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