1 | #include <errno.h> |
2 | #include <stdbool.h> |
3 | #include <stdio.h> |
4 | #include <string.h> |
5 | |
6 | #include "wren_common.h" |
7 | #include "wren_compiler.h" |
8 | #include "wren_vm.h" |
9 | |
10 | #if WREN_DEBUG_DUMP_COMPILED_CODE |
11 | #include "wren_debug.h" |
12 | #endif |
13 | |
14 | // This is written in bottom-up order, so the tokenization comes first, then |
15 | // parsing/code generation. This minimizes the number of explicit forward |
16 | // declarations needed. |
17 | |
18 | // The maximum number of local (i.e. not module level) variables that can be |
19 | // declared in a single function, method, or chunk of top level code. This is |
20 | // the maximum number of variables in scope at one time, and spans block scopes. |
21 | // |
22 | // Note that this limitation is also explicit in the bytecode. Since |
23 | // `CODE_LOAD_LOCAL` and `CODE_STORE_LOCAL` use a single argument byte to |
24 | // identify the local, only 256 can be in scope at one time. |
25 | #define MAX_LOCALS 256 |
26 | |
27 | // The maximum number of upvalues (i.e. variables from enclosing functions) |
28 | // that a function can close over. |
29 | #define MAX_UPVALUES 256 |
30 | |
31 | // The maximum number of distinct constants that a function can contain. This |
32 | // value is explicit in the bytecode since `CODE_CONSTANT` only takes a single |
33 | // two-byte argument. |
34 | #define MAX_CONSTANTS (1 << 16) |
35 | |
36 | // The maximum distance a CODE_JUMP or CODE_JUMP_IF instruction can move the |
37 | // instruction pointer. |
38 | #define MAX_JUMP (1 << 16) |
39 | |
40 | // The maximum depth that interpolation can nest. For example, this string has |
41 | // three levels: |
42 | // |
43 | // "outside %(one + "%(two + "%(three)")")" |
44 | #define MAX_INTERPOLATION_NESTING 8 |
45 | |
46 | // The buffer size used to format a compile error message, excluding the header |
47 | // with the module name and error location. Using a hardcoded buffer for this |
48 | // is kind of hairy, but fortunately we can control what the longest possible |
49 | // message is and handle that. Ideally, we'd use `snprintf()`, but that's not |
50 | // available in standard C++98. |
51 | #define ERROR_MESSAGE_SIZE (80 + MAX_VARIABLE_NAME + 15) |
52 | |
53 | typedef enum |
54 | { |
55 | TOKEN_LEFT_PAREN, |
56 | TOKEN_RIGHT_PAREN, |
57 | TOKEN_LEFT_BRACKET, |
58 | TOKEN_RIGHT_BRACKET, |
59 | TOKEN_LEFT_BRACE, |
60 | TOKEN_RIGHT_BRACE, |
61 | TOKEN_COLON, |
62 | TOKEN_DOT, |
63 | TOKEN_DOTDOT, |
64 | TOKEN_DOTDOTDOT, |
65 | TOKEN_COMMA, |
66 | TOKEN_STAR, |
67 | TOKEN_SLASH, |
68 | TOKEN_PERCENT, |
69 | TOKEN_HASH, |
70 | TOKEN_PLUS, |
71 | TOKEN_MINUS, |
72 | TOKEN_LTLT, |
73 | TOKEN_GTGT, |
74 | TOKEN_PIPE, |
75 | TOKEN_PIPEPIPE, |
76 | TOKEN_CARET, |
77 | TOKEN_AMP, |
78 | TOKEN_AMPAMP, |
79 | TOKEN_BANG, |
80 | TOKEN_TILDE, |
81 | TOKEN_QUESTION, |
82 | TOKEN_EQ, |
83 | TOKEN_LT, |
84 | TOKEN_GT, |
85 | TOKEN_LTEQ, |
86 | TOKEN_GTEQ, |
87 | TOKEN_EQEQ, |
88 | TOKEN_BANGEQ, |
89 | |
90 | TOKEN_BREAK, |
91 | TOKEN_CONTINUE, |
92 | TOKEN_CLASS, |
93 | TOKEN_CONSTRUCT, |
94 | TOKEN_ELSE, |
95 | TOKEN_FALSE, |
96 | TOKEN_FOR, |
97 | TOKEN_FOREIGN, |
98 | TOKEN_IF, |
99 | TOKEN_IMPORT, |
100 | TOKEN_AS, |
101 | TOKEN_IN, |
102 | TOKEN_IS, |
103 | TOKEN_NULL, |
104 | TOKEN_RETURN, |
105 | TOKEN_STATIC, |
106 | TOKEN_SUPER, |
107 | TOKEN_THIS, |
108 | TOKEN_TRUE, |
109 | TOKEN_VAR, |
110 | TOKEN_WHILE, |
111 | |
112 | TOKEN_FIELD, |
113 | TOKEN_STATIC_FIELD, |
114 | TOKEN_NAME, |
115 | TOKEN_NUMBER, |
116 | |
117 | // A string literal without any interpolation, or the last section of a |
118 | // string following the last interpolated expression. |
119 | TOKEN_STRING, |
120 | |
121 | // A portion of a string literal preceding an interpolated expression. This |
122 | // string: |
123 | // |
124 | // "a %(b) c %(d) e" |
125 | // |
126 | // is tokenized to: |
127 | // |
128 | // TOKEN_INTERPOLATION "a " |
129 | // TOKEN_NAME b |
130 | // TOKEN_INTERPOLATION " c " |
131 | // TOKEN_NAME d |
132 | // TOKEN_STRING " e" |
133 | TOKEN_INTERPOLATION, |
134 | |
135 | TOKEN_LINE, |
136 | |
137 | TOKEN_ERROR, |
138 | TOKEN_EOF |
139 | } TokenType; |
140 | |
141 | typedef struct |
142 | { |
143 | TokenType type; |
144 | |
145 | // The beginning of the token, pointing directly into the source. |
146 | const char* start; |
147 | |
148 | // The length of the token in characters. |
149 | int length; |
150 | |
151 | // The 1-based line where the token appears. |
152 | int line; |
153 | |
154 | // The parsed value if the token is a literal. |
155 | Value value; |
156 | } Token; |
157 | |
158 | typedef struct |
159 | { |
160 | WrenVM* vm; |
161 | |
162 | // The module being parsed. |
163 | ObjModule* module; |
164 | |
165 | // The source code being parsed. |
166 | const char* source; |
167 | |
168 | // The beginning of the currently-being-lexed token in [source]. |
169 | const char* tokenStart; |
170 | |
171 | // The current character being lexed in [source]. |
172 | const char* currentChar; |
173 | |
174 | // The 1-based line number of [currentChar]. |
175 | int currentLine; |
176 | |
177 | // The upcoming token. |
178 | Token next; |
179 | |
180 | // The most recently lexed token. |
181 | Token current; |
182 | |
183 | // The most recently consumed/advanced token. |
184 | Token previous; |
185 | |
186 | // Tracks the lexing state when tokenizing interpolated strings. |
187 | // |
188 | // Interpolated strings make the lexer not strictly regular: we don't know |
189 | // whether a ")" should be treated as a RIGHT_PAREN token or as ending an |
190 | // interpolated expression unless we know whether we are inside a string |
191 | // interpolation and how many unmatched "(" there are. This is particularly |
192 | // complex because interpolation can nest: |
193 | // |
194 | // " %( " %( inner ) " ) " |
195 | // |
196 | // This tracks that state. The parser maintains a stack of ints, one for each |
197 | // level of current interpolation nesting. Each value is the number of |
198 | // unmatched "(" that are waiting to be closed. |
199 | int parens[MAX_INTERPOLATION_NESTING]; |
200 | int numParens; |
201 | |
202 | // Whether compile errors should be printed to stderr or discarded. |
203 | bool printErrors; |
204 | |
205 | // If a syntax or compile error has occurred. |
206 | bool hasError; |
207 | } Parser; |
208 | |
209 | typedef struct |
210 | { |
211 | // The name of the local variable. This points directly into the original |
212 | // source code string. |
213 | const char* name; |
214 | |
215 | // The length of the local variable's name. |
216 | int length; |
217 | |
218 | // The depth in the scope chain that this variable was declared at. Zero is |
219 | // the outermost scope--parameters for a method, or the first local block in |
220 | // top level code. One is the scope within that, etc. |
221 | int depth; |
222 | |
223 | // If this local variable is being used as an upvalue. |
224 | bool isUpvalue; |
225 | } Local; |
226 | |
227 | typedef struct |
228 | { |
229 | // True if this upvalue is capturing a local variable from the enclosing |
230 | // function. False if it's capturing an upvalue. |
231 | bool isLocal; |
232 | |
233 | // The index of the local or upvalue being captured in the enclosing function. |
234 | int index; |
235 | } CompilerUpvalue; |
236 | |
237 | // Bookkeeping information for the current loop being compiled. |
238 | typedef struct sLoop |
239 | { |
240 | // Index of the instruction that the loop should jump back to. |
241 | int start; |
242 | |
243 | // Index of the argument for the CODE_JUMP_IF instruction used to exit the |
244 | // loop. Stored so we can patch it once we know where the loop ends. |
245 | int exitJump; |
246 | |
247 | // Index of the first instruction of the body of the loop. |
248 | int body; |
249 | |
250 | // Depth of the scope(s) that need to be exited if a break is hit inside the |
251 | // loop. |
252 | int scopeDepth; |
253 | |
254 | // The loop enclosing this one, or NULL if this is the outermost loop. |
255 | struct sLoop* enclosing; |
256 | } Loop; |
257 | |
258 | // The different signature syntaxes for different kinds of methods. |
259 | typedef enum |
260 | { |
261 | // A name followed by a (possibly empty) parenthesized parameter list. Also |
262 | // used for binary operators. |
263 | SIG_METHOD, |
264 | |
265 | // Just a name. Also used for unary operators. |
266 | SIG_GETTER, |
267 | |
268 | // A name followed by "=". |
269 | SIG_SETTER, |
270 | |
271 | // A square bracketed parameter list. |
272 | SIG_SUBSCRIPT, |
273 | |
274 | // A square bracketed parameter list followed by "=". |
275 | SIG_SUBSCRIPT_SETTER, |
276 | |
277 | // A constructor initializer function. This has a distinct signature to |
278 | // prevent it from being invoked directly outside of the constructor on the |
279 | // metaclass. |
280 | SIG_INITIALIZER |
281 | } SignatureType; |
282 | |
283 | typedef struct |
284 | { |
285 | const char* name; |
286 | int length; |
287 | SignatureType type; |
288 | int arity; |
289 | } Signature; |
290 | |
291 | // Bookkeeping information for compiling a class definition. |
292 | typedef struct |
293 | { |
294 | // The name of the class. |
295 | ObjString* name; |
296 | |
297 | // Attributes for the class itself |
298 | ObjMap* classAttributes; |
299 | // Attributes for methods in this class |
300 | ObjMap* methodAttributes; |
301 | |
302 | // Symbol table for the fields of the class. |
303 | SymbolTable fields; |
304 | |
305 | // Symbols for the methods defined by the class. Used to detect duplicate |
306 | // method definitions. |
307 | IntBuffer methods; |
308 | IntBuffer staticMethods; |
309 | |
310 | // True if the class being compiled is a foreign class. |
311 | bool isForeign; |
312 | |
313 | // True if the current method being compiled is static. |
314 | bool inStatic; |
315 | |
316 | // The signature of the method being compiled. |
317 | Signature* signature; |
318 | } ClassInfo; |
319 | |
320 | struct sCompiler |
321 | { |
322 | Parser* parser; |
323 | |
324 | // The compiler for the function enclosing this one, or NULL if it's the |
325 | // top level. |
326 | struct sCompiler* parent; |
327 | |
328 | // The currently in scope local variables. |
329 | Local locals[MAX_LOCALS]; |
330 | |
331 | // The number of local variables currently in scope. |
332 | int numLocals; |
333 | |
334 | // The upvalues that this function has captured from outer scopes. The count |
335 | // of them is stored in [numUpvalues]. |
336 | CompilerUpvalue upvalues[MAX_UPVALUES]; |
337 | |
338 | // The current level of block scope nesting, where zero is no nesting. A -1 |
339 | // here means top-level code is being compiled and there is no block scope |
340 | // in effect at all. Any variables declared will be module-level. |
341 | int scopeDepth; |
342 | |
343 | // The current number of slots (locals and temporaries) in use. |
344 | // |
345 | // We use this and maxSlots to track the maximum number of additional slots |
346 | // a function may need while executing. When the function is called, the |
347 | // fiber will check to ensure its stack has enough room to cover that worst |
348 | // case and grow the stack if needed. |
349 | // |
350 | // This value here doesn't include parameters to the function. Since those |
351 | // are already pushed onto the stack by the caller and tracked there, we |
352 | // don't need to double count them here. |
353 | int numSlots; |
354 | |
355 | // The current innermost loop being compiled, or NULL if not in a loop. |
356 | Loop* loop; |
357 | |
358 | // If this is a compiler for a method, keeps track of the class enclosing it. |
359 | ClassInfo* enclosingClass; |
360 | |
361 | // The function being compiled. |
362 | ObjFn* fn; |
363 | |
364 | // The constants for the function being compiled. |
365 | ObjMap* constants; |
366 | |
367 | // Whether or not the compiler is for a constructor initializer |
368 | bool isInitializer; |
369 | |
370 | // The number of attributes seen while parsing. |
371 | // We track this separately as compile time attributes |
372 | // are not stored, so we can't rely on attributes->count |
373 | // to enforce an error message when attributes are used |
374 | // anywhere other than methods or classes. |
375 | int numAttributes; |
376 | // Attributes for the next class or method. |
377 | ObjMap* attributes; |
378 | }; |
379 | |
380 | // Describes where a variable is declared. |
381 | typedef enum |
382 | { |
383 | // A local variable in the current function. |
384 | SCOPE_LOCAL, |
385 | |
386 | // A local variable declared in an enclosing function. |
387 | SCOPE_UPVALUE, |
388 | |
389 | // A top-level module variable. |
390 | SCOPE_MODULE |
391 | } Scope; |
392 | |
393 | // A reference to a variable and the scope where it is defined. This contains |
394 | // enough information to emit correct code to load or store the variable. |
395 | typedef struct |
396 | { |
397 | // The stack slot, upvalue slot, or module symbol defining the variable. |
398 | int index; |
399 | |
400 | // Where the variable is declared. |
401 | Scope scope; |
402 | } Variable; |
403 | |
404 | // Forward declarations |
405 | static void disallowAttributes(Compiler* compiler); |
406 | static void addToAttributeGroup(Compiler* compiler, Value group, Value key, Value value); |
407 | static void emitClassAttributes(Compiler* compiler, ClassInfo* classInfo); |
408 | static void copyAttributes(Compiler* compiler, ObjMap* into); |
409 | static void copyMethodAttributes(Compiler* compiler, bool isForeign, |
410 | bool isStatic, const char* fullSignature, int32_t length); |
411 | |
412 | // The stack effect of each opcode. The index in the array is the opcode, and |
413 | // the value is the stack effect of that instruction. |
414 | static const int stackEffects[] = { |
415 | #define OPCODE(_, effect) effect, |
416 | #include "wren_opcodes.h" |
417 | #undef OPCODE |
418 | }; |
419 | |
420 | static void printError(Parser* parser, int line, const char* label, |
421 | const char* format, va_list args) |
422 | { |
423 | parser->hasError = true; |
424 | if (!parser->printErrors) return; |
425 | |
426 | // Only report errors if there is a WrenErrorFn to handle them. |
427 | if (parser->vm->config.errorFn == NULL) return; |
428 | |
429 | // Format the label and message. |
430 | char message[ERROR_MESSAGE_SIZE]; |
431 | int length = sprintf(message, "%s: " , label); |
432 | length += vsprintf(message + length, format, args); |
433 | ASSERT(length < ERROR_MESSAGE_SIZE, "Error should not exceed buffer." ); |
434 | |
435 | ObjString* module = parser->module->name; |
436 | const char* module_name = module ? module->value : "<unknown>" ; |
437 | |
438 | parser->vm->config.errorFn(parser->vm, WREN_ERROR_COMPILE, |
