#include <limits.h>
#include <stdlib.h>
#include <string.h>
+#include <ctype.h>
#include "gmqcc.h"
+/* compile-time constant for constants */
+typedef struct {
+ char *name;
+ int type;
+ float value[3];
+ char *string; /* string value if constant is string literal */
+} constant;
+VECTOR_MAKE(constant, compile_constants);
+
/*
* These are not lexical tokens: These are parse tree types. Most people
* perform tokenizing on language punctuation which is wrong. That stuff
* work actually begins.
*/
#define PARSE_TREE_ADD(X) \
- do { \
- parsetree->next = mem_a(sizeof(struct parsenode)); \
- parsetree->next->next = NULL; \
- parsetree->next->type = (X); \
- parsetree = parsetree->next; \
- } while (0)
-
-/*
- * These are all the punctuation handled in the parser, these don't
- * need tokens, they're already tokens.
- */
-#if 0
- "&&", "||", "<=", ">=", "==", "!=", ";", ",", "!", "*",
- "/" , "(" , ")" , "-" , "+" , "=" , "[" , "]", "{", "}", "...",
- "." , "<" , ">" , "&" , "|" ,
-#endif
-
-#define STORE(X) { \
- printf(X); \
- break; \
+ do { \
+ parsetree->next = mem_a(sizeof(struct parsenode)); \
+ parsetree->next->next = NULL; \
+ parsetree->next->type = (X); \
+ parsetree = parsetree->next; \
+ } while (0)
+#define STORE(X) { \
+ printf(X); \
+ break; \
}
void parse_debug(struct parsenode *tree) {
- while (tree && tree->next != NULL) {
- /* skip blanks */
- if (tree->type == 0) {
- tree = tree->next;
- continue;
- }
-
+ long fill = 0;
+ while (tree) {
switch (tree->type) {
case PARSE_TYPE_ADD: STORE("OPERATOR: ADD \n");
case PARSE_TYPE_BAND: STORE("OPERATOR: BITAND \n");
case PARSE_TYPE_WHILE: STORE("LOOP: WHILE\n");
case PARSE_TYPE_FOR: STORE("LOOP: FOR\n");
case PARSE_TYPE_DO: STORE("LOOP: DO\n");
-
- case PARSE_TYPE_IDENT: STORE("IDENT: ???\n");
}
tree = tree->next;
}
}
/*
- * This just skips the token and throws it in the parse tree for later
- * checking / optimization / codegen, it doesn't do anything with it
- * like syntax check for legal use -- like it should as it's a TODO item
- * which is not implemented
+ * Performs a parse operation: This is a macro to prevent bugs, if the
+ * calls to lex_token are'nt exactly enough to feed to the end of the
+ * actual lexees for the current thing that is being parsed, the state
+ * of the next iteration in the creation of the parse tree will be wrong
+ * and everything will fail.
*/
-#define PARSE_TODO(X) { \
- token = lex_token(file); \
- PARSE_TREE_ADD(X); \
- break; \
+#define PARSE_PERFORM(X,C) { \
+ token = lex_token(file); \
+ { C } \
+ while (token != '\n') { \
+ token = lex_token(file); \
+ } \
+ PARSE_TREE_ADD(X); \
+ break; \
+}
+
+void parse_clear(struct parsenode *tree) {
+ if (!tree) return;
+ struct parsenode *temp = NULL;
+ while (tree != NULL) {
+ temp = tree;
+ tree = tree->next;
+ mem_d (temp);
+ }
+
+ /* free any potential typedefs */
+ typedef_clear();
}
-int parse(struct lex_file *file) {
+/*
+ * Generates a parse tree out of the lexees generated by the lexer. This
+ * is where the tree is built. This is where valid check is performed.
