BlitzLLVM/code_compiler/source/parser.cpp

/// AUTOGENERATED COPYRIGHT HEADER START
// Copyright (C) 2024 Michael Fabian 'Xaymar' Dirks <info@xaymar.com>
// AUTOGENERATED COPYRIGHT HEADER END
#include "parser.hpp"
#include <algorithm>
#include <cctype>
#include <cstdarg>
#include "error.hpp"

blitz::parser::~parser() {}

blitz::parser::parser(std::filesystem::path file) : _file(file), _lexer(), _expr()
{
	_lexer = std::make_shared<blitz::lexer>(file);
}

std::shared_ptr<blitz::ast::node> blitz::parser::current()
{
	return _expr;
}

std::shared_ptr<blitz::ast::node> blitz::parser::next()
{
	// This should return an entire "line" of expressions in one go, i.e.:
	// 1. Local a = 1, b = a, c = b+a
	// -> Local(Variable(a, Expression(Integer(1))), Variable(b, Expresssion(Variable(a))), Variable(c, Expression(Add(Variable(b), Variable(a)))
	// 2. Include "HelloWorld.bb"
	// -> Include(String("HelloWorld.bb"))
	// 3. Function HelloWorld()
	// -> Function(HelloWorld, ...)(
	//    Not quite sure if the above makes sense, we'd be returning many expressions outside of functions, but only one inside a function? Why even bother with the current/next crap then?
	// Handling Include becomes a problem too. I guess we should actually return expressions on a line by line basis, and let the "compiler" figure out scope and stuff.

	// Grab the next token to figure out what behavior we should have.
	while (true) {
		auto token = _lexer->next();
		try {
			switch (token.type) {
			case blitz::token::variant::ENDOFFILE:
				// End of file means there's nothing left to parse.
				_expr.reset();
				return nullptr;
			case blitz::token::variant::COMMENT:
			case blitz::token::variant::NEWLINE:
			case blitz::token::variant::SEPARATOR:
				// Ignore some things that aren't very useful right now.
				continue;
			case blitz::token::variant::TEXT:
				return try_parse(token);
			default:
				throw nullptr;
			}
		} catch (blitz::error const& ex) {
			throw ex;
		} catch (std::exception const& ex) {
			throw new blitz::error(_file, token.location, token.location, ex.what());
		} catch (...) {
			throw new blitz::error(_file, token.location, token.location, blitz::format("Token %s unexpected at this point.", token.to_string().c_str()));
		}
	}
}

std::shared_ptr<blitz::ast::node> blitz::parser::try_parse(blitz::token token)
{
	// ToDo: Switch to a proper Unicode library. Maybe Boost?
	std::string ltext;
	std::transform(token.text.begin(), token.text.end(), ltext.begin(), [](std::string::value_type c) { return std::tolower(c); });

	if ((ltext == "local") || (ltext == "global")) {
		// Local/Global have the same parsing, but different functionality.
		// Should be:
		// Text Text [Symbol(=) Expression] [Symbol(,) Text [Symbol(=) Expression] [Symbol(,) ...]]

	} else if (ltext == "global") {
		// Global ...

	} else if (ltext == "function") {
	} else if (ltext == "select") {
	} else if (ltext == "case") {
	} else if (ltext == "endselect") {
	} else if (ltext == "if") {
	} else if (ltext == "elif") {
	} else if (ltext == "endif") {

	} else if (ltext == "end") {
	}

	return nullptr;
}

std::shared_ptr<blitz::ast::node> blitz::parser::try_parse_variable()
{
	// Text [Symbol(=) Expression(...)] [Symbol(,) [Text [Symbol(=) Expression(...)]]]

	auto label = _lexer->next();
	if (label != blitz::token::variant::TEXT) {
		throw new blitz::error(_file, label.location, label.location, blitz::format("Unexpected %s, expected Text.", label.to_string().c_str()));
	}

	auto node = std::make_shared<blitz::ast::variable>(label);

	auto operand = _lexer->next();
	if (operand == "=") {
		//node->set_value(try_parse_expression());
	} else if (operand == blitz::token::variant::NEWLINE || operand == blitz::token::variant::SEPARATOR || (operand == blitz::token::variant::SYMBOL && operand == ",")) {
		return node;
	} else {
		throw new blitz::error(_file, label.location, operand.location, blitz::format("Unexpected %s, expected Symbol(=), NewLine, Separator, or Symbol(,).", operand.to_string().c_str()));
	}

	return node;
}