439 | module_name, line, message); |
440 | } |
441 | |
442 | // Outputs a lexical error. |
443 | static void lexError(Parser* parser, const char* format, ...) |
444 | { |
445 | va_list args; |
446 | va_start(args, format); |
447 | printError(parser, parser->currentLine, "Error" , format, args); |
448 | va_end(args); |
449 | } |
450 | |
451 | // Outputs a compile or syntax error. This also marks the compilation as having |
452 | // an error, which ensures that the resulting code will be discarded and never |
453 | // run. This means that after calling error(), it's fine to generate whatever |
454 | // invalid bytecode you want since it won't be used. |
455 | // |
456 | // You'll note that most places that call error() continue to parse and compile |
457 | // after that. That's so that we can try to find as many compilation errors in |
458 | // one pass as possible instead of just bailing at the first one. |
459 | static void error(Compiler* compiler, const char* format, ...) |
460 | { |
461 | Token* token = &compiler->parser->previous; |
462 | |
463 | // If the parse error was caused by an error token, the lexer has already |
464 | // reported it. |
465 | if (token->type == TOKEN_ERROR) return; |
466 | |
467 | va_list args; |
468 | va_start(args, format); |
469 | if (token->type == TOKEN_LINE) |
470 | { |
471 | printError(compiler->parser, token->line, "Error at newline" , format, args); |
472 | } |
473 | else if (token->type == TOKEN_EOF) |
474 | { |
475 | printError(compiler->parser, token->line, |
476 | "Error at end of file" , format, args); |
477 | } |
478 | else |
479 | { |
480 | // Make sure we don't exceed the buffer with a very long token. |
481 | char label[10 + MAX_VARIABLE_NAME + 4 + 1]; |
482 | if (token->length <= MAX_VARIABLE_NAME) |
483 | { |
484 | sprintf(label, "Error at '%.*s'" , token->length, token->start); |
485 | } |
486 | else |
487 | { |
488 | sprintf(label, "Error at '%.*s...'" , MAX_VARIABLE_NAME, token->start); |
489 | } |
490 | printError(compiler->parser, token->line, label, format, args); |
491 | } |
492 | va_end(args); |
493 | } |
494 | |
495 | // Adds [constant] to the constant pool and returns its index. |
496 | static int addConstant(Compiler* compiler, Value constant) |
497 | { |
498 | if (compiler->parser->hasError) return -1; |
499 | |
500 | // See if we already have a constant for the value. If so, reuse it. |
501 | if (compiler->constants != NULL) |
502 | { |
503 | Value existing = wrenMapGet(compiler->constants, constant); |
504 | if (IS_NUM(existing)) return (int)AS_NUM(existing); |
505 | } |
506 | |
507 | // It's a new constant. |
508 | if (compiler->fn->constants.count < MAX_CONSTANTS) |
509 | { |
510 | if (IS_OBJ(constant)) wrenPushRoot(compiler->parser->vm, AS_OBJ(constant)); |
511 | wrenValueBufferWrite(compiler->parser->vm, &compiler->fn->constants, |
512 | constant); |
513 | if (IS_OBJ(constant)) wrenPopRoot(compiler->parser->vm); |
514 | |
515 | if (compiler->constants == NULL) |
516 | { |
517 | compiler->constants = wrenNewMap(compiler->parser->vm); |
518 | } |
519 | wrenMapSet(compiler->parser->vm, compiler->constants, constant, |
520 | NUM_VAL(compiler->fn->constants.count - 1)); |
521 | } |
522 | else |
523 | { |
524 | error(compiler, "A function may only contain %d unique constants." , |
525 | MAX_CONSTANTS); |
526 | } |
527 | |
528 | return compiler->fn->constants.count - 1; |
529 | } |
530 | |
531 | // Initializes [compiler]. |
532 | static void initCompiler(Compiler* compiler, Parser* parser, Compiler* parent, |
533 | bool isMethod) |
534 | { |
535 | compiler->parser = parser; |
536 | compiler->parent = parent; |
537 | compiler->loop = NULL; |
538 | compiler->enclosingClass = NULL; |
539 | compiler->isInitializer = false; |
540 | |
541 | // Initialize these to NULL before allocating in case a GC gets triggered in |
542 | // the middle of initializing the compiler. |
543 | compiler->fn = NULL; |
544 | compiler->constants = NULL; |
545 | |
546 | parser->vm->compiler = compiler; |
547 | |
548 | // Declare a local slot for either the closure or method receiver so that we |
549 | // don't try to reuse that slot for a user-defined local variable. For |
550 | // methods, we name it "this", so that we can resolve references to that like |
551 | // a normal variable. For functions, they have no explicit "this", so we use |
552 | // an empty name. That way references to "this" inside a function walks up |
553 | // the parent chain to find a method enclosing the function whose "this" we |
554 | // can close over. |
555 | compiler->numLocals = 1; |
556 | compiler->numSlots = compiler->numLocals; |
557 | |
558 | if (isMethod) |
559 | { |
560 | compiler->locals[0].name = "this" ; |
561 | compiler->locals[0].length = 4; |
562 | } |
563 | else |
564 | { |
565 | compiler->locals[0].name = NULL; |
566 | compiler->locals[0].length = 0; |
567 | } |
568 | |
569 | compiler->locals[0].depth = -1; |
570 | compiler->locals[0].isUpvalue = false; |
571 | |
572 | if (parent == NULL) |
573 | { |
574 | // Compiling top-level code, so the initial scope is module-level. |
575 | compiler->scopeDepth = -1; |
576 | } |
577 | else |
578 | { |
579 | // The initial scope for functions and methods is local scope. |
580 | compiler->scopeDepth = 0; |
581 | } |
582 | |
583 | compiler->numAttributes = 0; |
584 | compiler->attributes = wrenNewMap(parser->vm); |
585 | compiler->fn = wrenNewFunction(parser->vm, parser->module, |
586 | compiler->numLocals); |
587 | } |
588 | |
589 | // Lexing ---------------------------------------------------------------------- |
590 | |
591 | typedef struct |
592 | { |
593 | const char* identifier; |
594 | size_t length; |
595 | TokenType tokenType; |
596 | } Keyword; |
597 | |
598 | // The table of reserved words and their associated token types. |
599 | static Keyword keywords[] = |
600 | { |
601 | {"break" , 5, TOKEN_BREAK}, |
602 | {"continue" , 8, TOKEN_CONTINUE}, |
603 | {"class" , 5, TOKEN_CLASS}, |
604 | {"construct" , 9, TOKEN_CONSTRUCT}, |
605 | {"else" , 4, TOKEN_ELSE}, |
606 | {"false" , 5, TOKEN_FALSE}, |
607 | {"for" , 3, TOKEN_FOR}, |
608 | {"foreign" , 7, TOKEN_FOREIGN}, |
609 | {"if" , 2, TOKEN_IF}, |
610 | {"import" , 6, TOKEN_IMPORT}, |
611 | {"as" , 2, TOKEN_AS}, |
612 | {"in" , 2, TOKEN_IN}, |
613 | {"is" , 2, TOKEN_IS}, |
614 | {"null" , 4, TOKEN_NULL}, |
615 | {"return" , 6, TOKEN_RETURN}, |
616 | {"static" , 6, TOKEN_STATIC}, |
617 | {"super" , 5, TOKEN_SUPER}, |
618 | {"this" , 4, TOKEN_THIS}, |
619 | {"true" , 4, TOKEN_TRUE}, |
620 | {"var" , 3, TOKEN_VAR}, |
621 | {"while" , 5, TOKEN_WHILE}, |
622 | {NULL, 0, TOKEN_EOF} // Sentinel to mark the end of the array. |
623 | }; |
624 | |
625 | // Returns true if [c] is a valid (non-initial) identifier character. |
626 | static bool isName(char c) |
627 | { |
628 | return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_'; |
629 | } |
630 | |
631 | // Returns true if [c] is a digit. |
632 | static bool isDigit(char c) |
633 | { |
634 | return c >= '0' && c <= '9'; |
635 | } |
636 | |
637 | // Returns the current character the parser is sitting on. |
638 | static char peekChar(Parser* parser) |
639 | { |
640 | return *parser->currentChar; |
641 | } |
642 | |
643 | // Returns the character after the current character. |
644 | static char peekNextChar(Parser* parser) |
645 | { |
646 | // If we're at the end of the source, don't read past it. |
647 | if (peekChar(parser) == '\0') return '\0'; |
648 | return *(parser->currentChar + 1); |
649 | } |
650 | |
651 | // Advances the parser forward one character. |
652 | static char nextChar(Parser* parser) |
653 | { |
654 | char c = peekChar(parser); |
655 | parser->currentChar++; |
656 | if (c == '\n') parser->currentLine++; |
657 | return c; |
658 | } |
659 | |
660 | // If the current character is [c], consumes it and returns `true`. |
661 | static bool matchChar(Parser* parser, char c) |
662 | { |
663 | if (peekChar(parser) != c) return false; |
664 | nextChar(parser); |
665 | return true; |
666 | } |
667 | |
668 | // Sets the parser's current token to the given [type] and current character |
669 | // range. |
670 | static void makeToken(Parser* parser, TokenType type) |
671 | { |
672 | parser->next.type = type; |
673 | parser->next.start = parser->tokenStart; |
674 | parser->next.length = (int)(parser->currentChar - parser->tokenStart); |
675 | parser->next.line = parser->currentLine; |
676 | |
677 | // Make line tokens appear on the line containing the "\n". |
678 | if (type == TOKEN_LINE) parser->next.line--; |
679 | } |
680 | |
681 | // If the current character is [c], then consumes it and makes a token of type |
682 | // [two]. Otherwise makes a token of type [one]. |
683 | static void twoCharToken(Parser* parser, char c, TokenType two, TokenType one) |
684 | { |
685 | makeToken(parser, matchChar(parser, c) ? two : one); |
686 | } |
687 | |
688 | // Skips the rest of the current line. |
689 | static void (Parser* parser) |
690 | { |
691 | while (peekChar(parser) != '\n' && peekChar(parser) != '\0') |
692 | { |
693 | nextChar(parser); |
694 | } |
695 | } |
696 | |
697 | // Skips the rest of a block comment. |
698 | static void (Parser* parser) |
699 | { |
700 | int nesting = 1; |
701 | while (nesting > 0) |
702 | { |
703 | if (peekChar(parser) == '\0') |
704 | { |
705 | lexError(parser, "Unterminated block comment." ); |
706 | return; |
707 | } |
708 | |
709 | if (peekChar(parser) == '/' && peekNextChar(parser) == '*') |
710 | { |
711 | nextChar(parser); |
712 | nextChar(parser); |
713 | nesting++; |
714 | continue; |
715 | } |
716 | |
717 | if (peekChar(parser) == '*' && peekNextChar(parser) == '/') |
718 | { |
719 | nextChar(parser); |
720 | nextChar(parser); |
721 | nesting--; |
722 | continue; |
723 | } |
724 | |
725 | // Regular comment character. |
726 | nextChar(parser); |
727 | } |
728 | } |
729 | |
730 | // Reads the next character, which should be a hex digit (0-9, a-f, or A-F) and |
731 | // returns its numeric value. If the character isn't a hex digit, returns -1. |
732 | static int readHexDigit(Parser* parser) |
733 | { |
734 | char c = nextChar(parser); |
735 | if (c >= '0' && c <= '9') return c - '0'; |
736 | if (c >= 'a' && c <= 'f') return c - 'a' + 10; |
737 | if (c >= 'A' && c <= 'F') return c - 'A' + 10; |
738 | |
739 | // Don't consume it if it isn't expected. Keeps us from reading past the end |
740 | // of an unterminated string. |
741 | parser->currentChar--; |
742 | return -1; |
743 | } |
744 | |
745 | // Parses the numeric value of the current token. |
746 | static void makeNumber(Parser* parser, bool isHex) |
747 | { |
748 | errno = 0; |
749 | |
750 | if (isHex) |
751 | { |
752 | parser->next.value = NUM_VAL((double)strtoll(parser->tokenStart, NULL, 16)); |
753 | } |
754 | else |
755 | { |
756 | parser->next.value = NUM_VAL(strtod(parser->tokenStart, NULL)); |
757 | } |
758 | |
759 | if (errno == ERANGE) |
760 | { |
761 | lexError(parser, "Number literal was too large (%d)." , sizeof(long int)); |
762 | parser->next.value = NUM_VAL(0); |
763 | } |
764 | |
765 | // We don't check that the entire token is consumed after calling strtoll() |
766 | // or strtod() because we've already scanned it ourselves and know it's valid. |
767 | |
768 | makeToken(parser, TOKEN_NUMBER); |
769 | } |
770 | |
771 | // Finishes lexing a hexadecimal number literal. |
772 | static void readHexNumber(Parser* parser) |
773 | { |
774 | // Skip past the `x` used to denote a hexadecimal literal. |
775 | nextChar(parser); |
776 | |
777 | // Iterate over all the valid hexadecimal digits found. |
778 | while (readHexDigit(parser) != -1) continue; |
779 | |
780 | makeNumber(parser, true); |
781 | } |
782 | |
783 | // Finishes lexing a number literal. |
784 | static void readNumber(Parser* parser) |
785 | { |
786 | while (isDigit(peekChar(parser))) nextChar(parser); |
787 | |
788 | // See if it has a floating point. Make sure there is a digit after the "." |
789 | // so we don't get confused by method calls on number literals. |
790 | if (peekChar(parser) == '.' && isDigit(peekNextChar(parser))) |
791 | { |
792 | nextChar(parser); |
793 | while (isDigit(peekChar(parser))) nextChar(parser); |
794 | } |
795 | |
796 | // See if the number is in scientific notation. |
797 | if (matchChar(parser, 'e') || matchChar(parser, 'E')) |
798 | { |
799 | // Allow a single positive/negative exponent symbol. |
800 | if(!matchChar(parser, '+')) |
801 | { |
802 | matchChar(parser, '-'); |
803 | } |
804 | |
805 | if (!isDigit(peekChar(parser))) |
806 | { |
807 | lexError(parser, "Unterminated scientific notation." ); |
808 | } |
809 | |
810 | while (isDigit(peekChar(parser))) nextChar(parser); |
811 | } |
812 | |
813 | makeNumber(parser, false); |
814 | } |
815 | |
816 | // Finishes lexing an identifier. Handles reserved words. |
817 | static void readName(Parser* parser, TokenType type, char firstChar) |
818 | { |
819 | ByteBuffer string; |
820 | wrenByteBufferInit(&string); |
821 | wrenByteBufferWrite(parser->vm, &string, firstChar); |
822 | |
823 | while (isName(peekChar(parser)) || isDigit(peekChar(parser))) |
824 | { |
825 | char c = nextChar(parser); |
826 | wrenByteBufferWrite(parser->vm, &string, c); |
827 | } |
828 | |
829 | // Update the type if it's a keyword. |
830 | size_t length = parser->currentChar - parser->tokenStart; |
831 | for (int i = 0; keywords[i].identifier != NULL; i++) |
832 | { |
833 | if (length == keywords[i].length && |
834 | memcmp(parser->tokenStart, keywords[i].identifier, length) == 0) |
835 | { |
836 | type = keywords[i].tokenType; |
837 | break; |
838 | } |
839 | } |
840 | |
841 | parser->next.value = wrenNewStringLength(parser->vm, |
842 | (char*)string.data, string.count); |
843 | |
844 | wrenByteBufferClear(parser->vm, &string); |
845 | makeToken(parser, type); |
846 | } |
847 | |
848 | // Reads [digits] hex digits in a string literal and returns their number value. |
849 | static int readHexEscape(Parser* parser, int digits, const char* description) |
850 | { |
851 | int value = 0; |
852 | for (int i = 0; i < digits; i++) |
853 | { |
854 | if (peekChar(parser) == '"' || peekChar(parser) == '\0') |
855 | { |
856 | lexError(parser, "Incomplete %s escape sequence." , description); |
857 | |
858 | // Don't consume it if it isn't expected. Keeps us from reading past the |
859 | // end of an unterminated string. |
860 | parser->currentChar--; |
861 | break; |
862 | } |
863 | |
864 | int digit = readHexDigit(parser); |
865 | if (digit == -1) |
866 | { |
867 | lexError(parser, "Invalid %s escape sequence." , description); |
868 | break; |
869 | } |
870 | |
871 | value = (value * 16) | digit; |
872 | } |
873 | |
874 | return value; |
875 | } |
876 | |
877 | // Reads a hex digit Unicode escape sequence in a string literal. |
878 | static void readUnicodeEscape(Parser* parser, ByteBuffer* string, int length) |
