How to get both Production and parse_tree?

[Life4gal]

We can print the result of source file parsing through visualize.
But the printed object must be parse_tree, which needs to be generated by parse_as_tree.

But it will discard all generated values.

To get the result of the parsing, we need parse. (I tried to store the results in the state, so that the return of parse would be insignificant. But I found that the state also had no values.)

This causes that if you want to get both the result and parse_tree, you must parse them twice.

auto file = lexy::read_file<lexy::utf8_encoding>(filename.data());

if (!file)
{
	throw std::runtime_error{"Cannot read file!"};
}

ParseState state{string::string{filename}, std::move(file).buffer()};

if (const auto result =
			lexy::parse<grammar::module_declaration>(state.buffer, state, lexy_ext::report_error.opts({.flags = lexy::visualize_fancy}).path(state.filename.c_str()));
	!result.is_success())
{
	throw std::runtime_error{"Cannot parse file!"};
}

#ifdef PARSE_AS_TREE
lexy::parse_tree_for<decltype(state.buffer)> tree{};
lexy::parse_as_tree<grammar::module_declaration>(tree, state.buffer, state, lexy_ext::report_error.opts({.flags = lexy::visualize_fancy}).path(state.filename.c_str()));
lexy::visualize(stdout, tree, {lexy::visualize_fancy});
#endif

[foonathan]

That's not supported by the library at the moment. I could add a lexy::parse_with_tree function that does it, but want to understand the use case further.

Why do you need both the tree and the value? If you have the tree, you have all the information you need to compute the values yourself.

[foonathan]

Ah, if it's for debugging, I might rather add the ability to combine tracing with parsing: https://lexy.foonathan.net/reference/action/trace/#trace

Would that work for you as well?

[Life4gal]

I'm not sure what I missed, I don't quite understand what you mean.

Am I supposed to be like this?

auto file = lexy::read_file<lexy::utf8_encoding>(filename.data());

if (!file)
{
	throw std::runtime_error{"Cannot read file!"};
}

ParseState state{string::string{filename}, std::move(file).buffer()};

#ifdef TRACE_PARSE
lexy::trace<grammar::module_declaration>(
				stdout,
				state.buffer,
				state,
				{.flags = lexy::visualize_fancy});
#else
if (const auto result =
			lexy::parse<grammar::module_declaration>(state.buffer, state, lexy_ext::report_error.opts({.flags = lexy::visualize_fancy}).path(state.filename.c_str()));
	!result.is_success())
{
	throw std::runtime_error{"Cannot parse file!"};
}
#endif

Or this?

auto file = lexy::read_file<lexy::utf8_encoding>(filename.data());

if (!file)
{
	throw std::runtime_error{"Cannot read file!"};
}

ParseState state{string::string{filename}, std::move(file).buffer()};

if (const auto result =
			lexy::parse<grammar::module_declaration>(state.buffer, state, lexy_ext::report_error.opts({.flags = lexy::visualize_fancy}).path(state.filename.c_str()));
	!result.is_success())
{
	throw std::runtime_error{"Cannot parse file!"};
}

#ifdef TRACE_PARSE
lexy::trace<grammar::module_declaration>(
				stdout,
				state.buffer,
				state,
				{.flags = lexy::visualize_fancy});
#endif

I noticed that trace does not produce any value, either the return value or the value of state.
So obviously the first piece of code is wrong, but if it is the second piece of code, what is the difference between it and parse_as_tree?

[Life4gal]

My idea is whether there is a way to get both the parsed value and print the parse tree at the same time during a source file parsing.

[foonathan]

Yes, currently that's not supported.

My question was: do you explicitly want to dump the parse tree, or is tracing okay as well? Since tracing does not return a value currently it's easier to compose with an existing parse action.

The idea is to enable an interface like lexy::trace<grammar::module_declaration>(lexy::parse_action(state, report_error), stdout, input) which traces parsing and returns the value.

[Life4gal]

Tracing is also OK. In fact, I just hope that there is a way to print the parse tree (if necessary) at the same time when get the value of the parse source file.

[foonathan]

I've started implementing something like that, but it's not trivial. I don't think it's necessary: if you just want to visualize the parse tree for debugging purposes, just parse it twice - performance doesn't matter there anyway.