parse.c

   1 /*
   2  * Copyright (C) 2012
   3  *      Dale Weiler
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a copy of
   6  * this software and associated documentation files (the "Software"), to deal in
   7  * the Software without restriction, including without limitation the rights to
   8  * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
   9  * of the Software, and to permit persons to whom the Software is furnished to do
  10  * so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice shall be included in all
  13  * copies or substantial portions of the Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  18  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  */
  23 #include <limits.h>
  24 #include <stdlib.h>
  25 #include <string.h>
  26 #include "gmqcc.h"
  27
  28 /*
  29  * These are not lexical tokens:  These are parse tree types.  Most people
  30  * perform tokenizing on language punctuation which is wrong.  That stuff
  31  * is technically already tokenized, it just needs to be parsed into a tree
  32  */
  33 #define PARSE_TYPE_DO       0
  34 #define PARSE_TYPE_ELSE     1
  35 #define PARSE_TYPE_IF       2
  36 #define PARSE_TYPE_WHILE    3
  37 #define PARSE_TYPE_BREAK    4
  38 #define PARSE_TYPE_CONTINUE 5
  39 #define PARSE_TYPE_RETURN   6
  40 #define PARSE_TYPE_GOTO     7
  41 #define PARSE_TYPE_FOR      8
  42 #define PARSE_TYPE_VOID     9
  43 #define PARSE_TYPE_STRING   10
  44 #define PARSE_TYPE_FLOAT    11
  45 #define PARSE_TYPE_VECTOR   12
  46 #define PARSE_TYPE_ENTITY   13
  47 #define PARSE_TYPE_LAND     14
  48 #define PARSE_TYPE_LOR      15
  49 #define PARSE_TYPE_LTEQ     16
  50 #define PARSE_TYPE_GTEQ     17
  51 #define PARSE_TYPE_EQEQ     18
  52 #define PARSE_TYPE_LNEQ     19
  53 #define PARSE_TYPE_COMMA    20
  54 #define PARSE_TYPE_LNOT     21
  55 #define PARSE_TYPE_STAR     22
  56 #define PARSE_TYPE_DIVIDE   23
  57 #define PARSE_TYPE_LPARTH   24
  58 #define PARSE_TYPE_RPARTH   25
  59 #define PARSE_TYPE_MINUS    26
  60 #define PARSE_TYPE_ADD      27
  61 #define PARSE_TYPE_EQUAL    28
  62 #define PARSE_TYPE_LBS      29
  63 #define PARSE_TYPE_RBS      30
  64 #define PARSE_TYPE_ELIP     31
  65 #define PARSE_TYPE_DOT      32
  66 #define PARSE_TYPE_LT       33
  67 #define PARSE_TYPE_GT       34
  68 #define PARSE_TYPE_BAND     35
  69 #define PARSE_TYPE_BOR      36
  70 #define PARSE_TYPE_DONE     37
  71 #define PARSE_TYPE_IDENT    38
  72
  73 int parse[PARSE_TYPE_IDENT] = {
  74         0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
  75         0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
  76         0,0,0,0,0,0,0,0
  77 };
  78
  79 /*
  80  * Adds a parse type to the parse tree, this is where all the hard
  81  * work actually begins.
  82  */
  83 #define PARSE_TREE_ADD(X)                                        \
  84         do {                                                         \
  85                 parsetree->next       = mem_a(sizeof(struct parsenode)); \
  86                 parsetree->next->next = NULL;                            \
  87                 parsetree->next->type = (X);                             \
  88                 parsetree             = parsetree->next;                 \
  89         } while (0)
  90
  91 #define PARSE_TREE_CHK(X,Y,Z)                                        \
  92         do { \
  93                 if(parse[X]) { \
  94                         error(ERROR_PARSE, "Expected %c for %c\n", Y, Z); \
  95                         exit (1); \
  96                 } \
  97         } while (0)
  98
  99 #define PARSE_TREE_PUT(X) do { parse[X] = 1; } while (0)
 100
 101 /*
 102  * This is all the punctuation handled in the parser, these don't
 103  * need tokens, they're already tokens.
