COMPLETEJAR org.antlr.Tool -o output Expr.g
$ gcc -o expr output/*.c -I $LIBANTLR3C/ -I $LIBANTLR3C/include $LIBANTLR3C/.libs/libantlr3c.a


ANTLR ships with a tool called grun that can help you to debug your grammar. Using that however, is a little different. Here is how one can use it with our Expr grammar.

First, grun requires all files in the same directory, and the files are expected to be in Java, so we copy Expr.g4 to a new directory, and we generate the Expr Java files and compile them first.

$ java -jar antlr-4.8-complete.jar Expr.g4
$ javac -cp antlr-4.8-complete.jar Expr*.java

Now, you can use the grun as below. The grun is implemented by org.antlr.v4.gui.TestRig so we use it directly.

echo -n "1+3" | java -cp .:./antlr-4.8-complete.jar org.antlr.v4.gui.TestRig Expr main -tree
(main (expr (expr 1) + (expr 3)) <EOF>)

You can also get a tree view by

echo -n "1+3" | java -cp .:./antlr-4.8-complete.jar org.antlr.v4.gui.TestRig Expr main -gui

This should correctly pop the tree