879 | { |
880 | int value = readHexEscape(parser, length, "Unicode" ); |
881 | |
882 | // Grow the buffer enough for the encoded result. |
883 | int numBytes = wrenUtf8EncodeNumBytes(value); |
884 | if (numBytes != 0) |
885 | { |
886 | wrenByteBufferFill(parser->vm, string, 0, numBytes); |
887 | wrenUtf8Encode(value, string->data + string->count - numBytes); |
888 | } |
889 | } |
890 | |
891 | static void readRawString(Parser* parser) |
892 | { |
893 | ByteBuffer string; |
894 | wrenByteBufferInit(&string); |
895 | TokenType type = TOKEN_STRING; |
896 | |
897 | //consume the second and third " |
898 | nextChar(parser); |
899 | nextChar(parser); |
900 | |
901 | int skipStart = 0; |
902 | int firstNewline = -1; |
903 | |
904 | int skipEnd = -1; |
905 | int lastNewline = -1; |
906 | |
907 | for (;;) |
908 | { |
909 | char c = nextChar(parser); |
910 | char c1 = peekChar(parser); |
911 | char c2 = peekNextChar(parser); |
912 | |
913 | if(c == '\n') { |
914 | lastNewline = string.count; |
915 | skipEnd = lastNewline; |
916 | firstNewline = firstNewline == -1 ? string.count : firstNewline; |
917 | } |
918 | |
919 | if(c == '"' && c1 == '"' && c2 == '"') break; |
920 | |
921 | bool isWhitespace = c == ' ' || c == '\t'; |
922 | skipEnd = c == '\n' || isWhitespace ? skipEnd : -1; |
923 | |
924 | // If we haven't seen a newline or other character yet, |
925 | // and still seeing whitespace, count the characters |
926 | // as skippable till we know otherwise |
927 | bool skippable = skipStart != -1 && isWhitespace && firstNewline == -1; |
928 | skipStart = skippable ? string.count + 1 : skipStart; |
929 | |
930 | // We've counted leading whitespace till we hit something else, |
931 | // but it's not a newline, so we reset skipStart since we need these characters |
932 | if (firstNewline == -1 && !isWhitespace && c != '\n') skipStart = -1; |
933 | |
934 | if (c == '\0' || c1 == '\0' || c2 == '\0') |
935 | { |
936 | lexError(parser, "Unterminated raw string." ); |
937 | |
938 | // Don't consume it if it isn't expected. Keeps us from reading past the |
939 | // end of an unterminated string. |
940 | parser->currentChar--; |
941 | break; |
942 | } |
943 | |
944 | wrenByteBufferWrite(parser->vm, &string, c); |
945 | } |
946 | |
947 | //consume the second and third " |
948 | nextChar(parser); |
949 | nextChar(parser); |
950 | |
951 | int offset = 0; |
952 | int count = string.count; |
953 | |
954 | if(firstNewline != -1 && skipStart == firstNewline) offset = firstNewline + 1; |
955 | if(lastNewline != -1 && skipEnd == lastNewline) count = lastNewline; |
956 | |
957 | count -= (offset > count) ? count : offset; |
958 | |
959 | parser->next.value = wrenNewStringLength(parser->vm, |
960 | ((char*)string.data) + offset, count); |
961 | |
962 | wrenByteBufferClear(parser->vm, &string); |
963 | makeToken(parser, type); |
964 | } |
965 | |
966 | // Finishes lexing a string literal. |
967 | static void readString(Parser* parser) |
968 | { |
969 | ByteBuffer string; |
970 | TokenType type = TOKEN_STRING; |
971 | wrenByteBufferInit(&string); |
972 | |
973 | for (;;) |
974 | { |
975 | char c = nextChar(parser); |
976 | if (c == '"') break; |
977 | |
978 | if (c == '\0') |
979 | { |
980 | lexError(parser, "Unterminated string." ); |
981 | |
982 | // Don't consume it if it isn't expected. Keeps us from reading past the |
983 | // end of an unterminated string. |
984 | parser->currentChar--; |
985 | break; |
986 | } |
987 | |
988 | if (c == '%') |
989 | { |
990 | if (parser->numParens < MAX_INTERPOLATION_NESTING) |
991 | { |
992 | // TODO: Allow format string. |
993 | if (nextChar(parser) != '(') lexError(parser, "Expect '(' after '%%'." ); |
994 | |
995 | parser->parens[parser->numParens++] = 1; |
996 | type = TOKEN_INTERPOLATION; |
997 | break; |
998 | } |
999 | |
1000 | lexError(parser, "Interpolation may only nest %d levels deep." , |
1001 | MAX_INTERPOLATION_NESTING); |
1002 | } |
1003 | |
1004 | if (c == '\\') |
1005 | { |
1006 | switch (nextChar(parser)) |
1007 | { |
1008 | case '"': wrenByteBufferWrite(parser->vm, &string, '"'); break; |
1009 | case '\\': wrenByteBufferWrite(parser->vm, &string, '\\'); break; |
1010 | case '%': wrenByteBufferWrite(parser->vm, &string, '%'); break; |
1011 | case '0': wrenByteBufferWrite(parser->vm, &string, '\0'); break; |
1012 | case 'a': wrenByteBufferWrite(parser->vm, &string, '\a'); break; |
1013 | case 'b': wrenByteBufferWrite(parser->vm, &string, '\b'); break; |
1014 | case 'e': wrenByteBufferWrite(parser->vm, &string, '\33'); break; |
1015 | case 'f': wrenByteBufferWrite(parser->vm, &string, '\f'); break; |
1016 | case 'n': wrenByteBufferWrite(parser->vm, &string, '\n'); break; |
1017 | case 'r': wrenByteBufferWrite(parser->vm, &string, '\r'); break; |
1018 | case 't': wrenByteBufferWrite(parser->vm, &string, '\t'); break; |
1019 | case 'u': readUnicodeEscape(parser, &string, 4); break; |
1020 | case 'U': readUnicodeEscape(parser, &string, 8); break; |
1021 | case 'v': wrenByteBufferWrite(parser->vm, &string, '\v'); break; |
1022 | case 'x': |
1023 | wrenByteBufferWrite(parser->vm, &string, |
1024 | (uint8_t)readHexEscape(parser, 2, "byte" )); |
1025 | break; |
1026 | |
1027 | default: |
1028 | lexError(parser, "Invalid escape character '%c'." , |
1029 | *(parser->currentChar - 1)); |
1030 | break; |
1031 | } |
1032 | } |
1033 | else |
1034 | { |
1035 | wrenByteBufferWrite(parser->vm, &string, c); |
1036 | } |
1037 | } |
1038 | |
1039 | parser->next.value = wrenNewStringLength(parser->vm, |
1040 | (char*)string.data, string.count); |
1041 | |
1042 | wrenByteBufferClear(parser->vm, &string); |
1043 | makeToken(parser, type); |
1044 | } |
1045 | |
1046 | // Lex the next token and store it in [parser.next]. |
1047 | static void nextToken(Parser* parser) |
1048 | { |
1049 | parser->previous = parser->current; |
1050 | parser->current = parser->next; |
1051 | |
1052 | // If we are out of tokens, don't try to tokenize any more. We *do* still |
1053 | // copy the TOKEN_EOF to previous so that code that expects it to be consumed |
1054 | // will still work. |
1055 | if (parser->next.type == TOKEN_EOF) return; |
1056 | if (parser->current.type == TOKEN_EOF) return; |
1057 | |
1058 | while (peekChar(parser) != '\0') |
1059 | { |
1060 | parser->tokenStart = parser->currentChar; |
1061 | |
1062 | char c = nextChar(parser); |
1063 | switch (c) |
1064 | { |
1065 | case '(': |
1066 | // If we are inside an interpolated expression, count the unmatched "(". |
1067 | if (parser->numParens > 0) parser->parens[parser->numParens - 1]++; |
1068 | makeToken(parser, TOKEN_LEFT_PAREN); |
1069 | return; |
1070 | |
1071 | case ')': |
1072 | // If we are inside an interpolated expression, count the ")". |
1073 | if (parser->numParens > 0 && |
1074 | --parser->parens[parser->numParens - 1] == 0) |
1075 | { |
1076 | // This is the final ")", so the interpolation expression has ended. |
1077 | // This ")" now begins the next section of the template string. |
1078 | parser->numParens--; |
1079 | readString(parser); |
1080 | return; |
1081 | } |
1082 | |
1083 | makeToken(parser, TOKEN_RIGHT_PAREN); |
1084 | return; |
1085 | |
1086 | case '[': makeToken(parser, TOKEN_LEFT_BRACKET); return; |
1087 | case ']': makeToken(parser, TOKEN_RIGHT_BRACKET); return; |
1088 | case '{': makeToken(parser, TOKEN_LEFT_BRACE); return; |
1089 | case '}': makeToken(parser, TOKEN_RIGHT_BRACE); return; |
1090 | case ':': makeToken(parser, TOKEN_COLON); return; |
1091 | case ',': makeToken(parser, TOKEN_COMMA); return; |
1092 | case '*': makeToken(parser, TOKEN_STAR); return; |
1093 | case '%': makeToken(parser, TOKEN_PERCENT); return; |
1094 | case '#': { |
1095 | // Ignore shebang on the first line. |
1096 | if (parser->currentLine == 1 && peekChar(parser) == '!' && peekNextChar(parser) == '/') |
1097 | { |
1098 | skipLineComment(parser); |
1099 | break; |
1100 | } |
1101 | // Otherwise we treat it as a token a token |
1102 | makeToken(parser, TOKEN_HASH); |
1103 | return; |
1104 | } |
1105 | case '^': makeToken(parser, TOKEN_CARET); return; |
1106 | case '+': makeToken(parser, TOKEN_PLUS); return; |
1107 | case '-': makeToken(parser, TOKEN_MINUS); return; |
1108 | case '~': makeToken(parser, TOKEN_TILDE); return; |
1109 | case '?': makeToken(parser, TOKEN_QUESTION); return; |
1110 | |
1111 | case '|': twoCharToken(parser, '|', TOKEN_PIPEPIPE, TOKEN_PIPE); return; |
1112 | case '&': twoCharToken(parser, '&', TOKEN_AMPAMP, TOKEN_AMP); return; |
1113 | case '=': twoCharToken(parser, '=', TOKEN_EQEQ, TOKEN_EQ); return; |
1114 | case '!': twoCharToken(parser, '=', TOKEN_BANGEQ, TOKEN_BANG); return; |
1115 | |
1116 | case '.': |
1117 | if (matchChar(parser, '.')) |
1118 | { |
1119 | twoCharToken(parser, '.', TOKEN_DOTDOTDOT, TOKEN_DOTDOT); |
1120 | return; |
1121 | } |
1122 | |
1123 | makeToken(parser, TOKEN_DOT); |
1124 | return; |
1125 | |
1126 | case '/': |
1127 | if (matchChar(parser, '/')) |
1128 | { |
1129 | skipLineComment(parser); |
1130 | break; |
1131 | } |
1132 | |
1133 | if (matchChar(parser, '*')) |
1134 | { |
1135 | skipBlockComment(parser); |
1136 | break; |
1137 | } |
1138 | |
1139 | makeToken(parser, TOKEN_SLASH); |
1140 | return; |
1141 | |
1142 | case '<': |
1143 | if (matchChar(parser, '<')) |
1144 | { |
1145 | makeToken(parser, TOKEN_LTLT); |
1146 | } |
1147 | else |
1148 | { |
1149 | twoCharToken(parser, '=', TOKEN_LTEQ, TOKEN_LT); |
1150 | } |
1151 | return; |
1152 | |
1153 | case '>': |
1154 | if (matchChar(parser, '>')) |
1155 | { |
1156 | makeToken(parser, TOKEN_GTGT); |
1157 | } |
1158 | else |
1159 | { |
1160 | twoCharToken(parser, '=', TOKEN_GTEQ, TOKEN_GT); |
1161 | } |
1162 | return; |
1163 | |
1164 | case '\n': |
1165 | makeToken(parser, TOKEN_LINE); |
1166 | return; |
1167 | |
1168 | case ' ': |
1169 | case '\r': |
1170 | case '\t': |
1171 | // Skip forward until we run out of whitespace. |
1172 | while (peekChar(parser) == ' ' || |
1173 | peekChar(parser) == '\r' || |
1174 | peekChar(parser) == '\t') |
1175 | { |
1176 | nextChar(parser); |
1177 | } |
1178 | break; |
1179 | |
1180 | case '"': { |
1181 | if(peekChar(parser) == '"' && peekNextChar(parser) == '"') { |
1182 | readRawString(parser); |
1183 | return; |
1184 | } |
1185 | readString(parser); return; |
1186 | } |
1187 | case '_': |
1188 | readName(parser, |
1189 | peekChar(parser) == '_' ? TOKEN_STATIC_FIELD : TOKEN_FIELD, c); |
1190 | return; |
1191 | |
1192 | case '0': |
1193 | if (peekChar(parser) == 'x') |
1194 | { |
1195 | readHexNumber(parser); |
1196 | return; |
1197 | } |
1198 | |
1199 | readNumber(parser); |
1200 | return; |
1201 | |
1202 | default: |
1203 | if (isName(c)) |
1204 | { |
1205 | readName(parser, TOKEN_NAME, c); |
1206 | } |
1207 | else if (isDigit(c)) |
1208 | { |
1209 | readNumber(parser); |
1210 | } |
1211 | else |
1212 | { |
1213 | if (c >= 32 && c <= 126) |
1214 | { |
1215 | lexError(parser, "Invalid character '%c'." , c); |
1216 | } |
1217 | else |
1218 | { |
1219 | // Don't show non-ASCII values since we didn't UTF-8 decode the |
1220 | // bytes. Since there are no non-ASCII byte values that are |
1221 | // meaningful code units in Wren, the lexer works on raw bytes, |
1222 | // even though the source code and console output are UTF-8. |
1223 | lexError(parser, "Invalid byte 0x%x." , (uint8_t)c); |
1224 | } |
1225 | parser->next.type = TOKEN_ERROR; |
1226 | parser->next.length = 0; |
1227 | } |
1228 | return; |
1229 | } |
1230 | } |
1231 | |
1232 | // If we get here, we're out of source, so just make EOF tokens. |
1233 | parser->tokenStart = parser->currentChar; |
1234 | makeToken(parser, TOKEN_EOF); |
1235 | } |
1236 | |
1237 | // Parsing --------------------------------------------------------------------- |
1238 | |
1239 | // Returns the type of the current token. |
1240 | static TokenType peek(Compiler* compiler) |
1241 | { |
1242 | return compiler->parser->current.type; |
1243 | } |
1244 | |
1245 | // Returns the type of the current token. |
1246 | static TokenType peekNext(Compiler* compiler) |
1247 | { |
1248 | return compiler->parser->next.type; |
1249 | } |
1250 | |
1251 | // Consumes the current token if its type is [expected]. Returns true if a |
1252 | // token was consumed. |
1253 | static bool match(Compiler* compiler, TokenType expected) |
1254 | { |
1255 | if (peek(compiler) != expected) return false; |
1256 | |
1257 | nextToken(compiler->parser); |
1258 | return true; |
1259 | } |
1260 | |
1261 | // Consumes the current token. Emits an error if its type is not [expected]. |
1262 | static void consume(Compiler* compiler, TokenType expected, |
1263 | const char* errorMessage) |
1264 | { |
1265 | nextToken(compiler->parser); |
1266 | if (compiler->parser->previous.type != expected) |
1267 | { |
1268 | error(compiler, errorMessage); |
1269 | |
1270 | // If the next token is the one we want, assume the current one is just a |
1271 | // spurious error and discard it to minimize the number of cascaded errors. |
1272 | if (compiler->parser->current.type == expected) nextToken(compiler->parser); |
1273 | } |
1274 | } |
1275 | |
1276 | // Matches one or more newlines. Returns true if at least one was found. |
1277 | static bool matchLine(Compiler* compiler) |
1278 | { |
1279 | if (!match(compiler, TOKEN_LINE)) return false; |
1280 | |
1281 | while (match(compiler, TOKEN_LINE)); |
1282 | return true; |
1283 | } |
1284 | |
1285 | // Discards any newlines starting at the current token. |
1286 | static void ignoreNewlines(Compiler* compiler) |
1287 | { |
1288 | matchLine(compiler); |
1289 | } |
1290 | |
1291 | // Consumes the current token. Emits an error if it is not a newline. Then |
1292 | // discards any duplicate newlines following it. |
1293 | static void consumeLine(Compiler* compiler, const char* errorMessage) |
1294 | { |
1295 | consume(compiler, TOKEN_LINE, errorMessage); |
1296 | ignoreNewlines(compiler); |
1297 | } |
1298 | |
1299 | static void allowLineBeforeDot(Compiler* compiler) { |
1300 | if (peek(compiler) == TOKEN_LINE && peekNext(compiler) == TOKEN_DOT) { |
1301 | nextToken(compiler->parser); |
1302 | } |
1303 | } |
1304 | |
1305 | // Variables and scopes -------------------------------------------------------- |
1306 | |
1307 | // Emits one single-byte argument. Returns its index. |
1308 | static int emitByte(Compiler* compiler, int byte) |
1309 | { |
1310 | wrenByteBufferWrite(compiler->parser->vm, &compiler->fn->code, (uint8_t)byte); |
1311 | |
1312 | // Assume the instruction is associated with the most recently consumed token. |
1313 | wrenIntBufferWrite(compiler->parser->vm, &compiler->fn->debug->sourceLines, |
1314 | compiler->parser->previous.line); |
1315 | |
1316 | return compiler->fn->code.count - 1; |
1317 | } |
1318 | |
1319 | // Emits one bytecode instruction. |
