Parse a formula using ANTLR4

Question

I am trying to parse a mathematical formula to a subset of LaTeX using ANTLR4. For example it should parse (a+4)/(b*10) to \frac{a+4}{b\cdot 10}.

My simple grammar creates a tree like this:

Now I am trying to implement parse tree listeners to somehow construct the LaTeX String while the tree is traversed. Here, I am failing because to construct a String like \frac{}{} it has to be built recursively. The parse tree walker, however, visits one tree node after the other (in a breadth-first way as far as I can tell).

I've read about parse tree visitors that might be what I need. But I wasn't able to find some examples how these visitors are applied.

Could you provide an example how parse tree listeners/visitors can be used in this particular case? Do you think my approach to use ANTLR for the parser makes sense in the first place?

Decencies

Answer 1

You can create a parse tree walker by implementing the ParseTreeVisitor interface. For ease of use, you can specify Antlr to generate a base visitor when compiling the grammar (in Antlrworks, Run->Generate Recognizer->Next->Generate Visitor->Next->Finish). The base visitor will be called MyGrammarBaseVisitor. Note that the visitor has a generic type T, which every single visit method should return. I recommend using Void for manual manipulation or String for ease of use during code generation.

After you extend the base visitor (I'll assume here we're dealing with String), you need to override the visit methods. These methods are named after the grammar rules you have. Each of these methods will receive a ParserContext ctx parameter, which you use to visit child rules and/or get terminal values. For example, you could do this:

class MyVisitor extends MyGrammarBaseVisitor<String> {
    @Override
    public String visitMultiplicative(MyGrammarParser.MultiplicativeContext ctx) {
        if (ctx.opMult().getText().equals("/")) return "\\frac{" + visit(ctx.expr(0)) + "}{" + visit(ctx.expr(1)) + "}";
        else return visit(ctx.expr(0)) + "\\cdot " + visit(ctx.expr(1));
    }
    // visit methods for other rules...
}

I'm assuming your multiplicative rule looks like multiplicative: expr opMult expr; opMult: '*' | '/'; You can find more information in The Definitive Antlr 4 Reference. You may also find more information and examples in the Antlr documentation.

Mizzen

This works great, thank you! I'm now using Antlrworks like you suggested (I compiled the grammar in the command line before) and now the *Visitor classes are properly created. – Decencies

@Decencies you can still do that in command-line, just use the -visitor argument! See: theantlrguy.atlassian.net/wiki/display/ANTLR4/… – Mizzen

Thanks for pointing that out. But using Antlrworks directly as NetBeans plugin is more convenient now ;-) Still, the command line will be an option for a possible build script. – Decencies