mips/masm/parse.c

#include <mlimits.h>
#include <stdint.h>
#include <merror.h>
#include <stddef.h>

#include "lex.h"
#include "parse.h"

///
/// Token Functions
/// either get a token, peek a token,
/// or assert a token was returned
///

/* get the next token from the lexer */
static int next_token(struct parser *parser, struct token *tok)
{
	// return peeked first
	if (parser->peek.type != TOK_EOF) {
		if (tok != NULL)
			*tok = parser->peek;
		else
			token_free(&parser->peek);

		parser->peek.type = TOK_EOF;
		return M_SUCCESS;
	}

	// get next token
	struct token token;
	if (lexer_next(&parser->lexer, &token))
		return M_ERROR;

	// return value if given pointer
	// else free
	if (tok != NULL) {
		*tok = token;
	} else {
		token_free(&token);
	}

	return M_SUCCESS;
}

/* peek the next token from the lexer */
static int peek_token(struct parser *parser, struct token *tok)
{
	// if we dont have a saved token
	// get the next one
	if (parser->peek.type == TOK_EOF) {
		if (next_token(parser, &parser->peek))
			return M_ERROR;
	}

	// return it if we were given
	// a pointer
	if (tok != NULL)
		*tok = parser->peek;

	return M_SUCCESS;
}

/* get the next token from the lexer, and assert its of type <type> */
static int assert_token(struct parser *parser, enum token_type type,
		 struct token *tok)
{
	// get next token
	struct token token;
	if (next_token(parser, &token))
		return M_ERROR;

	// assert its of type <type>
	if (token.type != type) {
		ERROR_POS(token, "expected a token of type '%s', got '%s'",
			token_str(type), token_str(token.type));
		token_free(&token);
		return M_ERROR;
	}

	// return value if given pointer
	// else free
	if (tok != NULL) {
		*tok = token;
	} else {
		token_free(&token);
	}

	return M_SUCCESS;
}

/* get the next token from the lexer, and assert its of type NL */
static int assert_eol(struct parser *parser)
{
	struct token token;
	if (next_token(parser, &token))
		return M_ERROR;
	if (token.type != TOK_NL && token.type != TOK_EOF) {
		ERROR_POS(token, "expected a new line or end of file, got '%s'",
			token_str(token.type));
		return M_ERROR;
	}
	token_free(&token);
	return M_SUCCESS;
}

/* peek the next token and return SUCCESS on eol */
static int peek_eol(struct parser *parser)
{
	struct token token;
	if (peek_token(parser, &token))
		return M_ERROR;
	int res = (token.type == TOK_NL || token.type == TOK_EOF) ?
		M_SUCCESS : M_ERROR;
	return res;
}

///
/// PARSER FUNCTIONS
/// parses each type of expression
///

static int parse_directive_whb(struct parser *parser, void *data, uint32_t *res,
			       size_t length, size_t max_size)
{
	struct token token;
	int len = 0;

	while (1) {
		if (peek_eol(parser) == M_SUCCESS)
			break;

		if (assert_token(parser, TOK_NUMBER, &token))
			return M_ERROR;

		if ((uint64_t)token.number > max_size) {
			ERROR_POS(token, "number cannot exceed max size of %zu",
				  max_size);
			return M_ERROR;
		}

		if (len >= MAX_ARG_LENGTH) {
			ERROR_POS(token, "exceeded max argument length for "
				  "directives");
			return M_ERROR;
		}

		// BUG: does this only work on little endian???
		memcpy((uint8_t *) data + (len++ * length), &token.number,
		       max_size);
	}

	*res = len;
	return M_SUCCESS;
}

static int parse_immd(struct parser *parser, uint16_t *num)
{
	struct token token;
	if (assert_token(parser, TOK_NUMBER, &token))
		return M_ERROR;

	// TOK_NUMBER does not need to be freed
	*num = token.number;
	return M_SUCCESS;
}

static int parse_ident(struct parser *parser, struct string *ident)
{
	struct token token;
	if (assert_token(parser, TOK_IDENT, &token))
		return M_ERROR;
	string_move(ident, &token.string);
	return M_SUCCESS;
}

static int parse_string(struct parser *parser, struct string *string)
{
	struct token token;
	if (assert_token(parser, TOK_STRING, &token))
		return M_ERROR;
	string_move(string, &token.string);
	return M_SUCCESS;
}

/* parses a directive */
static int parse_directive(struct parser *parser, struct string *name,
			   struct expr_directive *expr)
{
	#define CHK(n) if (strcmp(name->str, #n) == 0)

