DEVELOPING.md

Developing

This document gives some pointers and tips for those interested in contributing to this project.

If you notice that this document is outdated, or if something isn't clear to you, please open an issue!

Architecture

This package is a wrapper around a parser generated by ANTLR. This makes it hard to describe the components, since there are two parsers: the one exported by this project, and the one generated by ANTLR. To avoid confusion, we will call "parser" to the one exported by this project, and "generated parser" to the one generated by ANTLR.

The grammar description used as input for ANTLR lives in this repo, which is included here as a git submodule. The grammar file is Solidity.g4. This means that some issues can be fixed just in this repository, while others depend on updates to the ANTLR grammar. In the latter case, the fix usually involves updating the grammar, updating the submodule in this repository, updating the generated parser and probably updating the parser to handle these changes.

For example, an issue like "the list of state variables in the generated AST code should be in the same order than in the solidity file" probably can be done by just updating the code in the parser. On the other hand, an issue like "support this new keyword" will need an update to the grammar.

The entry point of the project is src/index.js, but most of the important logic is in src/ASTBuilder.js. ASTBuilder is a visitor object that traverses the AST returned by the generated parser, and produces the AST that will be returned by the parser.

Setup

# clone the repository
git clone git@github.com:solidity-parser/parser.git solidity-parser

# initialize the submodule
cd solidity-parser
git submodule update --init

Updating the grammar

The scripts/antlr.sh script should download ANTLR and generate the grammar. For this to work, you need Java (1.6 or higher) installed.

Updating the parser

Run yarn build to build the parser.

Quick testing

Right now the easier way to test something is to start the node REPL and import the project:

yarn build
node
> parser = require('.')
> ast = parser.parse('contract Foo {}')
> console.log(ast.children[0].name)

Understanding the generated parser

The AST returned by the generated parser is very low level and tightly coupled to the grammar definition, that's why using this wrapper is much easier. But if you want to contribute to this parser, you need to deal with this low level representation. Let's see some examples of how to interpret it.

Example 1: PragmaDirective

At the moment of writing this, the function that handles pragma directives in ASTBuilder is this one:

  PragmaDirective(ctx) {
    return {
      name: toText(ctx.pragmaName()),
      value: toText(ctx.pragmaValue())
    }
  },

In this function, we are in the context of a pragma directive. In Solidity.g4, a pragma directive is defined like this:

pragmaDirective : 'pragma' pragmaName pragmaValue ';' ;

This means that the context will have a pragmaName and a pragmaValue function. Since we are only interested in their verbatim content, we use the toText helper to obtain their values.

Example 2: FunctionCall

Let's look now at how FunctionCall is handled:

  FunctionCall(ctx) {
    let args = []
    const names = []

    const ctxArgs = ctx.functionCallArguments()
    if (ctxArgs.expressionList()) {
      args = ctxArgs
        .expressionList()
        .expression()
        .map(exprCtx => this.visit(exprCtx))
    } else if (ctxArgs.nameValueList()) {
      for (const nameValue of ctxArgs.nameValueList().nameValue()) {
        args.push(this.visit(nameValue.expression()))
        names.push(toText(nameValue.identifier()))
      }
    }

    return {
      expression: this.visit(ctx.expression()),
      arguments: args,
      names
    }
  },

This function is a little more involved. The relevant parts of the grammar are:

functionCall : expression '(' functionCallArguments ')' ;

functionCallArguments
  : '{' nameValueList? '}'
  | expressionList? ;

nameValueList
  : nameValue (',' nameValue)* ','? ;

nameValue
  : identifier ':' expression ;

Let's focus on the return. Here we are returning our parser's representation of a FunctionCall node. We decide to return the expression that represents the function called (and for this we recursively use this.visit), along with the names and values of the arguments. names only makes sense when using the f({a: 1}) syntax, otherwise it will be empty. If the f(1, 2) syntax is used, then ctxArgs.expressionList() will be truthy, and then we visit each expression in turn to generate the list of arguments. Otherwise, if ctxArgs.nameValueList() is truthy, we traverse the list of nameValues and use their identifiers as argument names and their expressions as values.

Using ANTLR for visualizing the AST

Sometimes is helpful to visualize the AST generated by the ANTLR parser. For example, let's say we are interested in this contract:

contract Example is Base1, Base2(0) {}

Go to the antlr directory and run:

./build.sh
echo 'contract Example is Base1, Base2(0) {}' > example.sol
java -classpath antlr4.jar:target/ org.antlr.v4.gui.TestRig Solidity sourceUnit -gui < example.sol

And you'll get a window where you can see an inspect the generated tree.