4coder/test/4cpp_new_lexer.h

688 lines
21 KiB
C
Raw Normal View History

2016-03-10 02:59:58 +00:00
// TOP
#ifndef FCPP_NEW_LEXER_INC
#define FCPP_NEW_LEXER_INC
#include "../4cpp_lexer_types.h"
2016-03-10 16:28:44 +00:00
namespace new_lex{
//
2016-03-10 02:59:58 +00:00
#define lexer_link static
lexer_link Cpp_Get_Token_Result
cpp_get_token(Cpp_Token_Stack *token_stack, int pos){
Cpp_Get_Token_Result result = {};
Cpp_Token *token_array = token_stack->tokens;
Cpp_Token *token = 0;
int first = 0;
int count = token_stack->count;
int last = count;
int this_start = 0, next_start = 0;
if (count > 0){
for (;;){
result.token_index = (first + last)/2;
token = token_array + result.token_index;
this_start = token->start;
if (result.token_index + 1 < count){
next_start = (token + 1)->start;
}
else{
next_start = this_start + token->size;
}
if (this_start <= pos && pos < next_start){
break;
}
else if (pos < this_start){
last = result.token_index;
}
else{
first = result.token_index + 1;
}
if (first == last){
result.token_index = first;
break;
}
}
if (result.token_index == count){
--result.token_index;
result.in_whitespace = 1;
}
else{
if (token->start + token->size <= pos){
result.in_whitespace = 1;
}
}
}
else{
result.token_index = -1;
result.in_whitespace = 1;
}
return(result);
}
lexer_link void
cpp_shift_token_starts(Cpp_Token_Stack *stack, int from_token_i, int shift_amount){
Cpp_Token *token = stack->tokens + from_token_i;
int count = stack->count, i;
for (i = from_token_i; i < count; ++i, ++token){
token->start += shift_amount;
}
}
enum Lex_State{
LS_default,
2016-03-10 16:28:44 +00:00
LS_identifier,
LS_char,
LS_string,
LS_number,
2016-03-11 03:33:19 +00:00
LS_float,
2016-03-10 02:59:58 +00:00
LS_comment_pre,
LS_comment,
LS_comment_block,
LS_comment_block_ending,
LS_dot,
2016-03-10 16:28:44 +00:00
LS_ellipsis,
2016-03-10 02:59:58 +00:00
LS_less,
2016-03-10 16:28:44 +00:00
LS_less_less,
2016-03-10 02:59:58 +00:00
LS_more,
2016-03-10 16:28:44 +00:00
LS_more_more,
LS_minus,
LS_arrow,
LS_and,
LS_or,
LS_plus,
LS_colon,
LS_star,
LS_modulo,
LS_caret,
2016-03-11 03:33:19 +00:00
LS_eq,
LS_bang,
2016-03-10 02:59:58 +00:00
};
struct Lex_Data{
int token_start;
int token_end;
int completed;
};
lexer_link Lex_Data
cpp_lex_nonalloc(char *chunk, int file_absolute_pos, int size, Cpp_Token_Stack *token_stack_out){
Cpp_Token *out_tokens = token_stack_out->tokens;
int token_i = token_stack_out->count;
int max_token_i = token_stack_out->max_count;
Cpp_Token token = {};
int pos = file_absolute_pos;
int end_pos = size + file_absolute_pos;
unsigned short state = LS_default;
unsigned short pp_state = 0;
Lex_Data lex_data = {};
int emit_token = 0;
2016-03-10 16:28:44 +00:00
char c = 0;
2016-03-10 02:59:58 +00:00
chunk -= file_absolute_pos;
2016-03-10 16:28:44 +00:00
for (; pos < end_pos && token_i < max_token_i;){
c = chunk[pos];
if (c == ' ' || c == '\n' || c == '\t' || c == '\r' || c == '\f' || c == '\v'){
for (; pos < end_pos;){
c = chunk[pos++];
if (!(c == ' ' || c == '\n' || c == '\t' || c == '\r' || c == '\f' || c == '\v')) break;
}
--pos;
2016-03-10 02:59:58 +00:00
}
lex_data.token_start = pos;
state = LS_default;
emit_token = 0;
2016-03-10 16:28:44 +00:00
for (; emit_token == 0 && pos <= end_pos;){
if (pos < end_pos){
c = chunk[pos++];
}
else{
c = 0;
++pos;
}
2016-03-10 02:59:58 +00:00
switch (state){
case LS_default:
2016-03-10 16:28:44 +00:00
if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_'){
state = LS_identifier;
}
else if (c >= '1' && c <= '9'){
state = LS_number;
}
else
2016-03-10 02:59:58 +00:00
switch (c){
2016-03-10 16:28:44 +00:00
case '\'': state = LS_char; break;
case '"': state = LS_string; break;
2016-03-10 02:59:58 +00:00
case '/': state = LS_comment_pre; break;
2016-03-10 16:28:44 +00:00
case '.': state = LS_dot; break;
case '<': state = LS_less; break;
case '>': state = LS_more; break;
case '-': state = LS_minus; break;
case '&': state = LS_and; break;
case '|': state = LS_or; break;
case '+': state = LS_plus; break;
case ':': state = LS_colon; break;
case '*': state = LS_star; break;
case '%': state = LS_modulo; break;
case '^': state = LS_caret; break;
2016-03-11 03:33:19 +00:00
case '=': state = LS_eq; break;
case '!': state = LS_bang; break;
2016-03-10 02:59:58 +00:00
#define OperCase(op,type) case op: emit_token = 1; break;
OperCase('{', CPP_TOKEN_BRACE_OPEN);
OperCase('}', CPP_TOKEN_BRACE_CLOSE);