1320 | static void emitOp(Compiler* compiler, Code instruction) |
1321 | { |
1322 | emitByte(compiler, instruction); |
1323 | |
1324 | // Keep track of the stack's high water mark. |
1325 | compiler->numSlots += stackEffects[instruction]; |
1326 | if (compiler->numSlots > compiler->fn->maxSlots) |
1327 | { |
1328 | compiler->fn->maxSlots = compiler->numSlots; |
1329 | } |
1330 | } |
1331 | |
1332 | // Emits one 16-bit argument, which will be written big endian. |
1333 | static void emitShort(Compiler* compiler, int arg) |
1334 | { |
1335 | emitByte(compiler, (arg >> 8) & 0xff); |
1336 | emitByte(compiler, arg & 0xff); |
1337 | } |
1338 | |
1339 | // Emits one bytecode instruction followed by a 8-bit argument. Returns the |
1340 | // index of the argument in the bytecode. |
1341 | static int emitByteArg(Compiler* compiler, Code instruction, int arg) |
1342 | { |
1343 | emitOp(compiler, instruction); |
1344 | return emitByte(compiler, arg); |
1345 | } |
1346 | |
1347 | // Emits one bytecode instruction followed by a 16-bit argument, which will be |
1348 | // written big endian. |
1349 | static void emitShortArg(Compiler* compiler, Code instruction, int arg) |
1350 | { |
1351 | emitOp(compiler, instruction); |
1352 | emitShort(compiler, arg); |
1353 | } |
1354 | |
1355 | // Emits [instruction] followed by a placeholder for a jump offset. The |
1356 | // placeholder can be patched by calling [jumpPatch]. Returns the index of the |
1357 | // placeholder. |
1358 | static int emitJump(Compiler* compiler, Code instruction) |
1359 | { |
1360 | emitOp(compiler, instruction); |
1361 | emitByte(compiler, 0xff); |
1362 | return emitByte(compiler, 0xff) - 1; |
1363 | } |
1364 | |
1365 | // Creates a new constant for the current value and emits the bytecode to load |
1366 | // it from the constant table. |
1367 | static void emitConstant(Compiler* compiler, Value value) |
1368 | { |
1369 | int constant = addConstant(compiler, value); |
1370 | |
1371 | // Compile the code to load the constant. |
1372 | emitShortArg(compiler, CODE_CONSTANT, constant); |
1373 | } |
1374 | |
1375 | // Create a new local variable with [name]. Assumes the current scope is local |
1376 | // and the name is unique. |
1377 | static int addLocal(Compiler* compiler, const char* name, int length) |
1378 | { |
1379 | Local* local = &compiler->locals[compiler->numLocals]; |
1380 | local->name = name; |
1381 | local->length = length; |
1382 | local->depth = compiler->scopeDepth; |
1383 | local->isUpvalue = false; |
1384 | return compiler->numLocals++; |
1385 | } |
1386 | |
1387 | // Declares a variable in the current scope whose name is the given token. |
1388 | // |
1389 | // If [token] is `NULL`, uses the previously consumed token. Returns its symbol. |
1390 | static int declareVariable(Compiler* compiler, Token* token) |
1391 | { |
1392 | if (token == NULL) token = &compiler->parser->previous; |
1393 | |
1394 | if (token->length > MAX_VARIABLE_NAME) |
1395 | { |
1396 | error(compiler, "Variable name cannot be longer than %d characters." , |
1397 | MAX_VARIABLE_NAME); |
1398 | } |
1399 | |
1400 | // Top-level module scope. |
1401 | if (compiler->scopeDepth == -1) |
1402 | { |
1403 | int line = -1; |
1404 | int symbol = wrenDefineVariable(compiler->parser->vm, |
1405 | compiler->parser->module, |
1406 | token->start, token->length, |
1407 | NULL_VAL, &line); |
1408 | |
1409 | if (symbol == -1) |
1410 | { |
1411 | error(compiler, "Module variable is already defined." ); |
1412 | } |
1413 | else if (symbol == -2) |
1414 | { |
1415 | error(compiler, "Too many module variables defined." ); |
1416 | } |
1417 | else if (symbol == -3) |
1418 | { |
1419 | error(compiler, |
1420 | "Variable '%.*s' referenced before this definition (first use at line %d)." , |
1421 | token->length, token->start, line); |
1422 | } |
1423 | |
1424 | return symbol; |
1425 | } |
1426 | |
1427 | // See if there is already a variable with this name declared in the current |
1428 | // scope. (Outer scopes are OK: those get shadowed.) |
1429 | for (int i = compiler->numLocals - 1; i >= 0; i--) |
1430 | { |
1431 | Local* local = &compiler->locals[i]; |
1432 | |
1433 | // Once we escape this scope and hit an outer one, we can stop. |
1434 | if (local->depth < compiler->scopeDepth) break; |
1435 | |
1436 | if (local->length == token->length && |
1437 | memcmp(local->name, token->start, token->length) == 0) |
1438 | { |
1439 | error(compiler, "Variable is already declared in this scope." ); |
1440 | return i; |
1441 | } |
1442 | } |
1443 | |
1444 | if (compiler->numLocals == MAX_LOCALS) |
1445 | { |
1446 | error(compiler, "Cannot declare more than %d variables in one scope." , |
1447 | MAX_LOCALS); |
1448 | return -1; |
1449 | } |
1450 | |
1451 | return addLocal(compiler, token->start, token->length); |
1452 | } |
1453 | |
1454 | // Parses a name token and declares a variable in the current scope with that |
1455 | // name. Returns its slot. |
1456 | static int declareNamedVariable(Compiler* compiler) |
1457 | { |
1458 | consume(compiler, TOKEN_NAME, "Expect variable name." ); |
1459 | return declareVariable(compiler, NULL); |
1460 | } |
1461 | |
1462 | // Stores a variable with the previously defined symbol in the current scope. |
1463 | static void defineVariable(Compiler* compiler, int symbol) |
1464 | { |
1465 | // Store the variable. If it's a local, the result of the initializer is |
1466 | // in the correct slot on the stack already so we're done. |
1467 | if (compiler->scopeDepth >= 0) return; |
1468 | |
1469 | // It's a module-level variable, so store the value in the module slot and |
1470 | // then discard the temporary for the initializer. |
1471 | emitShortArg(compiler, CODE_STORE_MODULE_VAR, symbol); |
1472 | emitOp(compiler, CODE_POP); |
1473 | } |
1474 | |
1475 | // Starts a new local block scope. |
1476 | static void pushScope(Compiler* compiler) |
1477 | { |
1478 | compiler->scopeDepth++; |
1479 | } |
1480 | |
1481 | // Generates code to discard local variables at [depth] or greater. Does *not* |
1482 | // actually undeclare variables or pop any scopes, though. This is called |
1483 | // directly when compiling "break" statements to ditch the local variables |
1484 | // before jumping out of the loop even though they are still in scope *past* |
1485 | // the break instruction. |
1486 | // |
1487 | // Returns the number of local variables that were eliminated. |
1488 | static int discardLocals(Compiler* compiler, int depth) |
1489 | { |
1490 | ASSERT(compiler->scopeDepth > -1, "Cannot exit top-level scope." ); |
1491 | |
1492 | int local = compiler->numLocals - 1; |
1493 | while (local >= 0 && compiler->locals[local].depth >= depth) |
1494 | { |
1495 | // If the local was closed over, make sure the upvalue gets closed when it |
1496 | // goes out of scope on the stack. We use emitByte() and not emitOp() here |
1497 | // because we don't want to track that stack effect of these pops since the |
1498 | // variables are still in scope after the break. |
1499 | if (compiler->locals[local].isUpvalue) |
1500 | { |
1501 | emitByte(compiler, CODE_CLOSE_UPVALUE); |
1502 | } |
1503 | else |
1504 | { |
1505 | emitByte(compiler, CODE_POP); |
1506 | } |
1507 | |
1508 | |
1509 | local--; |
1510 | } |
1511 | |
1512 | return compiler->numLocals - local - 1; |
1513 | } |
1514 | |
1515 | // Closes the last pushed block scope and discards any local variables declared |
1516 | // in that scope. This should only be called in a statement context where no |
1517 | // temporaries are still on the stack. |
1518 | static void popScope(Compiler* compiler) |
1519 | { |
1520 | int popped = discardLocals(compiler, compiler->scopeDepth); |
1521 | compiler->numLocals -= popped; |
1522 | compiler->numSlots -= popped; |
1523 | compiler->scopeDepth--; |
1524 | } |
1525 | |
1526 | // Attempts to look up the name in the local variables of [compiler]. If found, |
1527 | // returns its index, otherwise returns -1. |
1528 | static int resolveLocal(Compiler* compiler, const char* name, int length) |
1529 | { |
1530 | // Look it up in the local scopes. Look in reverse order so that the most |
1531 | // nested variable is found first and shadows outer ones. |
1532 | for (int i = compiler->numLocals - 1; i >= 0; i--) |
1533 | { |
1534 | if (compiler->locals[i].length == length && |
1535 | memcmp(name, compiler->locals[i].name, length) == 0) |
1536 | { |
1537 | return i; |
1538 | } |
1539 | } |
1540 | |
1541 | return -1; |
1542 | } |
1543 | |
1544 | // Adds an upvalue to [compiler]'s function with the given properties. Does not |
1545 | // add one if an upvalue for that variable is already in the list. Returns the |
1546 | // index of the upvalue. |
1547 | static int addUpvalue(Compiler* compiler, bool isLocal, int index) |
1548 | { |
1549 | // Look for an existing one. |
1550 | for (int i = 0; i < compiler->fn->numUpvalues; i++) |
1551 | { |
1552 | CompilerUpvalue* upvalue = &compiler->upvalues[i]; |
1553 | if (upvalue->index == index && upvalue->isLocal == isLocal) return i; |
1554 | } |
1555 | |
1556 | // If we got here, it's a new upvalue. |
1557 | compiler->upvalues[compiler->fn->numUpvalues].isLocal = isLocal; |
1558 | compiler->upvalues[compiler->fn->numUpvalues].index = index; |
1559 | return compiler->fn->numUpvalues++; |
1560 | } |
1561 | |
1562 | // Attempts to look up [name] in the functions enclosing the one being compiled |
1563 | // by [compiler]. If found, it adds an upvalue for it to this compiler's list |
1564 | // of upvalues (unless it's already in there) and returns its index. If not |
1565 | // found, returns -1. |
1566 | // |
1567 | // If the name is found outside of the immediately enclosing function, this |
1568 | // will flatten the closure and add upvalues to all of the intermediate |
1569 | // functions so that it gets walked down to this one. |
1570 | // |
1571 | // If it reaches a method boundary, this stops and returns -1 since methods do |
1572 | // not close over local variables. |
1573 | static int findUpvalue(Compiler* compiler, const char* name, int length) |
1574 | { |
1575 | // If we are at the top level, we didn't find it. |
1576 | if (compiler->parent == NULL) return -1; |
1577 | |
1578 | // If we hit the method boundary (and the name isn't a static field), then |
1579 | // stop looking for it. We'll instead treat it as a self send. |
1580 | if (name[0] != '_' && compiler->parent->enclosingClass != NULL) return -1; |
1581 | |
1582 | // See if it's a local variable in the immediately enclosing function. |
1583 | int local = resolveLocal(compiler->parent, name, length); |
1584 | if (local != -1) |
1585 | { |
1586 | // Mark the local as an upvalue so we know to close it when it goes out of |
1587 | // scope. |
1588 | compiler->parent->locals[local].isUpvalue = true; |
1589 | |
1590 | return addUpvalue(compiler, true, local); |
1591 | } |
1592 | |
1593 | // See if it's an upvalue in the immediately enclosing function. In other |
1594 | // words, if it's a local variable in a non-immediately enclosing function. |
1595 | // This "flattens" closures automatically: it adds upvalues to all of the |
1596 | // intermediate functions to get from the function where a local is declared |
1597 | // all the way into the possibly deeply nested function that is closing over |
1598 | // it. |
1599 | int upvalue = findUpvalue(compiler->parent, name, length); |
1600 | if (upvalue != -1) |
1601 | { |
1602 | return addUpvalue(compiler, false, upvalue); |
1603 | } |
1604 | |
1605 | // If we got here, we walked all the way up the parent chain and couldn't |
1606 | // find it. |
1607 | return -1; |
1608 | } |
1609 | |
1610 | // Look up [name] in the current scope to see what variable it refers to. |
1611 | // Returns the variable either in local scope, or the enclosing function's |
1612 | // upvalue list. Does not search the module scope. Returns a variable with |
1613 | // index -1 if not found. |
1614 | static Variable resolveNonmodule(Compiler* compiler, |
1615 | const char* name, int length) |
1616 | { |
1617 | // Look it up in the local scopes. |
1618 | Variable variable; |
1619 | variable.scope = SCOPE_LOCAL; |
1620 | variable.index = resolveLocal(compiler, name, length); |
1621 | if (variable.index != -1) return variable; |
1622 | |
1623 | // Tt's not a local, so guess that it's an upvalue. |
1624 | variable.scope = SCOPE_UPVALUE; |
1625 | variable.index = findUpvalue(compiler, name, length); |
1626 | return variable; |
1627 | } |
1628 | |
1629 | // Look up [name] in the current scope to see what variable it refers to. |
1630 | // Returns the variable either in module scope, local scope, or the enclosing |
1631 | // function's upvalue list. Returns a variable with index -1 if not found. |
1632 | static Variable resolveName(Compiler* compiler, const char* name, int length) |
1633 | { |
1634 | Variable variable = resolveNonmodule(compiler, name, length); |
1635 | if (variable.index != -1) return variable; |
1636 | |
1637 | variable.scope = SCOPE_MODULE; |
1638 | variable.index = wrenSymbolTableFind(&compiler->parser->module->variableNames, |
1639 | name, length); |
1640 | return variable; |
1641 | } |
1642 | |
1643 | static void loadLocal(Compiler* compiler, int slot) |
1644 | { |
1645 | if (slot <= 8) |
1646 | { |
1647 | emitOp(compiler, (Code)(CODE_LOAD_LOCAL_0 + slot)); |
1648 | return; |
1649 | } |
1650 | |
1651 | emitByteArg(compiler, CODE_LOAD_LOCAL, slot); |
1652 | } |
1653 | |
1654 | // Finishes [compiler], which is compiling a function, method, or chunk of top |
1655 | // level code. If there is a parent compiler, then this emits code in the |
1656 | // parent compiler to load the resulting function. |
1657 | static ObjFn* endCompiler(Compiler* compiler, |
1658 | const char* debugName, int debugNameLength) |
1659 | { |
1660 | // If we hit an error, don't finish the function since it's borked anyway. |
1661 | if (compiler->parser->hasError) |
1662 | { |
1663 | compiler->parser->vm->compiler = compiler->parent; |
1664 | return NULL; |
1665 | } |
1666 | |
1667 | // Mark the end of the bytecode. Since it may contain multiple early returns, |
1668 | // we can't rely on CODE_RETURN to tell us we're at the end. |
1669 | emitOp(compiler, CODE_END); |
1670 | |
1671 | wrenFunctionBindName(compiler->parser->vm, compiler->fn, |
1672 | debugName, debugNameLength); |
1673 | |
1674 | // In the function that contains this one, load the resulting function object. |
1675 | if (compiler->parent != NULL) |
1676 | { |
1677 | int constant = addConstant(compiler->parent, OBJ_VAL(compiler->fn)); |
1678 | |
1679 | // Wrap the function in a closure. We do this even if it has no upvalues so |
1680 | // that the VM can uniformly assume all called objects are closures. This |
1681 | // makes creating a function a little slower, but makes invoking them |