 104  */
 105 #if 0
 106         "&&", "||", "<=", ">=", "==", "!=", ";", ",", "!", "*",
 107         "/" , "(" , ")" , "-" , "+" , "=" , "[" , "]", "{", "}", "...",
 108         "." , "<" , ">" , "&" , "|" ,
 109 #endif
 110
 111 #define STORE(X) {  \
 112         printf(X);      \
 113         break;          \
 114 }
 115
 116 void parse_debug(struct parsenode *tree) {
 117         while (tree) {
 118                 switch (tree->type) {
 119                         case PARSE_TYPE_ADD:       STORE("OPERATOR:  ADD    \n");
 120                         case PARSE_TYPE_BAND:      STORE("OPERATOR:  BITAND \n");
 121                         case PARSE_TYPE_BOR:       STORE("OPERATOR:  BITOR  \n");
 122                         case PARSE_TYPE_COMMA:     STORE("OPERATOR:  SEPERATOR\n");
 123                         case PARSE_TYPE_DOT:       STORE("OPERATOR:  DOT\n");
 124                         case PARSE_TYPE_DIVIDE:    STORE("OPERATOR:  DIVIDE\n");
 125                         case PARSE_TYPE_EQUAL:     STORE("OPERATOR:  ASSIGNMENT\n");
 126
 127                         case PARSE_TYPE_BREAK:     STORE("STATEMENT: BREAK  \n");
 128                         case PARSE_TYPE_CONTINUE:  STORE("STATEMENT: CONTINUE\n");
 129                         case PARSE_TYPE_GOTO:      STORE("STATEMENT: GOTO\n");
 130                         case PARSE_TYPE_RETURN:    STORE("STATEMENT: RETURN\n");
 131                         case PARSE_TYPE_DONE:      STORE("STATEMENT: DONE\n");
 132
 133                         case PARSE_TYPE_VOID:      STORE("DECLTYPE:  VOID\n");
 134                         case PARSE_TYPE_STRING:    STORE("DECLTYPE:  STRING\n");
 135                         case PARSE_TYPE_ELIP:      STORE("DECLTYPE:  VALIST\n");
 136                         case PARSE_TYPE_ENTITY:    STORE("DECLTYPE:  ENTITY\n");
 137                         case PARSE_TYPE_FLOAT:     STORE("DECLTYPE:  FLOAT\n");
 138                         case PARSE_TYPE_VECTOR:    STORE("DECLTYPE:  VECTOR\n");
 139
 140                         case PARSE_TYPE_GT:        STORE("TEST:      GREATER THAN\n");
 141                         case PARSE_TYPE_LT:        STORE("TEST:      LESS THAN\n");
 142                         case PARSE_TYPE_GTEQ:      STORE("TEST:      GREATER THAN OR EQUAL\n");
 143                         case PARSE_TYPE_LTEQ:      STORE("TEST:      LESS THAN OR EQUAL\n");
 144                         case PARSE_TYPE_LNEQ:      STORE("TEST:      NOT EQUAL\n");
 145                         case PARSE_TYPE_EQEQ:      STORE("TEST:      EQUAL-EQUAL\n");
 146
 147                         case PARSE_TYPE_LBS:       STORE("BLOCK:     BEG\n");
 148                         case PARSE_TYPE_RBS:       STORE("BLOCK:     END\n");
 149                         case PARSE_TYPE_ELSE:      STORE("BLOCK:     ELSE\n");
 150                         case PARSE_TYPE_IF:        STORE("BLOCK:     IF\n");
 151
 152                         case PARSE_TYPE_LAND:      STORE("LOGICAL:   AND\n");
 153                         case PARSE_TYPE_LNOT:      STORE("LOGICAL:   NOT\n");
 154                         case PARSE_TYPE_LOR:       STORE("LOGICAL:   OR\n");
 155
 156                         case PARSE_TYPE_LPARTH:    STORE("PARTH:     BEG\n");
 157                         case PARSE_TYPE_RPARTH:    STORE("PARTH:     END\n");
 158
 159                         case PARSE_TYPE_WHILE:     STORE("LOOP:      WHILE\n");
 160                         case PARSE_TYPE_FOR:       STORE("LOOP:      FOR\n");
 161                         case PARSE_TYPE_DO:        STORE("LOOP:      DO\n");
 162
 163                         //case PARSE_TYPE_IDENT:     STORE("IDENT:     ???\n");
 164                 }
 165                 tree = tree->next;
 166         }
 167 }
 168
 169 /*
 170  * Performs a parse operation:  This is a macro to prevent bugs, if the
 171  * calls to lex_token are'nt exactly enough to feed to the end of the
 172  * actual lexees for the current thing that is being parsed, the state
 173  * of the next iteration in the creation of the parse tree will be wrong
 174  * and everything will fail.
