path: root/buildtools/codegen/main.cpp

#include "CodegenLookupTable.h"
#include "CodegenOutput.hpp"
#include "Macros.hpp"
#include "ScopeGuard.hpp"
#include "Utils.hpp"

#include <frozen/unordered_map.h>
#include <stb_c_lexer.h>
#include <cinttypes>
#include <cstdlib>
#include <filesystem>
#include <memory>
#include <string>
#include <string_view>

using namespace std::literals;
namespace fs = std::filesystem;

enum InputOpcode {
	IOP_ProcessSingleFile,
	IOP_ProcessRecursively,
	IOP_COUNT,
};

enum CodegenDirectives {
	CD_ClassInfo, // BRUSSEL_CLASS
	CD_EnumInfo, // BRUSSEL_ENUM
	// TODO implement
	CD_GlobalSequencer, // BRUSSEL_INIT
	CD_COUNT,
};

enum EnumUnderlyingType {
	EUT_Int8,
	EUT_Int16,
	EUT_Int32,
	EUT_Int64,
	EUT_Uint8,
	EUT_Uint16,
	EUT_Uint32,
	EUT_Uint64,
	EUT_COUNT,
};

InputOpcode ParseInputOpcode(std::string_view text) {
	if (text == "single"sv) {
		return IOP_ProcessSingleFile;
	} else if (text == "rec"sv) {
		return IOP_ProcessRecursively;
	}
	return IOP_COUNT;
}

struct InputDefinitionStruct {
	std::string name;
};

struct InputDefinitionEnumElement {
	std::string name;
	uint64_t value;
};

struct InputDefinitionEnum {
	std::string name;
	std::vector<InputDefinitionEnumElement> elements;
	EnumUnderlyingType underlyingType;
};

struct StbLexerToken {
	std::string text;
	// Can either be CLEX_* values, or just chars for single character tokens
	int type;
};

void CheckBraceDepth(int braceDpeth) {
	if (braceDpeth < 0) {
		printf("[WARNING] unbalanced brace");
	}
}

void HandleInputFile(std::string_view source) {
	stb_lexer lexer;
	char stringStorage[65536];
	const char* srcBegin = source.data();
	const char* srcEnd = srcBegin + source.length();
	stb_c_lexer_init(&lexer, srcBegin, srcEnd, stringStorage, sizeof(stringStorage));

	std::vector<StbLexerToken> tokens;
	std::vector<InputDefinitionStruct> foundStructs;
	InputDefinitionStruct currStruct;
	std::vector<InputDefinitionEnum> foundEnums;
	InputDefinitionEnum currEnum;

	auto PushFoundStruct = [&]() {
		foundStructs.push_back(std::move(currStruct));
		currStruct = {};
	};
	auto PushFoundEnum = [&]() {
		foundEnums.push_back(std::move(currEnum));
		currEnum = {};
	};

	enum NextMatchingConstruct {
		NMC_None,
		NMC_Enum,
		NMC_StructClass,
	} matchingConstruct = NMC_None;
	bool matchingConstructInBody = false;

	bool matchingDirectiveParams = false;

	int bracePairDepth = 0;

	while (true) {
		// See stb_c_lexer.h's comments, here are a few additinos that aren't made clear in the file:
		// - `lexer->token` (noted as "token" below) after calling stb_c_lexer_get_token() contains either:
		//     1. 0 <= token < 256:          an ASCII character (more precisely a single char that the lexer ate; technically can be an incomplete code unit)
		//     2. token < 0:                 an unknown token
		//     3. One of the `CLEX_*` enums: a special, recognized token such as an operator

		int stbToken = stb_c_lexer_get_token(&lexer);
		if (stbToken == 0) {
			// EOF
			break;
		}

		// TODO needed?
		// StbLexerToken token;
		// token.type = lexer.token;
		// token.text = std::string(lexer.string, lexer.string_len);
		// tokens.push_back(token);

		switch (lexer.token) {
			case CLEX_id: {
				// WORKAROUND: stb_c_lexer doens't set string_len properly when parsing identifiers
				std::string_view idenText(lexer.string);
				// std::string_view idenText(lexer.string, lexer.string_len);
				switch (matchingConstruct) {
					case NMC_StructClass: {
						if (matchingConstructInBody) {
							// TODO
						}
					} break;

					case NMC_Enum: {
						if (matchingConstructInBody) {
							printf("[DEBUG] found enum element '%s'\n", lexer.string);
							currEnum.elements.push_back(InputDefinitionEnumElement{
								.name = std::string(idenText),
								.value = 0, // TODO parse
							});
						} else {
							currEnum.name = std::string(idenText);
						}
					} break;