1682 | // faster. Given that functions are invoked more often than they are |
1683 | // created, this is a win. |
1684 | emitShortArg(compiler->parent, CODE_CLOSURE, constant); |
1685 | |
1686 | // Emit arguments for each upvalue to know whether to capture a local or |
1687 | // an upvalue. |
1688 | for (int i = 0; i < compiler->fn->numUpvalues; i++) |
1689 | { |
1690 | emitByte(compiler->parent, compiler->upvalues[i].isLocal ? 1 : 0); |
1691 | emitByte(compiler->parent, compiler->upvalues[i].index); |
1692 | } |
1693 | } |
1694 | |
1695 | // Pop this compiler off the stack. |
1696 | compiler->parser->vm->compiler = compiler->parent; |
1697 | |
1698 | #if WREN_DEBUG_DUMP_COMPILED_CODE |
1699 | wrenDumpCode(compiler->parser->vm, compiler->fn); |
1700 | #endif |
1701 | |
1702 | return compiler->fn; |
1703 | } |
1704 | |
1705 | // Grammar --------------------------------------------------------------------- |
1706 | |
1707 | typedef enum |
1708 | { |
1709 | PREC_NONE, |
1710 | PREC_LOWEST, |
1711 | PREC_ASSIGNMENT, // = |
1712 | PREC_CONDITIONAL, // ?: |
1713 | PREC_LOGICAL_OR, // || |
1714 | PREC_LOGICAL_AND, // && |
1715 | PREC_EQUALITY, // == != |
1716 | PREC_IS, // is |
1717 | PREC_COMPARISON, // < > <= >= |
1718 | PREC_BITWISE_OR, // | |
1719 | PREC_BITWISE_XOR, // ^ |
1720 | PREC_BITWISE_AND, // & |
1721 | PREC_BITWISE_SHIFT, // << >> |
1722 | PREC_RANGE, // .. ... |
1723 | PREC_TERM, // + - |
1724 | PREC_FACTOR, // * / % |
1725 | PREC_UNARY, // unary - ! ~ |
1726 | PREC_CALL, // . () [] |
1727 | PREC_PRIMARY |
1728 | } Precedence; |
1729 | |
1730 | typedef void (*GrammarFn)(Compiler*, bool canAssign); |
1731 | |
1732 | typedef void (*SignatureFn)(Compiler* compiler, Signature* signature); |
1733 | |
1734 | typedef struct |
1735 | { |
1736 | GrammarFn prefix; |
1737 | GrammarFn infix; |
1738 | SignatureFn method; |
1739 | Precedence precedence; |
1740 | const char* name; |
1741 | } GrammarRule; |
1742 | |
1743 | // Forward declarations since the grammar is recursive. |
1744 | static GrammarRule* getRule(TokenType type); |
1745 | static void expression(Compiler* compiler); |
1746 | static void statement(Compiler* compiler); |
1747 | static void definition(Compiler* compiler); |
1748 | static void parsePrecedence(Compiler* compiler, Precedence precedence); |
1749 | |
1750 | // Replaces the placeholder argument for a previous CODE_JUMP or CODE_JUMP_IF |
1751 | // instruction with an offset that jumps to the current end of bytecode. |
1752 | static void patchJump(Compiler* compiler, int offset) |
1753 | { |
1754 | // -2 to adjust for the bytecode for the jump offset itself. |
1755 | int jump = compiler->fn->code.count - offset - 2; |
1756 | if (jump > MAX_JUMP) error(compiler, "Too much code to jump over." ); |
1757 | |
1758 | compiler->fn->code.data[offset] = (jump >> 8) & 0xff; |
1759 | compiler->fn->code.data[offset + 1] = jump & 0xff; |
1760 | } |
1761 | |
1762 | // Parses a block body, after the initial "{" has been consumed. |
1763 | // |
1764 | // Returns true if it was a expression body, false if it was a statement body. |
1765 | // (More precisely, returns true if a value was left on the stack. An empty |
1766 | // block returns false.) |
1767 | static bool finishBlock(Compiler* compiler) |
1768 | { |
1769 | // Empty blocks do nothing. |
1770 | if (match(compiler, TOKEN_RIGHT_BRACE)) return false; |
1771 | |
1772 | // If there's no line after the "{", it's a single-expression body. |
1773 | if (!matchLine(compiler)) |
1774 | { |
1775 | expression(compiler); |
1776 | consume(compiler, TOKEN_RIGHT_BRACE, "Expect '}' at end of block." ); |
1777 | return true; |
1778 | } |
1779 | |
1780 | // Empty blocks (with just a newline inside) do nothing. |
1781 | if (match(compiler, TOKEN_RIGHT_BRACE)) return false; |
1782 | |
1783 | // Compile the definition list. |
1784 | do |
1785 | { |
1786 | definition(compiler); |
1787 | consumeLine(compiler, "Expect newline after statement." ); |
1788 | } |
1789 | while (peek(compiler) != TOKEN_RIGHT_BRACE && peek(compiler) != TOKEN_EOF); |
1790 | |
1791 | consume(compiler, TOKEN_RIGHT_BRACE, "Expect '}' at end of block." ); |
1792 | return false; |
1793 | } |
1794 | |
1795 | // Parses a method or function body, after the initial "{" has been consumed. |
1796 | // |
1797 | // If [Compiler->isInitializer] is `true`, this is the body of a constructor |
1798 | // initializer. In that case, this adds the code to ensure it returns `this`. |
1799 | static void finishBody(Compiler* compiler) |
1800 | { |
1801 | bool isExpressionBody = finishBlock(compiler); |
1802 | |
1803 | if (compiler->isInitializer) |
1804 | { |
1805 | // If the initializer body evaluates to a value, discard it. |
1806 | if (isExpressionBody) emitOp(compiler, CODE_POP); |
1807 | |
1808 | // The receiver is always stored in the first local slot. |
1809 | emitOp(compiler, CODE_LOAD_LOCAL_0); |
1810 | } |
1811 | else if (!isExpressionBody) |
1812 | { |
1813 | // Implicitly return null in statement bodies. |
1814 | emitOp(compiler, CODE_NULL); |
1815 | } |
1816 | |
1817 | emitOp(compiler, CODE_RETURN); |
1818 | } |
1819 | |
1820 | // The VM can only handle a certain number of parameters, so check that we |
1821 | // haven't exceeded that and give a usable error. |
1822 | static void validateNumParameters(Compiler* compiler, int numArgs) |
1823 | { |
1824 | if (numArgs == MAX_PARAMETERS + 1) |
1825 | { |
1826 | // Only show an error at exactly max + 1 so that we can keep parsing the |
1827 | // parameters and minimize cascaded errors. |
1828 | error(compiler, "Methods cannot have more than %d parameters." , |
1829 | MAX_PARAMETERS); |
1830 | } |
1831 | } |
1832 | |
1833 | // Parses the rest of a comma-separated parameter list after the opening |
1834 | // delimeter. Updates `arity` in [signature] with the number of parameters. |
1835 | static void finishParameterList(Compiler* compiler, Signature* signature) |
1836 | { |
1837 | do |
1838 | { |
1839 | ignoreNewlines(compiler); |
1840 | validateNumParameters(compiler, ++signature->arity); |
1841 | |
1842 | // Define a local variable in the method for the parameter. |
1843 | declareNamedVariable(compiler); |
1844 | } |
1845 | while (match(compiler, TOKEN_COMMA)); |
1846 | } |
1847 | |
1848 | // Gets the symbol for a method [name] with [length]. |
1849 | static int methodSymbol(Compiler* compiler, const char* name, int length) |
1850 | { |
1851 | return wrenSymbolTableEnsure(compiler->parser->vm, |
1852 | &compiler->parser->vm->methodNames, name, length); |
1853 | } |
1854 | |
1855 | // Appends characters to [name] (and updates [length]) for [numParams] "_" |
1856 | // surrounded by [leftBracket] and [rightBracket]. |
1857 | static void signatureParameterList(char name[MAX_METHOD_SIGNATURE], int* length, |
1858 | int numParams, char leftBracket, char rightBracket) |
1859 | { |
1860 | name[(*length)++] = leftBracket; |
1861 | |
1862 | // This function may be called with too many parameters. When that happens, |
1863 | // a compile error has already been reported, but we need to make sure we |
1864 | // don't overflow the string too, hence the MAX_PARAMETERS check. |
1865 | for (int i = 0; i < numParams && i < MAX_PARAMETERS; i++) |
1866 | { |
1867 | if (i > 0) name[(*length)++] = ','; |
1868 | name[(*length)++] = '_'; |
1869 | } |
1870 | name[(*length)++] = rightBracket; |
1871 | } |
1872 | |
1873 | // Fills [name] with the stringified version of [signature] and updates |
1874 | // [length] to the resulting length. |
1875 | static void signatureToString(Signature* signature, |
1876 | char name[MAX_METHOD_SIGNATURE], int* length) |
1877 | { |
1878 | *length = 0; |
1879 | |
1880 | // Build the full name from the signature. |
1881 | memcpy(name + *length, signature->name, signature->length); |
1882 | *length += signature->length; |
1883 | |
1884 | switch (signature->type) |
1885 | { |
1886 | case SIG_METHOD: |
1887 | signatureParameterList(name, length, signature->arity, '(', ')'); |
1888 | break; |
1889 | |
1890 | case SIG_GETTER: |
1891 | // The signature is just the name. |
1892 | break; |
1893 | |
1894 | case SIG_SETTER: |
1895 | name[(*length)++] = '='; |
1896 | signatureParameterList(name, length, 1, '(', ')'); |
1897 | break; |
1898 | |
1899 | case SIG_SUBSCRIPT: |
1900 | signatureParameterList(name, length, signature->arity, '[', ']'); |
1901 | break; |
1902 | |
1903 | case SIG_SUBSCRIPT_SETTER: |
1904 | signatureParameterList(name, length, signature->arity - 1, '[', ']'); |
1905 | name[(*length)++] = '='; |
1906 | signatureParameterList(name, length, 1, '(', ')'); |
1907 | break; |
1908 | |
1909 | case SIG_INITIALIZER: |
1910 | memcpy(name, "init " , 5); |
1911 | memcpy(name + 5, signature->name, signature->length); |
1912 | *length = 5 + signature->length; |
1913 | signatureParameterList(name, length, signature->arity, '(', ')'); |
1914 | break; |
1915 | } |
1916 | |
1917 | name[*length] = '\0'; |
1918 | } |
1919 | |
1920 | // Gets the symbol for a method with [signature]. |
1921 | static int signatureSymbol(Compiler* compiler, Signature* signature) |
1922 | { |
1923 | // Build the full name from the signature. |
1924 | char name[MAX_METHOD_SIGNATURE]; |
1925 | int length; |
1926 | signatureToString(signature, name, &length); |
1927 | |
1928 | return methodSymbol(compiler, name, length); |
1929 | } |
1930 | |
1931 | // Returns a signature with [type] whose name is from the last consumed token. |
1932 | static Signature signatureFromToken(Compiler* compiler, SignatureType type) |
1933 | { |
1934 | Signature signature; |
1935 | |
1936 | // Get the token for the method name. |
1937 | Token* token = &compiler->parser->previous; |
1938 | signature.name = token->start; |
1939 | signature.length = token->length; |
1940 | signature.type = type; |
1941 | signature.arity = 0; |
1942 | |
1943 | if (signature.length > MAX_METHOD_NAME) |
1944 | { |
1945 | error(compiler, "Method names cannot be longer than %d characters." , |
1946 | MAX_METHOD_NAME); |
1947 | signature.length = MAX_METHOD_NAME; |
1948 | } |
1949 | |
1950 | return signature; |
1951 | } |
1952 | |
1953 | // Parses a comma-separated list of arguments. Modifies [signature] to include |
1954 | // the arity of the argument list. |
1955 | static void finishArgumentList(Compiler* compiler, Signature* signature) |
1956 | { |
1957 | do |
1958 | { |
1959 | ignoreNewlines(compiler); |
1960 | validateNumParameters(compiler, ++signature->arity); |
1961 | expression(compiler); |
1962 | } |
1963 | while (match(compiler, TOKEN_COMMA)); |
1964 | |
1965 | // Allow a newline before the closing delimiter. |
1966 | ignoreNewlines(compiler); |
1967 | } |
1968 | |
1969 | // Compiles a method call with [signature] using [instruction]. |
1970 | static void callSignature(Compiler* compiler, Code instruction, |
1971 | Signature* signature) |
1972 | { |
1973 | int symbol = signatureSymbol(compiler, signature); |
1974 | emitShortArg(compiler, (Code)(instruction + signature->arity), symbol); |
1975 | |
1976 | if (instruction == CODE_SUPER_0) |
1977 | { |
1978 | // Super calls need to be statically bound to the class's superclass. This |
1979 | // ensures we call the right method even when a method containing a super |
1980 | // call is inherited by another subclass. |
1981 | // |
1982 | // We bind it at class definition time by storing a reference to the |
1983 | // superclass in a constant. So, here, we create a slot in the constant |
1984 | // table and store NULL in it. When the method is bound, we'll look up the |
1985 | // superclass then and store it in the constant slot. |
1986 | emitShort(compiler, addConstant(compiler, NULL_VAL)); |
1987 | } |
1988 | } |
1989 | |
1990 | // Compiles a method call with [numArgs] for a method with [name] with [length]. |
1991 | static void callMethod(Compiler* compiler, int numArgs, const char* name, |
1992 | int length) |
1993 | { |
1994 | int symbol = methodSymbol(compiler, name, length); |
1995 | emitShortArg(compiler, (Code)(CODE_CALL_0 + numArgs), symbol); |
1996 | } |
1997 | |
1998 | // Compiles an (optional) argument list for a method call with [methodSignature] |
1999 | // and then calls it. |
2000 | static void methodCall(Compiler* compiler, Code instruction, |
2001 | Signature* signature) |
2002 | { |
2003 | // Make a new signature that contains the updated arity and type based on |
2004 | // the arguments we find. |
2005 | Signature called = { signature->name, signature->length, SIG_GETTER, 0 }; |
2006 | |
2007 | // Parse the argument list, if any. |
2008 | if (match(compiler, TOKEN_LEFT_PAREN)) |
2009 | { |
2010 | called.type = SIG_METHOD; |
2011 | |
2012 | // Allow new line before an empty argument list |
2013 | ignoreNewlines(compiler); |
2014 | |
2015 | // Allow empty an argument list. |
2016 | if (peek(compiler) != TOKEN_RIGHT_PAREN) |
2017 | { |
2018 | finishArgumentList(compiler, &called); |
2019 | } |
2020 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after arguments." ); |
2021 | } |
2022 | |
2023 | // Parse the block argument, if any. |
2024 | if (match(compiler, TOKEN_LEFT_BRACE)) |
2025 | { |
2026 | // Include the block argument in the arity. |
2027 | called.type = SIG_METHOD; |
2028 | called.arity++; |
2029 | |
2030 | Compiler fnCompiler; |
2031 | initCompiler(&fnCompiler, compiler->parser, compiler, false); |
2032 | |
2033 | // Make a dummy signature to track the arity. |
2034 | Signature fnSignature = { "" , 0, SIG_METHOD, 0 }; |
2035 | |
2036 | // Parse the parameter list, if any. |
2037 | if (match(compiler, TOKEN_PIPE)) |
2038 | { |
2039 | finishParameterList(&fnCompiler, &fnSignature); |
2040 | consume(compiler, TOKEN_PIPE, "Expect '|' after function parameters." ); |
2041 | } |
2042 | |
2043 | fnCompiler.fn->arity = fnSignature.arity; |
2044 | |
2045 | finishBody(&fnCompiler); |
2046 | |
2047 | // Name the function based on the method its passed to. |
2048 | char blockName[MAX_METHOD_SIGNATURE + 15]; |
2049 | int blockLength; |
2050 | signatureToString(&called, blockName, &blockLength); |
2051 | memmove(blockName + blockLength, " block argument" , 16); |
2052 | |
2053 | endCompiler(&fnCompiler, blockName, blockLength + 15); |
2054 | } |
2055 | |
2056 | // TODO: Allow Grace-style mixfix methods? |
2057 | |
2058 | // If this is a super() call for an initializer, make sure we got an actual |
2059 | // argument list. |
2060 | if (signature->type == SIG_INITIALIZER) |
2061 | { |
2062 | if (called.type != SIG_METHOD) |
2063 | { |
2064 | error(compiler, "A superclass constructor must have an argument list." ); |
2065 | } |
2066 | |
2067 | called.type = SIG_INITIALIZER; |
2068 | } |
2069 | |
2070 | callSignature(compiler, instruction, &called); |
2071 | } |
2072 | |
2073 | // Compiles a call whose name is the previously consumed token. This includes |
2074 | // getters, method calls with arguments, and setter calls. |