+ */
+int parse_tree(struct lex_file *file) {
struct parsenode *parsetree = NULL;
struct parsenode *parseroot = NULL;
parseroot = mem_a(sizeof(struct parsenode));
if (!parseroot)
return error(ERROR_INTERNAL, "Ran out of memory", " ");
- parsetree = parseroot;
+ parsetree = parseroot;
+ parsetree->type = -1; /* not a valid type -- root element */
}
int token = 0;
token != ERROR_PARSE && \
token != ERROR_PREPRO && file->length >= 0) {
switch (token) {
- case TOKEN_IF:
- token = lex_token(file);
- //while ((token == ' ' || token == '\n') && file->length >= 0)
- // token = lex_token(file);
-
- //if (token != '(')
- // error(ERROR_PARSE, "Expected `(` after if\n", "");
-
- PARSE_TREE_ADD(PARSE_TYPE_IF);
- break;
- case TOKEN_ELSE:
- token = lex_token(file);
- //while ((token == ' ' || token == '\n') && file->length >= 0)
- // token = lex_token(file);
-
- PARSE_TREE_ADD(PARSE_TYPE_ELSE);
- break;
- case TOKEN_FOR:
- token = lex_token(file);
- //while ((token == ' ' || token == '\n') && file->length >= 0)
- // token = lex_token(file);
-
- PARSE_TREE_ADD(PARSE_TYPE_FOR);
- break;
-
- /*
- * This is a quick and easy way to do typedefs at parse time
- * all power is in typedef_add(), in typedef.c. We handle
- * the tokens accordingly here.
- */
case TOKEN_TYPEDEF: {
- char *f = NULL;
- char *t = NULL;
+ char *f; /* from */
+ char *t; /* to */
+
token = lex_token(file);
- token = lex_token(file); f = strdup(file->lastok);
+ token = lex_token(file); f = util_strdup(file->lastok);
token = lex_token(file);
- token = lex_token(file); t = strdup(file->lastok);
+ token = lex_token(file); t = util_strdup(file->lastok);
typedef_add(f, t);
-
- /* free stdup strings */
mem_d(f);
mem_d(t);
+
+ token = lex_token(file);
+ if (token == ' ')
+ token = lex_token(file);
+
+ if (token != ';')
+ error(ERROR_PARSE, "%s:%d Expected a `;` at end of typedef statement\n", file->name, file->line);
+
+ token = lex_token(file);
break;
}
+
+ case TOKEN_VOID: PARSE_TREE_ADD(PARSE_TYPE_VOID); goto fall;
+ case TOKEN_STRING: PARSE_TREE_ADD(PARSE_TYPE_STRING); goto fall;
+ case TOKEN_VECTOR: PARSE_TREE_ADD(PARSE_TYPE_VECTOR); goto fall;
+ case TOKEN_ENTITY: PARSE_TREE_ADD(PARSE_TYPE_ENTITY); goto fall;
+ case TOKEN_FLOAT: PARSE_TREE_ADD(PARSE_TYPE_FLOAT); goto fall;
+ {
+ fall:;
+ char *name = NULL;
+ int type = token; /* story copy */
+
+ /* skip over space */
+ token = lex_token(file);
+ if (token == ' ')
+ token = lex_token(file);
+ /* save name */
+ name = util_strdup(file->lastok);
- case TOKEN_DO: PARSE_TODO(PARSE_TYPE_DO);
- case TOKEN_WHILE: PARSE_TODO(PARSE_TYPE_WHILE);
- case TOKEN_BREAK: PARSE_TODO(PARSE_TYPE_BREAK);
- case TOKEN_CONTINUE: PARSE_TODO(PARSE_TYPE_CONTINUE);
- case TOKEN_RETURN: PARSE_TODO(PARSE_TYPE_RETURN);
- case TOKEN_GOTO: PARSE_TODO(PARSE_TYPE_GOTO);
- case TOKEN_VOID: PARSE_TODO(PARSE_TYPE_VOID);
- case TOKEN_STRING: PARSE_TODO(PARSE_TYPE_STRING);
- case TOKEN_FLOAT: PARSE_TODO(PARSE_TYPE_FLOAT);
- case TOKEN_VECTOR: PARSE_TODO(PARSE_TYPE_VECTOR);
- case TOKEN_ENTITY: PARSE_TODO(PARSE_TYPE_ENTITY);
+ /* skip spaces */
+ token = lex_token(file);
+ if (token == ' ')
+ token = lex_token(file);
+
+ if (token == ';') {
+ /*
+ * Definitions go to the defs table, they don't have
+ * any sort of data with them yet.
