obtaining abstract syntax tree

[nhasabni]

Hi,

I was using SyntaxTree::parseFile and SyntaxNode::toString functions to print a syntax tree, but I saw that this ends up printing a concrete syntax tree with actual variable names. Is there API or a way to print an abstract syntax tree? Or I would need to traverse the tree and print it myself? Thanks.

[MikePopoloski]

I think of the SyntaxTree / SyntaxNode as the "parse tree" or equivalently "concrete syntax tree" as you mention. The abstract syntax tree is built out of symbols, expressions, statements, etc by the Compilation object. That "tree" is serializable to JSON via the ASTSerializer class but it's done in a kind of ad-hoc / not very generalizable fashion so if you want something other than JSON you need to walk the tree yourself (the ASTVisitor class can be of help here).

[nhasabni]

Okay, thanks.

[nhasabni]

Hi Mike,

I was able to write a simple tree traversal to print AST for a SyntaxNode. It works nicely in most cases, but for some reason, it is printing a space, when I am printing a token.

E.g., for expression wb_wr | rd_fifo_becoming_empty, the AST is printed is
(ConditionalPattern(BinaryOrExpression(IdentifierName(wb_wr))(IdentifierName( rd_fifo_becoming_empty))))

Notice the space before rd_fifo_becoming_empty in AST. It looks to me that the token includes space as well. Am I doing something incorrect?

Here is my print function:

 std::string Print(const slang::SyntaxNode& node) const {
    std::ostringstream stream;
    if (slang::toString(node.kind).length() > 0)
      stream << "(" << node.kind;

    auto print_token_for_syntax_kinds = [](slang::SyntaxKind kind) {
      // Print tokens for identifiers and literals
      return ((slang::NameSyntax::isKind(kind)) ||
              (slang::LiteralExpressionSyntax::isKind(kind)));
    };

    for (uint32_t i = 0; i < node.getChildCount(); i++) {
      if (node.childNode(i)) {
        stream << Print(*node.childNode(i));
      } else if (node.childToken(i)) {
        if (print_token_for_syntax_kinds(node.kind)) {
          slang::SyntaxPrinter syntax_printer;
          syntax_printer.print(node.childToken(i));
          stream << "(" << syntax_printer.str() << ")";
        }
      }
    }

    if (slang::toString(node.kind).length() > 0)
      stream << ")";
    return stream.str();
 }

Thanks.

[MikePopoloski]

The SyntaxPrinter class will, by default, print trivia when printing a token (trivia is stuff like whitespace and comments that come before the token). You can disable that with setIncludeTrivia(false).

[nhasabni]

Ah, I should have checked the doc. My bad. That worked. Thanks.