2075 | static void namedCall(Compiler* compiler, bool canAssign, Code instruction) |
2076 | { |
2077 | // Get the token for the method name. |
2078 | Signature signature = signatureFromToken(compiler, SIG_GETTER); |
2079 | |
2080 | if (canAssign && match(compiler, TOKEN_EQ)) |
2081 | { |
2082 | ignoreNewlines(compiler); |
2083 | |
2084 | // Build the setter signature. |
2085 | signature.type = SIG_SETTER; |
2086 | signature.arity = 1; |
2087 | |
2088 | // Compile the assigned value. |
2089 | expression(compiler); |
2090 | callSignature(compiler, instruction, &signature); |
2091 | } |
2092 | else |
2093 | { |
2094 | methodCall(compiler, instruction, &signature); |
2095 | allowLineBeforeDot(compiler); |
2096 | } |
2097 | } |
2098 | |
2099 | // Emits the code to load [variable] onto the stack. |
2100 | static void loadVariable(Compiler* compiler, Variable variable) |
2101 | { |
2102 | switch (variable.scope) |
2103 | { |
2104 | case SCOPE_LOCAL: |
2105 | loadLocal(compiler, variable.index); |
2106 | break; |
2107 | case SCOPE_UPVALUE: |
2108 | emitByteArg(compiler, CODE_LOAD_UPVALUE, variable.index); |
2109 | break; |
2110 | case SCOPE_MODULE: |
2111 | emitShortArg(compiler, CODE_LOAD_MODULE_VAR, variable.index); |
2112 | break; |
2113 | default: |
2114 | UNREACHABLE(); |
2115 | } |
2116 | } |
2117 | |
2118 | // Loads the receiver of the currently enclosing method. Correctly handles |
2119 | // functions defined inside methods. |
2120 | static void loadThis(Compiler* compiler) |
2121 | { |
2122 | loadVariable(compiler, resolveNonmodule(compiler, "this" , 4)); |
2123 | } |
2124 | |
2125 | // Pushes the value for a module-level variable implicitly imported from core. |
2126 | static void loadCoreVariable(Compiler* compiler, const char* name) |
2127 | { |
2128 | int symbol = wrenSymbolTableFind(&compiler->parser->module->variableNames, |
2129 | name, strlen(name)); |
2130 | ASSERT(symbol != -1, "Should have already defined core name." ); |
2131 | emitShortArg(compiler, CODE_LOAD_MODULE_VAR, symbol); |
2132 | } |
2133 | |
2134 | // A parenthesized expression. |
2135 | static void grouping(Compiler* compiler, bool canAssign) |
2136 | { |
2137 | expression(compiler); |
2138 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after expression." ); |
2139 | } |
2140 | |
2141 | // A list literal. |
2142 | static void list(Compiler* compiler, bool canAssign) |
2143 | { |
2144 | // Instantiate a new list. |
2145 | loadCoreVariable(compiler, "List" ); |
2146 | callMethod(compiler, 0, "new()" , 5); |
2147 | |
2148 | // Compile the list elements. Each one compiles to a ".add()" call. |
2149 | do |
2150 | { |
2151 | ignoreNewlines(compiler); |
2152 | |
2153 | // Stop if we hit the end of the list. |
2154 | if (peek(compiler) == TOKEN_RIGHT_BRACKET) break; |
2155 | |
2156 | // The element. |
2157 | expression(compiler); |
2158 | callMethod(compiler, 1, "addCore_(_)" , 11); |
2159 | } while (match(compiler, TOKEN_COMMA)); |
2160 | |
2161 | // Allow newlines before the closing ']'. |
2162 | ignoreNewlines(compiler); |
2163 | consume(compiler, TOKEN_RIGHT_BRACKET, "Expect ']' after list elements." ); |
2164 | } |
2165 | |
2166 | // A map literal. |
2167 | static void map(Compiler* compiler, bool canAssign) |
2168 | { |
2169 | // Instantiate a new map. |
2170 | loadCoreVariable(compiler, "Map" ); |
2171 | callMethod(compiler, 0, "new()" , 5); |
2172 | |
2173 | // Compile the map elements. Each one is compiled to just invoke the |
2174 | // subscript setter on the map. |
2175 | do |
2176 | { |
2177 | ignoreNewlines(compiler); |
2178 | |
2179 | // Stop if we hit the end of the map. |
2180 | if (peek(compiler) == TOKEN_RIGHT_BRACE) break; |
2181 | |
2182 | // The key. |
2183 | parsePrecedence(compiler, PREC_UNARY); |
2184 | consume(compiler, TOKEN_COLON, "Expect ':' after map key." ); |
2185 | ignoreNewlines(compiler); |
2186 | |
2187 | // The value. |
2188 | expression(compiler); |
2189 | callMethod(compiler, 2, "addCore_(_,_)" , 13); |
2190 | } while (match(compiler, TOKEN_COMMA)); |
2191 | |
2192 | // Allow newlines before the closing '}'. |
2193 | ignoreNewlines(compiler); |
2194 | consume(compiler, TOKEN_RIGHT_BRACE, "Expect '}' after map entries." ); |
2195 | } |
2196 | |
2197 | // Unary operators like `-foo`. |
2198 | static void unaryOp(Compiler* compiler, bool canAssign) |
2199 | { |
2200 | GrammarRule* rule = getRule(compiler->parser->previous.type); |
2201 | |
2202 | ignoreNewlines(compiler); |
2203 | |
2204 | // Compile the argument. |
2205 | parsePrecedence(compiler, (Precedence)(PREC_UNARY + 1)); |
2206 | |
2207 | // Call the operator method on the left-hand side. |
2208 | callMethod(compiler, 0, rule->name, 1); |
2209 | } |
2210 | |
2211 | static void boolean(Compiler* compiler, bool canAssign) |
2212 | { |
2213 | emitOp(compiler, |
2214 | compiler->parser->previous.type == TOKEN_FALSE ? CODE_FALSE : CODE_TRUE); |
2215 | } |
2216 | |
2217 | // Walks the compiler chain to find the compiler for the nearest class |
2218 | // enclosing this one. Returns NULL if not currently inside a class definition. |
2219 | static Compiler* getEnclosingClassCompiler(Compiler* compiler) |
2220 | { |
2221 | while (compiler != NULL) |
2222 | { |
2223 | if (compiler->enclosingClass != NULL) return compiler; |
2224 | compiler = compiler->parent; |
2225 | } |
2226 | |
2227 | return NULL; |
2228 | } |
2229 | |
2230 | // Walks the compiler chain to find the nearest class enclosing this one. |
2231 | // Returns NULL if not currently inside a class definition. |
2232 | static ClassInfo* getEnclosingClass(Compiler* compiler) |
2233 | { |
2234 | compiler = getEnclosingClassCompiler(compiler); |
2235 | return compiler == NULL ? NULL : compiler->enclosingClass; |
2236 | } |
2237 | |
2238 | static void field(Compiler* compiler, bool canAssign) |
2239 | { |
2240 | // Initialize it with a fake value so we can keep parsing and minimize the |
2241 | // number of cascaded errors. |
2242 | int field = MAX_FIELDS; |
2243 | |
2244 | ClassInfo* enclosingClass = getEnclosingClass(compiler); |
2245 | |
2246 | if (enclosingClass == NULL) |
2247 | { |
2248 | error(compiler, "Cannot reference a field outside of a class definition." ); |
2249 | } |
2250 | else if (enclosingClass->isForeign) |
2251 | { |
2252 | error(compiler, "Cannot define fields in a foreign class." ); |
2253 | } |
2254 | else if (enclosingClass->inStatic) |
2255 | { |
2256 | error(compiler, "Cannot use an instance field in a static method." ); |
2257 | } |
2258 | else |
2259 | { |
2260 | // Look up the field, or implicitly define it. |
2261 | field = wrenSymbolTableEnsure(compiler->parser->vm, &enclosingClass->fields, |
2262 | compiler->parser->previous.start, |
2263 | compiler->parser->previous.length); |
2264 | |
2265 | if (field >= MAX_FIELDS) |
2266 | { |
2267 | error(compiler, "A class can only have %d fields." , MAX_FIELDS); |
2268 | } |
2269 | } |
2270 | |
2271 | // If there's an "=" after a field name, it's an assignment. |
2272 | bool isLoad = true; |
2273 | if (canAssign && match(compiler, TOKEN_EQ)) |
2274 | { |
2275 | // Compile the right-hand side. |
2276 | expression(compiler); |
2277 | isLoad = false; |
2278 | } |
2279 | |
2280 | // If we're directly inside a method, use a more optimal instruction. |
2281 | if (compiler->parent != NULL && |
2282 | compiler->parent->enclosingClass == enclosingClass) |
2283 | { |
2284 | emitByteArg(compiler, isLoad ? CODE_LOAD_FIELD_THIS : CODE_STORE_FIELD_THIS, |
2285 | field); |
2286 | } |
2287 | else |
2288 | { |
2289 | loadThis(compiler); |
2290 | emitByteArg(compiler, isLoad ? CODE_LOAD_FIELD : CODE_STORE_FIELD, field); |
2291 | } |
2292 | |
2293 | allowLineBeforeDot(compiler); |
2294 | } |
2295 | |
2296 | // Compiles a read or assignment to [variable]. |
2297 | static void bareName(Compiler* compiler, bool canAssign, Variable variable) |
2298 | { |
2299 | // If there's an "=" after a bare name, it's a variable assignment. |
2300 | if (canAssign && match(compiler, TOKEN_EQ)) |
2301 | { |
2302 | // Compile the right-hand side. |
2303 | expression(compiler); |
2304 | |
2305 | // Emit the store instruction. |
2306 | switch (variable.scope) |
2307 | { |
2308 | case SCOPE_LOCAL: |
2309 | emitByteArg(compiler, CODE_STORE_LOCAL, variable.index); |
2310 | break; |
2311 | case SCOPE_UPVALUE: |
2312 | emitByteArg(compiler, CODE_STORE_UPVALUE, variable.index); |
2313 | break; |
2314 | case SCOPE_MODULE: |
2315 | emitShortArg(compiler, CODE_STORE_MODULE_VAR, variable.index); |
2316 | break; |
2317 | default: |
2318 | UNREACHABLE(); |
2319 | } |
2320 | return; |
2321 | } |
2322 | |
2323 | // Emit the load instruction. |
2324 | loadVariable(compiler, variable); |
2325 | |
2326 | allowLineBeforeDot(compiler); |
2327 | } |
2328 | |
2329 | static void staticField(Compiler* compiler, bool canAssign) |
2330 | { |
2331 | Compiler* classCompiler = getEnclosingClassCompiler(compiler); |
2332 | if (classCompiler == NULL) |
2333 | { |
2334 | error(compiler, "Cannot use a static field outside of a class definition." ); |
2335 | return; |
2336 | } |
2337 | |
2338 | // Look up the name in the scope chain. |
2339 | Token* token = &compiler->parser->previous; |
2340 | |
2341 | // If this is the first time we've seen this static field, implicitly |
2342 | // define it as a variable in the scope surrounding the class definition. |
2343 | if (resolveLocal(classCompiler, token->start, token->length) == -1) |
2344 | { |
2345 | int symbol = declareVariable(classCompiler, NULL); |
2346 | |
2347 | // Implicitly initialize it to null. |
2348 | emitOp(classCompiler, CODE_NULL); |
2349 | defineVariable(classCompiler, symbol); |
2350 | } |
2351 | |
2352 | // It definitely exists now, so resolve it properly. This is different from |
2353 | // the above resolveLocal() call because we may have already closed over it |
2354 | // as an upvalue. |
2355 | Variable variable = resolveName(compiler, token->start, token->length); |
2356 | bareName(compiler, canAssign, variable); |
2357 | } |
2358 | |
2359 | // Compiles a variable name or method call with an implicit receiver. |
2360 | static void name(Compiler* compiler, bool canAssign) |
2361 | { |
2362 | // Look for the name in the scope chain up to the nearest enclosing method. |
2363 | Token* token = &compiler->parser->previous; |
2364 | |
2365 | Variable variable = resolveNonmodule(compiler, token->start, token->length); |
2366 | if (variable.index != -1) |
2367 | { |
2368 | bareName(compiler, canAssign, variable); |
2369 | return; |
2370 | } |
2371 | |
2372 | // TODO: The fact that we return above here if the variable is known and parse |
2373 | // an optional argument list below if not means that the grammar is not |
2374 | // context-free. A line of code in a method like "someName(foo)" is a parse |
2375 | // error if "someName" is a defined variable in the surrounding scope and not |
2376 | // if it isn't. Fix this. One option is to have "someName(foo)" always |
2377 | // resolve to a self-call if there is an argument list, but that makes |
2378 | // getters a little confusing. |
2379 | |
2380 | // If we're inside a method and the name is lowercase, treat it as a method |
2381 | // on this. |
2382 | if (wrenIsLocalName(token->start) && getEnclosingClass(compiler) != NULL) |
2383 | { |
2384 | loadThis(compiler); |
2385 | namedCall(compiler, canAssign, CODE_CALL_0); |
2386 | return; |
2387 | } |
2388 | |
2389 | // Otherwise, look for a module-level variable with the name. |
2390 | variable.scope = SCOPE_MODULE; |
2391 | variable.index = wrenSymbolTableFind(&compiler->parser->module->variableNames, |
2392 | token->start, token->length); |
2393 | if (variable.index == -1) |
2394 | { |
2395 | // Implicitly define a module-level variable in |
2396 | // the hopes that we get a real definition later. |
2397 | variable.index = wrenDeclareVariable(compiler->parser->vm, |
2398 | compiler->parser->module, |
2399 | token->start, token->length, |
2400 | token->line); |
2401 | |
2402 | if (variable.index == -2) |
2403 | { |
2404 | error(compiler, "Too many module variables defined." ); |
2405 | } |
2406 | } |
2407 | |
2408 | bareName(compiler, canAssign, variable); |
2409 | } |
2410 | |
2411 | static void null(Compiler* compiler, bool canAssign) |
2412 | { |
2413 | emitOp(compiler, CODE_NULL); |
2414 | } |
2415 | |
2416 | // A number or string literal. |
2417 | static void literal(Compiler* compiler, bool canAssign) |
2418 | { |
2419 | emitConstant(compiler, compiler->parser->previous.value); |
2420 | } |
2421 | |
2422 | // A string literal that contains interpolated expressions. |
2423 | // |
2424 | // Interpolation is syntactic sugar for calling ".join()" on a list. So the |
2425 | // string: |
2426 | // |
2427 | // "a %(b + c) d" |
2428 | // |
2429 | // is compiled roughly like: |
2430 | // |
2431 | // ["a ", b + c, " d"].join() |
2432 | static void stringInterpolation(Compiler* compiler, bool canAssign) |
2433 | { |
2434 | // Instantiate a new list. |
2435 | loadCoreVariable(compiler, "List" ); |
2436 | callMethod(compiler, 0, "new()" , 5); |
2437 | |
2438 | do |
2439 | { |
2440 | // The opening string part. |
2441 | literal(compiler, false); |
2442 | callMethod(compiler, 1, "addCore_(_)" , 11); |
2443 | |
2444 | // The interpolated expression. |
2445 | ignoreNewlines(compiler); |
2446 | expression(compiler); |
2447 | callMethod(compiler, 1, "addCore_(_)" , 11); |
2448 | |
2449 | ignoreNewlines(compiler); |
2450 | } while (match(compiler, TOKEN_INTERPOLATION)); |
2451 | |
2452 | // The trailing string part. |
2453 | consume(compiler, TOKEN_STRING, "Expect end of string interpolation." ); |
2454 | literal(compiler, false); |
2455 | callMethod(compiler, 1, "addCore_(_)" , 11); |
2456 | |
2457 | // The list of interpolated parts. |
2458 | callMethod(compiler, 0, "join()" , 6); |
2459 | } |
2460 | |
2461 | static void super_(Compiler* compiler, bool canAssign) |
2462 | { |
2463 | ClassInfo* enclosingClass = getEnclosingClass(compiler); |
2464 | if (enclosingClass == NULL) |
2465 | { |
2466 | error(compiler, "Cannot use 'super' outside of a method." ); |
2467 | } |
2468 | |
2469 | loadThis(compiler); |
2470 | |
2471 | // TODO: Super operator calls. |
2472 | // TODO: There's no syntax for invoking a superclass constructor with a |
2473 | // different name from the enclosing one. Figure that out. |
2474 | |
2475 | // See if it's a named super call, or an unnamed one. |
2476 | if (match(compiler, TOKEN_DOT)) |
2477 | { |
2478 | // Compile the superclass call. |
2479 | consume(compiler, TOKEN_NAME, "Expect method name after 'super.'." ); |
2480 | namedCall(compiler, canAssign, CODE_SUPER_0); |
2481 | } |
2482 | else if (enclosingClass != NULL) |
2483 | { |
2484 | // No explicit name, so use the name of the enclosing method. Make sure we |