 175  */
 176 #define PARSE_PERFORM(X,C) {     \
 177     token = lex_token(file);     \
 178     { C }                        \
 179     while (token != '\n') {      \
 180             token = lex_token(file); \
 181     }                            \
 182     PARSE_TREE_ADD(X);           \
 183     break;                       \
 184 }
 185
 186 void parse_clear(struct parsenode *tree) {
 187         if (!tree) return;
 188         struct parsenode *temp = NULL;
 189         while (tree != NULL) {
 190                 temp = tree;
 191                 tree = tree->next;
 192                 mem_d (temp);
 193         }
 194
 195         /* free any potential typedefs */
 196         typedef_clear();
 197 }
 198
 199 /*
 200  * Generates a parse tree out of the lexees generated by the lexer.  This
 201  * is where the tree is built.  This is where valid check is performed.
 202  */
 203 int parse_tree(struct lex_file *file) {
 204         struct parsenode *parsetree = NULL;
 205         struct parsenode *parseroot = NULL;
 206
 207         /*
 208          * Allocate memory for our parse tree:
 209          * the parse tree is just a singly linked list which will contain
 210          * all the data for code generation.
 211          */
 212         if (!parseroot) {
 213                 parseroot = mem_a(sizeof(struct parsenode));
 214                 if (!parseroot)
 215                         return error(ERROR_INTERNAL, "Ran out of memory", " ");
 216                 parsetree       = parseroot;
 217                 parsetree->type = -1; /* not a valid type -- root element */
 218         }
 219
 220         int     token = 0;
 221         while ((token = lex_token(file)) != ERROR_LEX      && \
 222                     token                    != ERROR_COMPILER && \
 223                     token                    != ERROR_INTERNAL && \
 224                     token                    != ERROR_PARSE    && \
 225                     token                    != ERROR_PREPRO   && file->length >= 0) {
 226                 switch (token) {
 227                         case TOKEN_IF:
 228                                 token = lex_token(file);
 229                                 if (token != '(')
 230                                         error(ERROR_PARSE, "Expected `(` on if statement:\n");
 231                                 PARSE_TREE_ADD(PARSE_TYPE_IF);
 232                                 PARSE_TREE_ADD(PARSE_TYPE_LPARTH);
 233                                 PARSE_TREE_CHK(PARSE_TYPE_LPARTH, ')', '(');
 234                                 PARSE_TREE_PUT(PARSE_TYPE_LPARTH);
 235                                 break;
 236                         case TOKEN_ELSE:
 237                                 token = lex_token(file);
 238                                 PARSE_TREE_ADD(PARSE_TYPE_ELSE);
 239                                 break;
 240                         case TOKEN_FOR:
 241                                 //token = lex_token(file);
 242                                 while ((token == ' ' || token == '\n') && file->length >= 0)
 243                                         token = lex_token(file);
 244                                 PARSE_TREE_ADD(PARSE_TYPE_FOR);
 245                                 break;
 246
 247                         /*
 248                          * This is a quick and easy way to do typedefs at parse time
 249                          * all power is in typedef_add(), in typedef.c.  We handle
 250                          * the tokens accordingly here.