+ */
+ } else if (token == '=') {
+ token = lex_token(file);
+ if (token == ' ')
+ token = lex_token(file);
+
+ /* strings are in file->lastok */
+ switch (type) {
+ case TOKEN_VOID:
+ return error(ERROR_PARSE, "%s:%d Cannot assign value to type void\n", file->name, file->line);
+
+ /* TODO: Validate (end quote), strip quotes for constant add, name constant */
+ case TOKEN_STRING:
+ if (*file->lastok != '"')
+ error(ERROR_PARSE, "%s:%d Expected a '\"' (quote) for string constant\n", file->name, file->line);
+ /* add the compile-time constant */
+ compile_constants_add((constant){
+ .name = util_strdup(name),
+ .type = TYPE_STRING,
+ .value = {0,0,0},
+ .string = util_strdup(file->lastok)
+ });
+ break;
+ /* TODO: name constant, old qc vec literals, whitespace fixes, name constant */
+ case TOKEN_VECTOR: {
+ float compile_calc_x = 0;
+ float compile_calc_y = 0;
+ float compile_calc_z = 0;
+ int compile_calc_d = 0; /* dot? */
+ int compile_calc_s = 0; /* sign (-, +) */
+
+ char compile_data[1024];
+ char *compile_eval = compile_data;
+
+ if (token != '{')
+ error(ERROR_PARSE, "%s:%d Expected initializer list `{`,`}` for vector constant\n", file->name, file->line);
+
+ /*
+ * This parses a single vector element: x,y & z. This will handle all the
+ * complicated mechanics of a vector, and can be extended as well. This
+ * is a rather large macro, and is #undef'd after it's use below.
+ */
+ #define PARSE_VEC_ELEMENT(NAME, BIT) \
+ token = lex_token(file); \
+ if (token == ' ') \
+ token = lex_token(file); \
+ if (token == '.') \
+ compile_calc_d = 1; \
+ if (!isdigit(token) && !compile_calc_d && token != '+' && token != '-') \
+ error(ERROR_PARSE,"%s:%d Invalid constant initializer element %c for vector, must be numeric\n", file->name, file->line, NAME); \
+ if (token == '+') \
+ compile_calc_s = '+'; \
+ if (token == '-' && !compile_calc_s) \
+ compile_calc_s = '-'; \
+ while (isdigit(token) || token == '.' || token == '+' || token == '-') { \
+ *compile_eval++ = token; \
+ token = lex_token(file); \
+ if (token == '.' && compile_calc_d) { \
+ error(ERROR_PARSE, "%s:%d Invalid constant initializer element %c for vector, must be numeric.\n", file->name, file->line, NAME); \
+ token = lex_token(file); \
+ } \
+ if ((token == '-' || token == '+') && compile_calc_s) { \
+ error(ERROR_PARSE, "%s:%d Invalid constant initializer sign for vector element %c\n", file->name, file->line, NAME); \
+ token = lex_token(file); \
+ } \
+ else if (token == '.' && !compile_calc_d) \
+ compile_calc_d = 1; \
+ else if (token == '-' && !compile_calc_s) \
+ compile_calc_s = '-'; \
+ else if (token == '+' && !compile_calc_s) \
+ compile_calc_s = '+'; \
+ } \
+ if (token == ' ') \
+ token = lex_token(file); \
+ if (NAME != 'z') { \
+ if (token != ',' && token != ' ') \
+ error(ERROR_PARSE, "%s:%d invalid constant initializer element %c for vector (missing spaces, or comma delimited list?)\n", file->name, file->line, NAME); \
+ } else if (token != '}') { \
+ error(ERROR_PARSE, "%s:%d Expected `}` on end of constant initialization for vector\n", file->name, file->line); \
+ } \
+ compile_calc_##BIT = atof(compile_data); \
+ compile_calc_d = 0; \
+ compile_calc_s = 0; \
+ compile_eval = &compile_data[0]; \
+ memset(compile_data, 0, sizeof(compile_data))
+
+ /*
+ * Parse all elements using the macro above.
+ * We must undef the macro afterwards.