2485 | // check that enclosingClass isn't NULL first. We've already reported the |
2486 | // error, but we don't want to crash here. |
2487 | methodCall(compiler, CODE_SUPER_0, enclosingClass->signature); |
2488 | } |
2489 | } |
2490 | |
2491 | static void this_(Compiler* compiler, bool canAssign) |
2492 | { |
2493 | if (getEnclosingClass(compiler) == NULL) |
2494 | { |
2495 | error(compiler, "Cannot use 'this' outside of a method." ); |
2496 | return; |
2497 | } |
2498 | |
2499 | loadThis(compiler); |
2500 | } |
2501 | |
2502 | // Subscript or "array indexing" operator like `foo[bar]`. |
2503 | static void subscript(Compiler* compiler, bool canAssign) |
2504 | { |
2505 | Signature signature = { "" , 0, SIG_SUBSCRIPT, 0 }; |
2506 | |
2507 | // Parse the argument list. |
2508 | finishArgumentList(compiler, &signature); |
2509 | consume(compiler, TOKEN_RIGHT_BRACKET, "Expect ']' after arguments." ); |
2510 | |
2511 | allowLineBeforeDot(compiler); |
2512 | |
2513 | if (canAssign && match(compiler, TOKEN_EQ)) |
2514 | { |
2515 | signature.type = SIG_SUBSCRIPT_SETTER; |
2516 | |
2517 | // Compile the assigned value. |
2518 | validateNumParameters(compiler, ++signature.arity); |
2519 | expression(compiler); |
2520 | } |
2521 | |
2522 | callSignature(compiler, CODE_CALL_0, &signature); |
2523 | } |
2524 | |
2525 | static void call(Compiler* compiler, bool canAssign) |
2526 | { |
2527 | ignoreNewlines(compiler); |
2528 | consume(compiler, TOKEN_NAME, "Expect method name after '.'." ); |
2529 | namedCall(compiler, canAssign, CODE_CALL_0); |
2530 | } |
2531 | |
2532 | static void and_(Compiler* compiler, bool canAssign) |
2533 | { |
2534 | ignoreNewlines(compiler); |
2535 | |
2536 | // Skip the right argument if the left is false. |
2537 | int jump = emitJump(compiler, CODE_AND); |
2538 | parsePrecedence(compiler, PREC_LOGICAL_AND); |
2539 | patchJump(compiler, jump); |
2540 | } |
2541 | |
2542 | static void or_(Compiler* compiler, bool canAssign) |
2543 | { |
2544 | ignoreNewlines(compiler); |
2545 | |
2546 | // Skip the right argument if the left is true. |
2547 | int jump = emitJump(compiler, CODE_OR); |
2548 | parsePrecedence(compiler, PREC_LOGICAL_OR); |
2549 | patchJump(compiler, jump); |
2550 | } |
2551 | |
2552 | static void conditional(Compiler* compiler, bool canAssign) |
2553 | { |
2554 | // Ignore newline after '?'. |
2555 | ignoreNewlines(compiler); |
2556 | |
2557 | // Jump to the else branch if the condition is false. |
2558 | int ifJump = emitJump(compiler, CODE_JUMP_IF); |
2559 | |
2560 | // Compile the then branch. |
2561 | parsePrecedence(compiler, PREC_CONDITIONAL); |
2562 | |
2563 | consume(compiler, TOKEN_COLON, |
2564 | "Expect ':' after then branch of conditional operator." ); |
2565 | ignoreNewlines(compiler); |
2566 | |
2567 | // Jump over the else branch when the if branch is taken. |
2568 | int elseJump = emitJump(compiler, CODE_JUMP); |
2569 | |
2570 | // Compile the else branch. |
2571 | patchJump(compiler, ifJump); |
2572 | |
2573 | parsePrecedence(compiler, PREC_ASSIGNMENT); |
2574 | |
2575 | // Patch the jump over the else. |
2576 | patchJump(compiler, elseJump); |
2577 | } |
2578 | |
2579 | void infixOp(Compiler* compiler, bool canAssign) |
2580 | { |
2581 | GrammarRule* rule = getRule(compiler->parser->previous.type); |
2582 | |
2583 | // An infix operator cannot end an expression. |
2584 | ignoreNewlines(compiler); |
2585 | |
2586 | // Compile the right-hand side. |
2587 | parsePrecedence(compiler, (Precedence)(rule->precedence + 1)); |
2588 | |
2589 | // Call the operator method on the left-hand side. |
2590 | Signature signature = { rule->name, (int)strlen(rule->name), SIG_METHOD, 1 }; |
2591 | callSignature(compiler, CODE_CALL_0, &signature); |
2592 | } |
2593 | |
2594 | // Compiles a method signature for an infix operator. |
2595 | void infixSignature(Compiler* compiler, Signature* signature) |
2596 | { |
2597 | // Add the RHS parameter. |
2598 | signature->type = SIG_METHOD; |
2599 | signature->arity = 1; |
2600 | |
2601 | // Parse the parameter name. |
2602 | consume(compiler, TOKEN_LEFT_PAREN, "Expect '(' after operator name." ); |
2603 | declareNamedVariable(compiler); |
2604 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after parameter name." ); |
2605 | } |
2606 | |
2607 | // Compiles a method signature for an unary operator (i.e. "!"). |
2608 | void unarySignature(Compiler* compiler, Signature* signature) |
2609 | { |
2610 | // Do nothing. The name is already complete. |
2611 | signature->type = SIG_GETTER; |
2612 | } |
2613 | |
2614 | // Compiles a method signature for an operator that can either be unary or |
2615 | // infix (i.e. "-"). |
2616 | void mixedSignature(Compiler* compiler, Signature* signature) |
2617 | { |
2618 | signature->type = SIG_GETTER; |
2619 | |
2620 | // If there is a parameter, it's an infix operator, otherwise it's unary. |
2621 | if (match(compiler, TOKEN_LEFT_PAREN)) |
2622 | { |
2623 | // Add the RHS parameter. |
2624 | signature->type = SIG_METHOD; |
2625 | signature->arity = 1; |
2626 | |
2627 | // Parse the parameter name. |
2628 | declareNamedVariable(compiler); |
2629 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after parameter name." ); |
2630 | } |
2631 | } |
2632 | |
2633 | // Compiles an optional setter parameter in a method [signature]. |
2634 | // |
2635 | // Returns `true` if it was a setter. |
2636 | static bool maybeSetter(Compiler* compiler, Signature* signature) |
2637 | { |
2638 | // See if it's a setter. |
2639 | if (!match(compiler, TOKEN_EQ)) return false; |
2640 | |
2641 | // It's a setter. |
2642 | if (signature->type == SIG_SUBSCRIPT) |
2643 | { |
2644 | signature->type = SIG_SUBSCRIPT_SETTER; |
2645 | } |
2646 | else |
2647 | { |
2648 | signature->type = SIG_SETTER; |
2649 | } |
2650 | |
2651 | // Parse the value parameter. |
2652 | consume(compiler, TOKEN_LEFT_PAREN, "Expect '(' after '='." ); |
2653 | declareNamedVariable(compiler); |
2654 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after parameter name." ); |
2655 | |
2656 | signature->arity++; |
2657 | |
2658 | return true; |
2659 | } |
2660 | |
2661 | // Compiles a method signature for a subscript operator. |
2662 | void subscriptSignature(Compiler* compiler, Signature* signature) |
2663 | { |
2664 | signature->type = SIG_SUBSCRIPT; |
2665 | |
2666 | // The signature currently has "[" as its name since that was the token that |
2667 | // matched it. Clear that out. |
2668 | signature->length = 0; |
2669 | |
2670 | // Parse the parameters inside the subscript. |
2671 | finishParameterList(compiler, signature); |
2672 | consume(compiler, TOKEN_RIGHT_BRACKET, "Expect ']' after parameters." ); |
2673 | |
2674 | maybeSetter(compiler, signature); |
2675 | } |
2676 | |
2677 | // Parses an optional parenthesized parameter list. Updates `type` and `arity` |
2678 | // in [signature] to match what was parsed. |
2679 | static void parameterList(Compiler* compiler, Signature* signature) |
2680 | { |
2681 | // The parameter list is optional. |
2682 | if (!match(compiler, TOKEN_LEFT_PAREN)) return; |
2683 | |
2684 | signature->type = SIG_METHOD; |
2685 | |
2686 | // Allow new line before an empty argument list |
2687 | ignoreNewlines(compiler); |
2688 | |
2689 | // Allow an empty parameter list. |
2690 | if (match(compiler, TOKEN_RIGHT_PAREN)) return; |
2691 | |
2692 | finishParameterList(compiler, signature); |
2693 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after parameters." ); |
2694 | } |
2695 | |
2696 | // Compiles a method signature for a named method or setter. |
2697 | void namedSignature(Compiler* compiler, Signature* signature) |
2698 | { |
2699 | signature->type = SIG_GETTER; |
2700 | |
2701 | // If it's a setter, it can't also have a parameter list. |
2702 | if (maybeSetter(compiler, signature)) return; |
2703 | |
2704 | // Regular named method with an optional parameter list. |
2705 | parameterList(compiler, signature); |
2706 | } |
2707 | |
2708 | // Compiles a method signature for a constructor. |
2709 | void constructorSignature(Compiler* compiler, Signature* signature) |
2710 | { |
2711 | consume(compiler, TOKEN_NAME, "Expect constructor name after 'construct'." ); |
2712 | |
2713 | // Capture the name. |
2714 | *signature = signatureFromToken(compiler, SIG_INITIALIZER); |
2715 | |
2716 | if (match(compiler, TOKEN_EQ)) |
2717 | { |
2718 | error(compiler, "A constructor cannot be a setter." ); |
2719 | } |
2720 | |
2721 | if (!match(compiler, TOKEN_LEFT_PAREN)) |
2722 | { |
2723 | error(compiler, "A constructor cannot be a getter." ); |
2724 | return; |
2725 | } |
2726 | |
2727 | // Allow an empty parameter list. |
2728 | if (match(compiler, TOKEN_RIGHT_PAREN)) return; |
2729 | |
2730 | finishParameterList(compiler, signature); |
2731 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after parameters." ); |
2732 | } |
2733 | |
2734 | // This table defines all of the parsing rules for the prefix and infix |
2735 | // expressions in the grammar. Expressions are parsed using a Pratt parser. |
2736 | // |
2737 | // See: http://journal.stuffwithstuff.com/2011/03/19/pratt-parsers-expression-parsing-made-easy/ |
2738 | #define UNUSED { NULL, NULL, NULL, PREC_NONE, NULL } |
2739 | #define PREFIX(fn) { fn, NULL, NULL, PREC_NONE, NULL } |
2740 | #define INFIX(prec, fn) { NULL, fn, NULL, prec, NULL } |
2741 | #define INFIX_OPERATOR(prec, name) { NULL, infixOp, infixSignature, prec, name } |
2742 | #define PREFIX_OPERATOR(name) { unaryOp, NULL, unarySignature, PREC_NONE, name } |
2743 | #define OPERATOR(name) { unaryOp, infixOp, mixedSignature, PREC_TERM, name } |
2744 | |
2745 | GrammarRule rules[] = |
2746 | { |
2747 | /* TOKEN_LEFT_PAREN */ PREFIX(grouping), |
2748 | /* TOKEN_RIGHT_PAREN */ UNUSED, |
2749 | /* TOKEN_LEFT_BRACKET */ { list, subscript, subscriptSignature, PREC_CALL, NULL }, |
2750 | /* TOKEN_RIGHT_BRACKET */ UNUSED, |
2751 | /* TOKEN_LEFT_BRACE */ PREFIX(map), |
2752 | /* TOKEN_RIGHT_BRACE */ UNUSED, |
2753 | /* TOKEN_COLON */ UNUSED, |
2754 | /* TOKEN_DOT */ INFIX(PREC_CALL, call), |
2755 | /* TOKEN_DOTDOT */ INFIX_OPERATOR(PREC_RANGE, ".." ), |
2756 | /* TOKEN_DOTDOTDOT */ INFIX_OPERATOR(PREC_RANGE, "..." ), |
2757 | /* TOKEN_COMMA */ UNUSED, |
2758 | /* TOKEN_STAR */ INFIX_OPERATOR(PREC_FACTOR, "*" ), |
2759 | /* TOKEN_SLASH */ INFIX_OPERATOR(PREC_FACTOR, "/" ), |
2760 | /* TOKEN_PERCENT */ INFIX_OPERATOR(PREC_FACTOR, "%" ), |
2761 | /* TOKEN_HASH */ UNUSED, |
2762 | /* TOKEN_PLUS */ INFIX_OPERATOR(PREC_TERM, "+" ), |
2763 | /* TOKEN_MINUS */ OPERATOR("-" ), |
2764 | /* TOKEN_LTLT */ INFIX_OPERATOR(PREC_BITWISE_SHIFT, "<<" ), |
2765 | /* TOKEN_GTGT */ INFIX_OPERATOR(PREC_BITWISE_SHIFT, ">>" ), |
2766 | /* TOKEN_PIPE */ INFIX_OPERATOR(PREC_BITWISE_OR, "|" ), |
2767 | /* TOKEN_PIPEPIPE */ INFIX(PREC_LOGICAL_OR, or_), |
2768 | /* TOKEN_CARET */ INFIX_OPERATOR(PREC_BITWISE_XOR, "^" ), |
2769 | /* TOKEN_AMP */ INFIX_OPERATOR(PREC_BITWISE_AND, "&" ), |
2770 | /* TOKEN_AMPAMP */ INFIX(PREC_LOGICAL_AND, and_), |
2771 | /* TOKEN_BANG */ PREFIX_OPERATOR("!" ), |
2772 | /* TOKEN_TILDE */ PREFIX_OPERATOR("~" ), |
2773 | /* TOKEN_QUESTION */ INFIX(PREC_ASSIGNMENT, conditional), |
2774 | /* TOKEN_EQ */ UNUSED, |
2775 | /* TOKEN_LT */ INFIX_OPERATOR(PREC_COMPARISON, "<" ), |
2776 | /* TOKEN_GT */ INFIX_OPERATOR(PREC_COMPARISON, ">" ), |
2777 | /* TOKEN_LTEQ */ INFIX_OPERATOR(PREC_COMPARISON, "<=" ), |
2778 | /* TOKEN_GTEQ */ INFIX_OPERATOR(PREC_COMPARISON, ">=" ), |
2779 | /* TOKEN_EQEQ */ INFIX_OPERATOR(PREC_EQUALITY, "==" ), |
2780 | /* TOKEN_BANGEQ */ INFIX_OPERATOR(PREC_EQUALITY, "!=" ), |
2781 | /* TOKEN_BREAK */ UNUSED, |
2782 | /* TOKEN_CONTINUE */ UNUSED, |
2783 | /* TOKEN_CLASS */ UNUSED, |
2784 | /* TOKEN_CONSTRUCT */ { NULL, NULL, constructorSignature, PREC_NONE, NULL }, |
2785 | /* TOKEN_ELSE */ UNUSED, |
2786 | /* TOKEN_FALSE */ PREFIX(boolean), |
2787 | /* TOKEN_FOR */ UNUSED, |
2788 | /* TOKEN_FOREIGN */ UNUSED, |
2789 | /* TOKEN_IF */ UNUSED, |
2790 | /* TOKEN_IMPORT */ UNUSED, |
2791 | /* TOKEN_AS */ UNUSED, |
2792 | /* TOKEN_IN */ UNUSED, |
2793 | /* TOKEN_IS */ INFIX_OPERATOR(PREC_IS, "is" ), |
2794 | /* TOKEN_NULL */ PREFIX(null), |
2795 | /* TOKEN_RETURN */ UNUSED, |
2796 | /* TOKEN_STATIC */ UNUSED, |
2797 | /* TOKEN_SUPER */ PREFIX(super_), |
2798 | /* TOKEN_THIS */ PREFIX(this_), |
2799 | /* TOKEN_TRUE */ PREFIX(boolean), |
2800 | /* TOKEN_VAR */ UNUSED, |
2801 | /* TOKEN_WHILE */ UNUSED, |
2802 | /* TOKEN_FIELD */ PREFIX(field), |
2803 | /* TOKEN_STATIC_FIELD */ PREFIX(staticField), |
2804 | /* TOKEN_NAME */ { name, NULL, namedSignature, PREC_NONE, NULL }, |
2805 | /* TOKEN_NUMBER */ PREFIX(literal), |
2806 | /* TOKEN_STRING */ PREFIX(literal), |
2807 | /* TOKEN_INTERPOLATION */ PREFIX(stringInterpolation), |
2808 | /* TOKEN_LINE */ UNUSED, |
2809 | /* TOKEN_ERROR */ UNUSED, |
2810 | /* TOKEN_EOF */ UNUSED |
2811 | }; |
2812 | |
2813 | // Gets the [GrammarRule] associated with tokens of [type]. |
2814 | static GrammarRule* getRule(TokenType type) |
2815 | { |
2816 | return &rules[type]; |
2817 | } |
2818 | |
2819 | // The main entrypoint for the top-down operator precedence parser. |
2820 | void parsePrecedence(Compiler* compiler, Precedence precedence) |
2821 | { |
2822 | nextToken(compiler->parser); |
2823 | GrammarFn prefix = rules[compiler->parser->previous.type].prefix; |
2824 | |
2825 | if (prefix == NULL) |
2826 | { |
2827 | error(compiler, "Expected expression." ); |
2828 | return; |
2829 | } |
2830 | |
2831 | // Track if the precendence of the surrounding expression is low enough to |
2832 | // allow an assignment inside this one. We can't compile an assignment like |
2833 | // a normal expression because it requires us to handle the LHS specially -- |
2834 | // it needs to be an lvalue, not an rvalue. So, for each of the kinds of |
2835 | // expressions that are valid lvalues -- names, subscripts, fields, etc. -- |
2836 | // we pass in whether or not it appears in a context loose enough to allow |
2837 | // "=". If so, it will parse the "=" itself and handle it appropriately. |
2838 | bool canAssign = precedence <= PREC_CONDITIONAL; |
2839 | prefix(compiler, canAssign); |
2840 | |
2841 | while (precedence <= rules[compiler->parser->current.type].precedence) |
2842 | { |
2843 | nextToken(compiler->parser); |
2844 | GrammarFn infix = rules[compiler->parser->previous.type].infix; |
2845 | infix(compiler, canAssign); |
2846 | } |
2847 | } |
2848 | |
2849 | // Parses an expression. Unlike statements, expressions leave a resulting value |
2850 | // on the stack. |
2851 | void expression(Compiler* compiler) |
2852 | { |
2853 | parsePrecedence(compiler, PREC_LOWEST); |
2854 | } |
2855 | |
2856 | // Returns the number of bytes for the arguments to the instruction |
2857 | // at [ip] in [fn]'s bytecode. |
2858 | static int getByteCountForArguments(const uint8_t* bytecode, |