 251                          */
 252                         case TOKEN_TYPEDEF: {
 253                                 char *f,*t;
 254
 255                                 token = lex_token(file);
 256                                 token = lex_token(file); f = util_strdup(file->lastok);
 257                                 token = lex_token(file);
 258                                 token = lex_token(file); t = util_strdup(file->lastok);
 259
 260                                 typedef_add(f, t);
 261
 262                                 mem_d(f);
 263                                 mem_d(t);
 264
 265                                 while (token != '\n')
 266                                         token = lex_token(file);
 267                                 break;
 268                         }
 269
 270                         /*
 271                          * Returns are addable as-is, statement checking is during
 272                          * the actual parse tree check.
 273                          */
 274                         case TOKEN_RETURN:
 275                                 token = lex_token(file);
 276                                 PARSE_TREE_ADD(PARSE_TYPE_RETURN);
 277                                 break;
 278                         case TOKEN_CONTINUE:
 279                                 PARSE_TREE_ADD(PARSE_TYPE_CONTINUE);
 280                                 break;
 281
 282                         case TOKEN_DO:        PARSE_PERFORM(PARSE_TYPE_DO,      {});
 283                         case TOKEN_WHILE:     PARSE_PERFORM(PARSE_TYPE_WHILE,   {});
 284                         case TOKEN_BREAK:     PARSE_PERFORM(PARSE_TYPE_BREAK,   {});
 285                         case TOKEN_GOTO:      PARSE_PERFORM(PARSE_TYPE_GOTO,    {});
 286                         case TOKEN_VOID:      PARSE_PERFORM(PARSE_TYPE_VOID,    {});
 287                         case TOKEN_STRING:    PARSE_PERFORM(PARSE_TYPE_STRING,  {});
 288                         case TOKEN_FLOAT:     PARSE_PERFORM(PARSE_TYPE_FLOAT,   {});
 289                         case TOKEN_VECTOR:    PARSE_PERFORM(PARSE_TYPE_VECTOR,  {});
 290                         case TOKEN_ENTITY:    PARSE_PERFORM(PARSE_TYPE_ENTITY,  {});
 291
 292                         /*
 293                          * From here down is all language punctuation:  There is no
 294                          * need to actual create tokens from these because they're already
 295                          * tokenized as these individual tokens (which are in a special area
 296                          * of the ascii table which doesn't conflict with our other tokens
 297                          * which are higer than the ascii table.)
 298                          */
 299                         case '#':
 300                                 /*
 301                                  * Skip the preprocessor for now:  We'll implement our own
 302                                  * eventually.  For now we need to make sure directives are
 303                                  * not accidently tokenized.