	CHK(.align) {
		expr->type = EXPR_DIRECTIVE_ALIGN;
		return parse_immd(parser, &expr->align);
	} else CHK(.space) {
		expr->type = EXPR_DIRECTIVE_SPACE;
		return parse_immd(parser, &expr->space);
	} else CHK(.word) {
		expr->type = EXPR_DIRECTIVE_WORD;
		return parse_directive_whb(parser, expr->words, &expr->len,
					   sizeof(uint32_t), UINT32_MAX);
	} else CHK(.half) {
		expr->type = EXPR_DIRECTIVE_HALF;
		return parse_directive_whb(parser, expr->halfs, &expr->len,
					   sizeof(uint16_t), UINT16_MAX);
	} else CHK(.byte) {
		expr->type = EXPR_DIRECTIVE_BYTE;
		return parse_directive_whb(parser, expr->bytes, &expr->len,
					   sizeof(uint8_t), UINT8_MAX);
	} else CHK(.extern) {
		expr->type = EXPR_DIRECTIVE_EXTERN;
		return parse_ident(parser, &expr->label);
	} else CHK(.globl) {
		expr->type = EXPR_DIRECTIVE_GLOBL;
		return parse_ident(parser, &expr->label);
	} else CHK(.ascii) {
		expr->type = EXPR_DIRECTIVE_ASCII;
		return parse_string(parser, &expr->string);
	} else CHK(.asciiz) {
		expr->type = EXPR_DIRECTIVE_ASCIIZ;
		return parse_string(parser, &expr->string);
	} else {
		expr->type = EXPR_DIRECTIVE_SECTION;
		string_move(&expr->section, name);
		return M_SUCCESS;
	}

	#undef CHK
}

static int parse_constant(struct parser *parser, struct string *name,
			  struct expr_const *constant)
{
	if (assert_token(parser, TOK_EQUAL, NULL))
		return M_ERROR;

	struct token token;
	if (assert_token(parser, TOK_NUMBER, &token))
		return M_ERROR;

	string_move(&constant->name, name);
	constant->num = token.number;
	return M_SUCCESS;
}

static int parse_offset(struct parser *parser,
			struct expr_ins_arg *arg,
			uint64_t immd)
{
	// the immediate has already been parsed
	// now parse (REG)

	if (assert_token(parser, TOK_LPAREN, NULL))
		return M_ERROR;

	struct token token;
	if (assert_token(parser, TOK_REG, &token))
		return M_ERROR;

	// set values
	string_move(&arg->offset.reg, &token.string);
	arg->offset.immd = immd;

	if (assert_token(parser, TOK_RPAREN, NULL)) {
		string_free(&arg->offset.reg);
		return M_ERROR;
	}

	return M_SUCCESS;
}

static int parse_instruction_arg(struct parser *parser,
				 struct expr_ins_arg *arg)
{
	// allowed token matches:
	//
	// register:
	// REG
	//
	// label:
	// IDENT
	//
	// immediate:
	// IMMD
	//
	// offset:
	// (REG)
	// IMMD(REG)

	struct token token;
	if (peek_token(parser, &token))
		return M_ERROR;

	// if its a left paren, were parsing
	// an offset
	if (token.type == TOK_LPAREN) {
		arg->type = EXPR_INS_ARG_OFFSET;
		return parse_offset(parser, arg, 0);
	}

	// token must now be either a number (immediate)
	// register, or label,,,
	// ... take ownership of the next token
	if (next_token(parser, &token))
		return M_ERROR;

	// if its a register... return
	// there are no other pathways
	if (token.type == TOK_REG) {
		arg->type = EXPR_INS_ARG_REGISTER;
		string_move(&arg->reg, &token.string);
		return M_SUCCESS;
	}

	// if it is a label... return
	// therea are no other pathways
	if (token.type == TOK_IDENT) {
		arg->type = EXPR_INS_ARG_LABEL;
		string_move(&arg->label, &token.string);
		return M_SUCCESS;
	}

	// now it must be a number...
	// throw a error if its now
	if (token.type != TOK_NUMBER) {
		ERROR_POS(token, "expected number, got %s",
                        token_str(token.type));
		token_free(&token);
		return M_ERROR;
	}

	uint64_t immd = (uint64_t)token.number;
	// now if the next token is a lparen
	// parse offset, else return immd
	if (peek_token(parser, &token))
		return M_ERROR;

	if (token.type == TOK_LPAREN) {
		arg->type = EXPR_INS_ARG_OFFSET;
		return parse_offset(parser, arg, immd);
	} else {
		arg->type = EXPR_INS_ARG_IMMEDIATE;
		arg->immd = immd;
		return M_SUCCESS;
	}