OperCase('[', CPP_TOKEN_BRACKET_OPEN);
OperCase(']', CPP_TOKEN_BRACKET_CLOSE);
OperCase('(', CPP_TOKEN_PARENTHESE_OPEN);
OperCase(')', CPP_TOKEN_PARENTHESE_CLOSE);
OperCase('~', CPP_TOKEN_TILDE);
OperCase(',', CPP_TOKEN_COMMA);
2016-03-10 16:28:44 +00:00
OperCase(';', CPP_TOKEN_SEMICOLON);
2016-03-10 02:59:58 +00:00
OperCase('?', CPP_TOKEN_TERNARY_QMARK);
#undef OperCase
}
break;
2016-03-10 16:28:44 +00:00
case LS_identifier:
if (!((c >= '0' && c <= '9') || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_')){
emit_token = 1;
}
break;
case LS_char:
// TODO
break;
case LS_string:
// TODO
break;
case LS_number:
if (c >= '0' && c <= '9'){
state = LS_number;
}
2016-03-11 03:33:19 +00:00
else{
emit_token = 1;
2016-03-10 16:28:44 +00:00
}
break;
2016-03-10 02:59:58 +00:00
case LS_dot:
2016-03-10 16:28:44 +00:00
switch (c){
case '.': state = LS_ellipsis; break;
case '*': emit_token = 1; break;
default: emit_token = 1; break;
}
2016-03-10 02:59:58 +00:00
break;
2016-03-10 16:28:44 +00:00
case LS_ellipsis:
emit_token = 1;
break;
2016-03-10 02:59:58 +00:00
case LS_less:
2016-03-10 16:28:44 +00:00
switch (c){
case '<': state = LS_less_less; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
2016-03-10 02:59:58 +00:00
break;
2016-03-10 16:28:44 +00:00
case LS_less_less:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
2016-03-10 02:59:58 +00:00
case LS_more:
2016-03-10 16:28:44 +00:00
switch (c){
case '>': state = LS_more_more; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
2016-03-10 02:59:58 +00:00
break;
2016-03-10 16:28:44 +00:00
case LS_more_more:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
2016-03-10 02:59:58 +00:00
case LS_comment_pre:
switch (c){
case '/': state = LS_comment; break;
case '*': state = LS_comment_block; break;
2016-03-10 16:28:44 +00:00
case '=': emit_token = 1; break;
default: emit_token = 1; break;
2016-03-10 02:59:58 +00:00
}
break;
case LS_comment:
switch (c){
case '\n': emit_token = 1; break;
}
break;
case LS_comment_block:
switch (c){
case '*': state = LS_comment_block_ending; break;
}
break;
case LS_comment_block_ending:
switch (c){
case '*': state = LS_comment_block_ending; break;
case '/': emit_token = 1; break;
default: state = LS_comment_block; break;
}
break;
2016-03-10 16:28:44 +00:00
case LS_minus:
switch (c){
case '>': state = LS_arrow; break;
case '-': emit_token = 1; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_arrow:
switch (c){
case '*': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_and:
switch (c){
case '&': emit_token = 1; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_or:
switch (c){
case '|': emit_token = 1; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_plus:
switch (c){
case '+': emit_token = 1; break;
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_colon:
switch (c){
case ':': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_star:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_modulo:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_caret:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
2016-03-11 03:33:19 +00:00
case LS_eq:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
case LS_bang:
switch (c){
case '=': emit_token = 1; break;
default: emit_token = 1; break;
}
break;
2016-03-10 02:59:58 +00:00
}
}
if (emit_token){
lex_data.token_end = pos;
switch (state){
case LS_default:
2016-03-10 16:28:44 +00:00
switch (c){
2016-03-10 02:59:58 +00:00
#define OperCase(op,t) case op: token.type = t; break;
OperCase('{', CPP_TOKEN_BRACE_OPEN);
OperCase('}', CPP_TOKEN_BRACE_CLOSE);
OperCase('[', CPP_TOKEN_BRACKET_OPEN);
OperCase(']', CPP_TOKEN_BRACKET_CLOSE);
OperCase('(', CPP_TOKEN_PARENTHESE_OPEN);
OperCase(')', CPP_TOKEN_PARENTHESE_CLOSE);
OperCase('~', CPP_TOKEN_TILDE);
OperCase(',', CPP_TOKEN_COMMA);
2016-03-10 16:28:44 +00:00
OperCase(';', CPP_TOKEN_SEMICOLON);
2016-03-10 02:59:58 +00:00
OperCase('?', CPP_TOKEN_TERNARY_QMARK);
#undef OperCase
}
token.flags = CPP_TFLAG_IS_OPERATOR;
break;
2016-03-10 16:28:44 +00:00
case LS_identifier:
token.type = CPP_TOKEN_IDENTIFIER;
token.flags = 0;
--lex_data.token_end;
--pos;
break;
2016-03-11 03:33:19 +00:00
case LS_number:
token.type = CPP_TOKEN_INTEGER_CONSTANT;
token.flags = 0;
--lex_data.token_end;
--pos;
break;
2016-03-10 16:28:44 +00:00
case LS_comment_pre:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_DIVEQ; break;
default:
token.type = CPP_TOKEN_DIV;