					default: {
						if (idenText == "struct"sv || idenText == "class"sv) {
							printf("[DEBUG] found struct named\n");
							matchingConstruct = NMC_StructClass;
						} else if (idenText == "enum"sv) {
							printf("[DEBUG] found enum\n");
							matchingConstruct = NMC_Enum;
						} else if (idenText == "BRUSSEL_CLASS"sv) {
							// TODO
							printf("[DEBUG] found BRUSSEL_CLASS\n");
						} else if (idenText == "BRUSSEL_ENUM"sv) {
							matchingDirectiveParams = true;
							printf("[DEBUG] found BRUSSEL_ENUM\n");
						}

						if (matchingDirectiveParams) {
							for (auto& foundEnum : foundEnums) {
								if (foundEnum.name == idenText) {
									// TODO generate data
									break;
								}
							}
							matchingDirectiveParams = false;
						}
					} break;
				}
			} break;

			case CLEX_intlit:
			case CLEX_floatlit:
			case CLEX_dqstring:
			case CLEX_sqstring:
			case CLEX_charlit:
			case CLEX_eq:
			case CLEX_noteq:
			case CLEX_lesseq:
			case CLEX_greatereq:
			case CLEX_andand:
			case CLEX_oror:
			case CLEX_shl:
			case CLEX_shr:
			case CLEX_plusplus:
			case CLEX_minusminus:
			case CLEX_pluseq:
			case CLEX_minuseq:
			case CLEX_muleq:
			case CLEX_diveq:
			case CLEX_modeq:
			case CLEX_andeq:
			case CLEX_oreq:
			case CLEX_xoreq:
			case CLEX_arrow:
			case CLEX_eqarrow:
			case CLEX_shleq:
			case CLEX_shreq: {

			} break;

			case '{': {
				bracePairDepth++;
				CheckBraceDepth(bracePairDepth);

				switch (matchingConstruct) {
					case NMC_StructClass:
					case NMC_Enum: {
						matchingConstructInBody = true;
					} break;

					default: break;
				}
			} break;

			case '}': {
				bracePairDepth--;
				CheckBraceDepth(bracePairDepth);

				switch (matchingConstruct) {
					case NMC_StructClass: {
						matchingConstruct = NMC_None;
						matchingConstructInBody = false;
					} break;

					case NMC_Enum: {
						printf("[DEBUG] committed enum '%s'\n", currEnum.name.c_str());
						for (auto& elm : currEnum.elements) {
							printf("        - element %s = %" PRId64 "\n", elm.name.c_str(), elm.value);
						}

						matchingConstruct = NMC_None;
						matchingConstructInBody = false;
						PushFoundEnum();
					} break;

					default: break;
				}
			} break;

			case CLEX_parse_error: {
				fprintf(stderr, "[ERROR] stb_c_lexer countered a parse error.");
				// TODO how to handle?
			} break;
		}
	}

	if (bracePairDepth != 0) {
		printf("[WARNING] unbalanced brace at end of file.");
	}
}

std::string ReadFileAtOnce(const fs::path& path) {
	auto file = Utils::OpenCstdioFile(path, Utils::Read);
	if (!file) throw std::runtime_error("Failed to open source file.");
	DEFER { fclose(file); };

	fseek(file, 0, SEEK_END);
	auto fileSize = ftell(file);
	rewind(file);

	std::string result(fileSize, '\0');
	fread(result.data(), fileSize, 1, file);

	return result;
}

void HandleArgument(InputOpcode opcode, std::string_view operand) {
	switch (opcode) {
		case IOP_ProcessSingleFile: {
			fs::path filePath(operand);
			auto source = ReadFileAtOnce(filePath);
			HandleInputFile(source);
		} break;

		case IOP_ProcessRecursively: {
			fs::path startPath(operand);
			for (auto& item : fs::recursive_directory_iterator(startPath)) {
				if (!item.is_regular_file()) {
					continue;
				}

				auto& path = item.path();
				auto filename = path.filename().string();
				if (filename != ".c" ||
					filename != ".cpp")
				{
					continue;
				}

				auto source = ReadFileAtOnce(path);
				HandleInputFile(source);
			}
		} break;

		case IOP_COUNT: break;
	}
}

int main(int argc, char* argv[]) {
	// TODO better arg parser
	//      option 1: use cxxopts and positional arguments
	//      option 1: take one argument only, being a json objecet

	// If no cli is provided (argv[0]), this loop will do nothing
	// Otherwise, start with the 2nd element which is the 1st argument
	for (int i = 1; i < argc; ++i) {
		std::string_view arg(argv[i]);
		auto separatorLoc = arg.find(':');
		if (separatorLoc != std::string_view::npos) {
			auto opcode = ParseInputOpcode(arg.substr(0, separatorLoc));
			auto operand = arg.substr(separatorLoc + 1);
			HandleArgument(opcode, operand);
		}
	}

	return 0;
}