}

static int parse_instruction(struct parser *parser, struct string *name,
			     struct expr_ins *ins)
{
	int len = 0;

	if (peek_eol(parser) == M_SUCCESS)
		goto skip_args;

	while (1) {
		if (len >= MAX_ARG_LENGTH) {
			ERROR_POS(parser->lexer,
				  "reached max argument length");
			return M_ERROR;
		}


		if (parse_instruction_arg(parser, &ins->args[len++]))
			return M_ERROR;

		if (peek_eol(parser) == M_SUCCESS)
			break;

		if (assert_token(parser, TOK_COMMA, NULL))
			return M_ERROR;
	}

skip_args:

	string_move(&ins->name, name);
	ins->args_len = len;
	return M_SUCCESS;
}

/* gets the next value from the parser */
int parser_next(struct parser *parser, struct expr *expr)
{
	// the next token being looked at
	struct token token = {
		.type = TOK_NL
	};

	// the result to return
	int res = M_SUCCESS;

	// skip all new lines
	while (1) {
		if (next_token(parser, &token))
			return M_ERROR;
		if (token.type != TOK_NL)
			break;
		token_free(&token);
	}

	expr->line_no = parser->lexer.y;
	expr->byte_start = token.off;
	expr->byte_end = token.off;

	// if EOF, return M_EOF
	if (token.type == TOK_EOF)
		return M_EOF;

	// when a ident ends with a colon
	// parse a lebel
	else if (token.type == TOK_LABEL) {
		expr->type = EXPR_LABEL;
		// label now owns string
		string_move(&expr->label, &token.string);
	}

	// when a ident starts with a dot
	// parse a directive
	else if (token.type == TOK_DIRECTIVE) {
		expr->type = EXPR_DIRECTIVE;
		res = parse_directive(parser, &token.string, &expr->directive);
	}

	// peek the next token:
	// 1. = means parse constant
	// 2. else parse instruction
	else {
		if (token.type != TOK_IDENT) {
			ERROR_POS(token, "expected ident, got %s",
				  token_str(token.type));
			token_free(&token);
			return M_ERROR;
		}

		struct token peek;
		if (peek_token(parser, &peek)) {
			token_free(&token);
			return M_ERROR;
		}

		if (peek.type == TOK_EQUAL) {
			expr->type = EXPR_CONSTANT;
			res = parse_constant(parser, &token.string,
					     &expr->constant);
		} else {
			expr->type = EXPR_INS;
			res = parse_instruction(parser, &token.string,
						&expr->instruction);
		}
	}

	// update byte end for expr
	expr->byte_end = ftell(parser->lexer.file);

	// free tokens
	token_free(&token);

	// everything must end in a new line
	if (res == M_SUCCESS && assert_eol(parser))
		return M_ERROR;

	return res;
}

int parser_init(const char *file, struct parser *parser)
{
	parser->peek.type = TOK_EOF;
	if (lexer_init(file, &parser->lexer))
		return M_ERROR;
	return M_SUCCESS;
}


void parser_free(struct parser *parser)
{
	token_free(&parser->peek);
	lexer_free(&parser->lexer);
}

static inline void expr_directive_free(struct expr_directive *dir)
{
	switch (dir->type) {
	case EXPR_DIRECTIVE_SECTION:
		string_free(&dir->section);
		break;
	case EXPR_DIRECTIVE_EXTERN:
	case EXPR_DIRECTIVE_GLOBL:
		string_free(&dir->label);
		break;
	case EXPR_DIRECTIVE_ASCII:
	case EXPR_DIRECTIVE_ASCIIZ:
		string_free(&dir->string);
		break;
	default:
	}
}

static inline void expr_ins_arg_free(struct expr_ins_arg *arg)
{
	switch (arg->type) {
	case EXPR_INS_ARG_REGISTER:
		string_free(&arg->reg);
		break;
	case EXPR_INS_ARG_IMMEDIATE:
		break;
	case EXPR_INS_ARG_LABEL:
		string_free(&arg->label);
		break;
	case EXPR_INS_ARG_OFFSET:
		string_free(&arg->offset.reg);
		break;
	}
}

void expr_free(struct expr *expr)
{
	switch (expr->type) {
	case EXPR_DIRECTIVE:
		expr_directive_free(&expr->directive);
		break;
	case EXPR_CONSTANT:
		string_free(&expr->constant.name);
		break;
	case EXPR_INS:
		string_free(&expr->instruction.name);
		for (uint32_t i = 0; i < expr->instruction.args_len; i++)
			expr_ins_arg_free(&expr->instruction.args[i]);
		break;
	case EXPR_LABEL:
		string_free(&expr->label);
		break;
	}
}