--lex_data.token_end;
--pos;
break;
}
break;
2016-03-10 02:59:58 +00:00
case LS_comment: case LS_comment_block_ending:
token.type = CPP_TOKEN_COMMENT;
token.flags = 0;
2016-03-10 16:28:44 +00:00
c = chunk[--lex_data.token_end];
2016-03-10 02:59:58 +00:00
while (c == ' ' || c == '\n' || c == '\t' || c == '\r' || c == '\v' || c == '\f'){
2016-03-10 16:28:44 +00:00
--lex_data.token_end;
c = chunk[lex_data.token_end];
}
++lex_data.token_end;
break;
case LS_dot:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '*': token.type = CPP_TOKEN_PTRDOT; break;
default:
token.type = CPP_TOKEN_DOT;
--lex_data.token_end;
2016-03-10 02:59:58 +00:00
--pos;
2016-03-10 16:28:44 +00:00
break;
}
break;
case LS_ellipsis:
switch (c){
case '.':
token.flags = CPP_TFLAG_IS_OPERATOR;
token.type = CPP_TOKEN_ELLIPSIS;
break;
default:
token.type = CPP_TOKEN_JUNK;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_less:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_LESSEQ; break;
default:
token.type = CPP_TOKEN_LESS;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_less_less:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_LSHIFTEQ; break;
default:
token.type = CPP_TOKEN_LSHIFT;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_more:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_GRTREQ; break;
default:
token.type = CPP_TOKEN_GRTR;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_more_more:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_RSHIFTEQ; break;
default:
token.type = CPP_TOKEN_RSHIFT;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_minus:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '-': token.type = CPP_TOKEN_DECREMENT; break;
case '=': token.type = CPP_TOKEN_SUBEQ; break;
default:
token.type = CPP_TOKEN_MINUS;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_arrow:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '*': token.type = CPP_TOKEN_PTRARROW; break;
default:
token.type = CPP_TOKEN_ARROW;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_and:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '&': token.type = CPP_TOKEN_AND; break;
case '=': token.type = CPP_TOKEN_ANDEQ; break;
default:
token.type = CPP_TOKEN_AMPERSAND;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_or:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '|': token.type = CPP_TOKEN_OR; break;
case '=': token.type = CPP_TOKEN_OREQ; break;
default:
token.type = CPP_TOKEN_BIT_OR;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_plus:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '+': token.type = CPP_TOKEN_INCREMENT; break;
case '=': token.type = CPP_TOKEN_ADDEQ; break;
default:
token.type = CPP_TOKEN_PLUS;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_colon:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case ':': token.type = CPP_TOKEN_SCOPE; break;
default:
token.type = CPP_TOKEN_COLON;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_star:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_MULEQ; break;
default:
token.type = CPP_TOKEN_STAR;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_modulo:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_MODEQ; break;
default:
token.type = CPP_TOKEN_MOD;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_caret:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
2016-03-11 03:33:19 +00:00
case '=': token.type = CPP_TOKEN_XOREQ; break;
2016-03-10 16:28:44 +00:00
default:
token.type = CPP_TOKEN_BIT_XOR;
--lex_data.token_end;
--pos;
break;
2016-03-10 02:59:58 +00:00
}
break;
2016-03-11 03:33:19 +00:00
case LS_eq:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_EQEQ; break;
default:
token.type = CPP_TOKEN_EQ;
--lex_data.token_end;
--pos;
break;
}
break;
case LS_bang:
token.flags = CPP_TFLAG_IS_OPERATOR;
switch (c){
case '=': token.type = CPP_TOKEN_NOTEQ; break;
default:
token.type = CPP_TOKEN_BIT_NOT;
--lex_data.token_end;
--pos;
break;
}
break;
2016-03-10 02:59:58 +00:00
}
2016-03-10 16:28:44 +00:00
2016-03-10 02:59:58 +00:00
token.start = lex_data.token_start;
2016-03-10 16:28:44 +00:00
token.size = lex_data.token_end - lex_data.token_start;
2016-03-10 02:59:58 +00:00
token.state_flags = pp_state;
out_tokens[token_i++] = token;
}
}
token_stack_out->count = token_i;
if (pos == end_pos) lex_data.completed = 1;
return(lex_data);
}
2016-03-10 16:28:44 +00:00
}
2016-03-10 02:59:58 +00:00
#endif
// BOTTOM