 304                                  */
 305                                 token = lex_token(file);
 306                                 token = lex_token(file);
 307
 308                                 /* skip all tokens to end of directive */
 309                                 while (token != '\n')
 310                                         token = lex_token(file);
 311                                 break;
 312
 313                         case '.':
 314                                 //token = lex_token(file);
 315                                 PARSE_TREE_ADD(PARSE_TYPE_DOT);
 316                                 break;
 317                         case '(':
 318                                 //token = lex_token(file);
 319                                 PARSE_TREE_PUT(PARSE_TYPE_LPARTH);
 320                                 PARSE_TREE_ADD(PARSE_TYPE_LPARTH);
 321                                 break;
 322                         case ')':
 323                                 //token = lex_token(file);
 324                                 parse[PARSE_TYPE_LPARTH] = 0;
 325                                 PARSE_TREE_PUT(PARSE_TYPE_RPARTH);
 326                                 PARSE_TREE_ADD(PARSE_TYPE_RPARTH);
 327                                 break;
 328
 329                         case '&':                               /* &  */
 330                                 token = lex_token(file);
 331                                 if (token == '&') { /* && */
 332                                         token = lex_token(file);
 333                                         PARSE_TREE_ADD(PARSE_TYPE_LAND);
 334                                         break;
 335                                 }
 336                                 PARSE_TREE_ADD(PARSE_TYPE_BAND);
 337                                 break;
 338                         case '|':                               /* |  */
 339                                 token = lex_token(file);
 340                                 if (token == '|') { /* || */
 341                                         token = lex_token(file);
 342                                         PARSE_TREE_ADD(PARSE_TYPE_LOR);
 343                                         break;
 344                                 }
 345                                 PARSE_TREE_ADD(PARSE_TYPE_BOR);
 346                                 break;
 347                         case '!':                               /* !  */
 348                                 token = lex_token(file);
 349                                 if (token == '=') { /* != */
 350                                         token = lex_token(file);
 351                                         PARSE_TREE_ADD(PARSE_TYPE_LNEQ);
 352                                         break;
 353                                 }
 354                                 PARSE_TREE_ADD(PARSE_TYPE_LNOT);
 355                                 break;
 356                         case '<':                               /* <  */
 357                                 token = lex_token(file);
 358                                 if (token == '=') { /* <= */
 359                                         token = lex_token(file);
 360                                         PARSE_TREE_ADD(PARSE_TYPE_LTEQ);
 361                                         break;
 362                                 }
 363                                 PARSE_TREE_ADD(PARSE_TYPE_LT);
 364                                 break;
 365                         case '>':                               /* >  */
 366                                 token = lex_token(file);
 367                                 if (token == '=') { /* >= */
 368                                         token = lex_token(file);
 369                                         PARSE_TREE_ADD(PARSE_TYPE_GTEQ);
 370                                         break;
 371                                 }
 372                                 PARSE_TREE_ADD(PARSE_TYPE_GT);
 373                                 break;
 374                         case '=':                               /* =  */
 375                                 token = lex_token(file);
 376                                 if (token == '=') { /* == */
 377                                         token = lex_token(file);
 378                                         PARSE_TREE_ADD(PARSE_TYPE_EQEQ);
 379                                         break;
 380                                 }
 381                                 PARSE_TREE_ADD(PARSE_TYPE_EQUAL);
 382                                 break;
 383                         case ';':
 384                                 token = lex_token(file);
 385                                 PARSE_TREE_ADD(PARSE_TYPE_DONE);
 386                                 break;
 387                         case '-':
 388                                 token = lex_token(file);
 389                                 PARSE_TREE_ADD(PARSE_TYPE_MINUS);
 390                                 break;
 391                         case '+':
 392                                 token = lex_token(file);
 393                                 PARSE_TREE_ADD(PARSE_TYPE_ADD);
 394                                 break;
 395                         case '{':
 396                                 token = lex_token(file);
 397                                 PARSE_TREE_ADD(PARSE_TYPE_LBS);
 398                                 break;
 399                         case '}':
 400                                 token = lex_token(file);
 401                                 PARSE_TREE_ADD(PARSE_TYPE_RBS);
 402                                 break;
 403
 404                         /*
 405                          * TODO: Fix lexer to spit out ( ) as tokens, it seems the
 406                          * using '(' or ')' in parser doesn't work properly unless
 407                          * there are spaces before them to allow the lexer to properly
 408                          * seperate identifiers. -- otherwise it eats all of it.
 409                          */
 410                         case LEX_IDENT:
 411                                 token = lex_token(file);
 412                                 PARSE_TREE_ADD(PARSE_TYPE_IDENT);
 413                                 break;
 414                 }
 415         }
 416         parse_debug(parseroot);
 417         lex_reset(file);
 418         parse_clear(parseroot);
 419         return 1;
 420 }