+ */
+ PARSE_VEC_ELEMENT('x', x);
+ PARSE_VEC_ELEMENT('y', y);
+ PARSE_VEC_ELEMENT('z', z);
+ #undef PARSE_VEC_ELEMENT
+
+ /* Check for the semi-colon... */
+ token = lex_token(file);
+ if (token == ' ')
+ token = lex_token(file);
+ if (token != ';')
+ error(ERROR_PARSE, "%s:%d Expected `;` on end of constant initialization for vector\n", file->name, file->line);
+
+ /* add the compile-time constant */
+ compile_constants_add((constant){
+ .name = util_strdup(name),
+ .type = TYPE_VECTOR,
+ .value = {
+ [0] = compile_calc_x,
+ [1] = compile_calc_y,
+ [2] = compile_calc_z
+ },
+ .string = NULL
+ });
+ break;
+ }
+
+ case TOKEN_ENTITY:
+ case TOKEN_FLOAT: /*TODO: validate, constant generation, name constant */
+ if (!isdigit(token))
+ error(ERROR_PARSE, "%s:%d Expected numeric constant for float constant\n");
+ compile_constants_add((constant){
+ .name = util_strdup(name),
+ .type = TOKEN_FLOAT,
+ .value = {0,0,0},
+ .string = NULL
+ });
+ break;
+ }
+ } else if (token == '(') {
+ printf("FUNCTION ??\n");
+ }
+ mem_d(name);
+ }
/*
* From here down is all language punctuation: There is no
* which are higer than the ascii table.)
*/
case '#':
+ token = lex_token(file); /* skip '#' */
+ if (token == ' ')
+ token = lex_token(file);
/*
- * Skip the preprocessor for now: We'll implement our own
- * eventually. For now we need to make sure directives are
- * not accidently tokenized.
+ * If we make it here we found a directive, the supported
+ * directives so far are #include.
*/
- token = lex_token(file);
- token = lex_token(file);
-
+ if (strncmp(file->lastok, "include", sizeof("include")) == 0) {
+ /*
+ * We only suport include " ", not <> like in C (why?)
+ * because the latter is silly.
+ */
+ while (*file->lastok != '"' && token != '\n')
+ token = lex_token(file);
+ if (token == '\n')
+ return error(ERROR_PARSE, "%d: Invalid use of include preprocessor directive: wanted #include \"file.h\"\n", file->line-1);
+
+ char *copy = util_strdup(file->lastok);
+ struct lex_file *next = lex_include(file, copy);
+
+ if (!next) {
+ error(ERROR_INTERNAL, "Include subsystem failure\n");
+ exit (-1);
+ }
+ parse_tree(next);
+ mem_d (copy);
+ lex_close (next);
+ }
/* skip all tokens to end of directive */
while (token != '\n')
token = lex_token(file);
break;
- case '(':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LPARTH);
- break;
- case ')':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_RPARTH);
- break;
-
- case '&': /* & */
- token = lex_token(file);
- if (token == '&') { /* && */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LAND);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_BAND);
- break;
- case '|': /* | */
- token = lex_token(file);
- if (token == '|') { /* || */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LOR);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_BOR);
- break;
- case '!':
- token = lex_token(file);
- if (token == '=') { /* != */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LNEQ);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_LNOT);
- break;
- case '<': /* < */
- token = lex_token(file);
- if (token == '=') { /* <= */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LTEQ);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_LT);
- break;
- case '>': /* > */
- token = lex_token(file);
- if (token == '=') { /* >= */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_GTEQ);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_GT);
- break;
- case '=':
- token = lex_token(file);
- if (token == '=') { /* == */
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_EQEQ);
- break;
- }
- PARSE_TREE_ADD(PARSE_TYPE_EQUAL);
- break;
- case ';':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_DONE);
- break;
- case '-':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_MINUS);
- break;
- case '+':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_ADD);
- break;
- case '{':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_LBS);
- break;
- case '}':
- token = lex_token(file);
- PARSE_TREE_ADD(PARSE_TYPE_RBS);
- break;
-
- /*
- * TODO: Fix lexer to spit out ( ) as tokens, it seems the
- * using '(' or ')' in parser doesn't work properly unless
- * there are spaces before them to allow the lexer to properly
- * seperate identifiers. -- otherwise it eats all of it.
- */
case LEX_IDENT:
token = lex_token(file);
PARSE_TREE_ADD(PARSE_TYPE_IDENT);
}
parse_debug(parseroot);
lex_reset(file);
-
+ parse_clear(parseroot);
return 1;
}
projects / xonotic / gmqcc.git / blobdiff