2859 | const Value* constants, int ip) |
2860 | { |
2861 | Code instruction = (Code)bytecode[ip]; |
2862 | switch (instruction) |
2863 | { |
2864 | case CODE_NULL: |
2865 | case CODE_FALSE: |
2866 | case CODE_TRUE: |
2867 | case CODE_POP: |
2868 | case CODE_CLOSE_UPVALUE: |
2869 | case CODE_RETURN: |
2870 | case CODE_END: |
2871 | case CODE_LOAD_LOCAL_0: |
2872 | case CODE_LOAD_LOCAL_1: |
2873 | case CODE_LOAD_LOCAL_2: |
2874 | case CODE_LOAD_LOCAL_3: |
2875 | case CODE_LOAD_LOCAL_4: |
2876 | case CODE_LOAD_LOCAL_5: |
2877 | case CODE_LOAD_LOCAL_6: |
2878 | case CODE_LOAD_LOCAL_7: |
2879 | case CODE_LOAD_LOCAL_8: |
2880 | case CODE_CONSTRUCT: |
2881 | case CODE_FOREIGN_CONSTRUCT: |
2882 | case CODE_FOREIGN_CLASS: |
2883 | case CODE_END_MODULE: |
2884 | case CODE_END_CLASS: |
2885 | return 0; |
2886 | |
2887 | case CODE_LOAD_LOCAL: |
2888 | case CODE_STORE_LOCAL: |
2889 | case CODE_LOAD_UPVALUE: |
2890 | case CODE_STORE_UPVALUE: |
2891 | case CODE_LOAD_FIELD_THIS: |
2892 | case CODE_STORE_FIELD_THIS: |
2893 | case CODE_LOAD_FIELD: |
2894 | case CODE_STORE_FIELD: |
2895 | case CODE_CLASS: |
2896 | return 1; |
2897 | |
2898 | case CODE_CONSTANT: |
2899 | case CODE_LOAD_MODULE_VAR: |
2900 | case CODE_STORE_MODULE_VAR: |
2901 | case CODE_CALL_0: |
2902 | case CODE_CALL_1: |
2903 | case CODE_CALL_2: |
2904 | case CODE_CALL_3: |
2905 | case CODE_CALL_4: |
2906 | case CODE_CALL_5: |
2907 | case CODE_CALL_6: |
2908 | case CODE_CALL_7: |
2909 | case CODE_CALL_8: |
2910 | case CODE_CALL_9: |
2911 | case CODE_CALL_10: |
2912 | case CODE_CALL_11: |
2913 | case CODE_CALL_12: |
2914 | case CODE_CALL_13: |
2915 | case CODE_CALL_14: |
2916 | case CODE_CALL_15: |
2917 | case CODE_CALL_16: |
2918 | case CODE_JUMP: |
2919 | case CODE_LOOP: |
2920 | case CODE_JUMP_IF: |
2921 | case CODE_AND: |
2922 | case CODE_OR: |
2923 | case CODE_METHOD_INSTANCE: |
2924 | case CODE_METHOD_STATIC: |
2925 | case CODE_IMPORT_MODULE: |
2926 | case CODE_IMPORT_VARIABLE: |
2927 | return 2; |
2928 | |
2929 | case CODE_SUPER_0: |
2930 | case CODE_SUPER_1: |
2931 | case CODE_SUPER_2: |
2932 | case CODE_SUPER_3: |
2933 | case CODE_SUPER_4: |
2934 | case CODE_SUPER_5: |
2935 | case CODE_SUPER_6: |
2936 | case CODE_SUPER_7: |
2937 | case CODE_SUPER_8: |
2938 | case CODE_SUPER_9: |
2939 | case CODE_SUPER_10: |
2940 | case CODE_SUPER_11: |
2941 | case CODE_SUPER_12: |
2942 | case CODE_SUPER_13: |
2943 | case CODE_SUPER_14: |
2944 | case CODE_SUPER_15: |
2945 | case CODE_SUPER_16: |
2946 | return 4; |
2947 | |
2948 | case CODE_CLOSURE: |
2949 | { |
2950 | int constant = (bytecode[ip + 1] << 8) | bytecode[ip + 2]; |
2951 | ObjFn* loadedFn = AS_FN(constants[constant]); |
2952 | |
2953 | // There are two bytes for the constant, then two for each upvalue. |
2954 | return 2 + (loadedFn->numUpvalues * 2); |
2955 | } |
2956 | } |
2957 | |
2958 | UNREACHABLE(); |
2959 | return 0; |
2960 | } |
2961 | |
2962 | // Marks the beginning of a loop. Keeps track of the current instruction so we |
2963 | // know what to loop back to at the end of the body. |
2964 | static void startLoop(Compiler* compiler, Loop* loop) |
2965 | { |
2966 | loop->enclosing = compiler->loop; |
2967 | loop->start = compiler->fn->code.count - 1; |
2968 | loop->scopeDepth = compiler->scopeDepth; |
2969 | compiler->loop = loop; |
2970 | } |
2971 | |
2972 | // Emits the [CODE_JUMP_IF] instruction used to test the loop condition and |
2973 | // potentially exit the loop. Keeps track of the instruction so we can patch it |
2974 | // later once we know where the end of the body is. |
2975 | static void testExitLoop(Compiler* compiler) |
2976 | { |
2977 | compiler->loop->exitJump = emitJump(compiler, CODE_JUMP_IF); |
2978 | } |
2979 | |
2980 | // Compiles the body of the loop and tracks its extent so that contained "break" |
2981 | // statements can be handled correctly. |
2982 | static void loopBody(Compiler* compiler) |
2983 | { |
2984 | compiler->loop->body = compiler->fn->code.count; |
2985 | statement(compiler); |
2986 | } |
2987 | |
2988 | // Ends the current innermost loop. Patches up all jumps and breaks now that |
2989 | // we know where the end of the loop is. |
2990 | static void endLoop(Compiler* compiler) |
2991 | { |
2992 | // We don't check for overflow here since the forward jump over the loop body |
2993 | // will report an error for the same problem. |
2994 | int loopOffset = compiler->fn->code.count - compiler->loop->start + 2; |
2995 | emitShortArg(compiler, CODE_LOOP, loopOffset); |
2996 | |
2997 | patchJump(compiler, compiler->loop->exitJump); |
2998 | |
2999 | // Find any break placeholder instructions (which will be CODE_END in the |
3000 | // bytecode) and replace them with real jumps. |
3001 | int i = compiler->loop->body; |
3002 | while (i < compiler->fn->code.count) |
3003 | { |
3004 | if (compiler->fn->code.data[i] == CODE_END) |
3005 | { |
3006 | compiler->fn->code.data[i] = CODE_JUMP; |
3007 | patchJump(compiler, i + 1); |
3008 | i += 3; |
3009 | } |
3010 | else |
3011 | { |
3012 | // Skip this instruction and its arguments. |
3013 | i += 1 + getByteCountForArguments(compiler->fn->code.data, |
3014 | compiler->fn->constants.data, i); |
3015 | } |
3016 | } |
3017 | |
3018 | compiler->loop = compiler->loop->enclosing; |
3019 | } |
3020 | |
3021 | static void forStatement(Compiler* compiler) |
3022 | { |
3023 | // A for statement like: |
3024 | // |
3025 | // for (i in sequence.expression) { |
3026 | // System.print(i) |
3027 | // } |
3028 | // |
3029 | // Is compiled to bytecode almost as if the source looked like this: |
3030 | // |
3031 | // { |
3032 | // var seq_ = sequence.expression |
3033 | // var iter_ |
3034 | // while (iter_ = seq_.iterate(iter_)) { |
3035 | // var i = seq_.iteratorValue(iter_) |
3036 | // System.print(i) |
3037 | // } |
3038 | // } |
3039 | // |
3040 | // It's not exactly this, because the synthetic variables `seq_` and `iter_` |
3041 | // actually get names that aren't valid Wren identfiers, but that's the basic |
3042 | // idea. |
3043 | // |
3044 | // The important parts are: |
3045 | // - The sequence expression is only evaluated once. |
3046 | // - The .iterate() method is used to advance the iterator and determine if |
3047 | // it should exit the loop. |
3048 | // - The .iteratorValue() method is used to get the value at the current |
3049 | // iterator position. |
3050 | |
3051 | // Create a scope for the hidden local variables used for the iterator. |
3052 | pushScope(compiler); |
3053 | |
3054 | consume(compiler, TOKEN_LEFT_PAREN, "Expect '(' after 'for'." ); |
3055 | consume(compiler, TOKEN_NAME, "Expect for loop variable name." ); |
3056 | |
3057 | // Remember the name of the loop variable. |
3058 | const char* name = compiler->parser->previous.start; |
3059 | int length = compiler->parser->previous.length; |
3060 | |
3061 | consume(compiler, TOKEN_IN, "Expect 'in' after loop variable." ); |
3062 | ignoreNewlines(compiler); |
3063 | |
3064 | // Evaluate the sequence expression and store it in a hidden local variable. |
3065 | // The space in the variable name ensures it won't collide with a user-defined |
3066 | // variable. |
3067 | expression(compiler); |
3068 | |
3069 | // Verify that there is space to hidden local variables. |
3070 | // Note that we expect only two addLocal calls next to each other in the |
3071 | // following code. |
3072 | if (compiler->numLocals + 2 > MAX_LOCALS) |
3073 | { |
3074 | error(compiler, "Cannot declare more than %d variables in one scope. (Not enough space for for-loops internal variables)" , |
3075 | MAX_LOCALS); |
3076 | return; |
3077 | } |
3078 | int seqSlot = addLocal(compiler, "seq " , 4); |
3079 | |
3080 | // Create another hidden local for the iterator object. |
3081 | null(compiler, false); |
3082 | int iterSlot = addLocal(compiler, "iter " , 5); |
3083 | |
3084 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after loop expression." ); |
3085 | |
3086 | Loop loop; |
3087 | startLoop(compiler, &loop); |
3088 | |
3089 | // Advance the iterator by calling the ".iterate" method on the sequence. |
3090 | loadLocal(compiler, seqSlot); |
3091 | loadLocal(compiler, iterSlot); |
3092 | |
3093 | // Update and test the iterator. |
3094 | callMethod(compiler, 1, "iterate(_)" , 10); |
3095 | emitByteArg(compiler, CODE_STORE_LOCAL, iterSlot); |
3096 | testExitLoop(compiler); |
3097 | |
3098 | // Get the current value in the sequence by calling ".iteratorValue". |
3099 | loadLocal(compiler, seqSlot); |
3100 | loadLocal(compiler, iterSlot); |
3101 | callMethod(compiler, 1, "iteratorValue(_)" , 16); |
3102 | |
3103 | // Bind the loop variable in its own scope. This ensures we get a fresh |
3104 | // variable each iteration so that closures for it don't all see the same one. |
3105 | pushScope(compiler); |
3106 | addLocal(compiler, name, length); |
3107 | |
3108 | loopBody(compiler); |
3109 | |
3110 | // Loop variable. |
3111 | popScope(compiler); |
3112 | |
3113 | endLoop(compiler); |
3114 | |
3115 | // Hidden variables. |
3116 | popScope(compiler); |
3117 | } |
3118 | |
3119 | static void ifStatement(Compiler* compiler) |
3120 | { |
3121 | // Compile the condition. |
3122 | consume(compiler, TOKEN_LEFT_PAREN, "Expect '(' after 'if'." ); |
3123 | expression(compiler); |
3124 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after if condition." ); |
3125 | |
3126 | // Jump to the else branch if the condition is false. |
3127 | int ifJump = emitJump(compiler, CODE_JUMP_IF); |
3128 | |
3129 | // Compile the then branch. |
3130 | statement(compiler); |
3131 | |
3132 | // Compile the else branch if there is one. |
3133 | if (match(compiler, TOKEN_ELSE)) |
3134 | { |
3135 | // Jump over the else branch when the if branch is taken. |
3136 | int elseJump = emitJump(compiler, CODE_JUMP); |
3137 | patchJump(compiler, ifJump); |
3138 | |
3139 | statement(compiler); |
3140 | |
3141 | // Patch the jump over the else. |
3142 | patchJump(compiler, elseJump); |
3143 | } |
3144 | else |
3145 | { |
3146 | patchJump(compiler, ifJump); |
3147 | } |
3148 | } |
3149 | |
3150 | static void whileStatement(Compiler* compiler) |
3151 | { |
3152 | Loop loop; |
3153 | startLoop(compiler, &loop); |
3154 | |
3155 | // Compile the condition. |
3156 | consume(compiler, TOKEN_LEFT_PAREN, "Expect '(' after 'while'." ); |
3157 | expression(compiler); |
3158 | consume(compiler, TOKEN_RIGHT_PAREN, "Expect ')' after while condition." ); |
3159 | |
3160 | testExitLoop(compiler); |
3161 | loopBody(compiler); |
3162 | endLoop(compiler); |
3163 | } |
3164 | |
3165 | // Compiles a simple statement. These can only appear at the top-level or |
3166 | // within curly blocks. Simple statements exclude variable binding statements |
3167 | // like "var" and "class" which are not allowed directly in places like the |
3168 | // branches of an "if" statement. |
3169 | // |
3170 | // Unlike expressions, statements do not leave a value on the stack. |
3171 | void statement(Compiler* compiler) |
3172 | { |
3173 | if (match(compiler, TOKEN_BREAK)) |
3174 | { |
3175 | if (compiler->loop == NULL) |
3176 | { |
3177 | error(compiler, "Cannot use 'break' outside of a loop." ); |
3178 | return; |
3179 | } |
3180 | |
3181 | // Since we will be jumping out of the scope, make sure any locals in it |
3182 | // are discarded first. |
3183 | discardLocals(compiler, compiler->loop->scopeDepth + 1); |
3184 | |
3185 | // Emit a placeholder instruction for the jump to the end of the body. When |
3186 | // we're done compiling the loop body and know where the end is, we'll |
3187 | // replace these with `CODE_JUMP` instructions with appropriate offsets. |
3188 | // We use `CODE_END` here because that can't occur in the middle of |
3189 | // bytecode. |
3190 | emitJump(compiler, CODE_END); |
3191 | } |
3192 | else if (match(compiler, TOKEN_CONTINUE)) |
3193 | { |
3194 | if (compiler->loop == NULL) |
3195 | { |
3196 | error(compiler, "Cannot use 'continue' outside of a loop." ); |
3197 | return; |
3198 | } |
3199 | |
3200 | // Since we will be jumping out of the scope, make sure any locals in it |
3201 | // are discarded first. |
3202 | discardLocals(compiler, compiler->loop->scopeDepth + 1); |
3203 | |
3204 | // emit a jump back to the top of the loop |
3205 | int loopOffset = compiler->fn->code.count - compiler->loop->start + 2; |
3206 | emitShortArg(compiler, CODE_LOOP, loopOffset); |
3207 | } |
3208 | else if (match(compiler, TOKEN_FOR)) |
3209 | { |
3210 | forStatement(compiler); |
3211 | } |
3212 | else if (match(compiler, TOKEN_IF)) |
3213 | { |
3214 | ifStatement(compiler); |
3215 | } |
3216 | else if (match(compiler, TOKEN_RETURN)) |
3217 | { |
3218 | // Compile the return value. |
3219 | if (peek(compiler) == TOKEN_LINE) |
3220 | { |
3221 | // If there's no expression after return, initializers should |
3222 | // return 'this' and regular methods should return null |
3223 | Code result = compiler->isInitializer ? CODE_LOAD_LOCAL_0 : CODE_NULL; |
3224 | emitOp(compiler, result); |
3225 | } |
3226 | else |
3227 | { |
3228 | if (compiler->isInitializer) |
3229 | { |
3230 | error(compiler, "A constructor cannot return a value." ); |
3231 | } |
3232 | |
3233 | expression(compiler); |
3234 | } |
3235 | |
3236 | emitOp(compiler, CODE_RETURN); |
3237 | } |
3238 | else if (match(compiler, TOKEN_WHILE)) |
3239 | { |
3240 | whileStatement(compiler); |
3241 | } |
3242 | else if (match(compiler, TOKEN_LEFT_BRACE)) |
3243 | { |
3244 | // Block statement. |
3245 | pushScope(compiler); |
3246 | if (finishBlock(compiler)) |
3247 | { |
3248 | // Block was an expression, so discard it. |
3249 | emitOp(compiler, CODE_POP); |
3250 | } |
3251 | popScope(compiler); |
3252 | } |
3253 | else |
3254 | { |
3255 | // Expression statement. |
3256 | expression(compiler); |
3257 | emitOp(compiler, CODE_POP); |
3258 | } |
3259 | } |
3260 | |
3261 | // Creates a matching constructor method for an initializer with [signature] |
3262 | // and [initializerSymbol]. |
3263 | // |
3264 | // Construction is a two-stage process in Wren that involves two separate |
3265 | // methods. There is a static method that allocates a new instance of the class. |
3266 | // It then invokes an initializer method on the new instance, forwarding all of |
3267 | // the constructor arguments to it. |
3268 | // |
3269 | // The allocator method always has a fixed implementation: |
3270 | // |
3271 | // CODE_CONSTRUCT - Replace the class in slot 0 with a new instance of it. |
3272 | // CODE_CALL - Invoke the initializer on the new instance. |
3273 | // |
3274 | // This creates that method and calls the initializer with [initializerSymbol]. |
3275 | static void createConstructor(Compiler* compiler, Signature* signature, |
3276 | int initializerSymbol) |
3277 | { |
3278 | Compiler methodCompiler; |
3279 | initCompiler(&methodCompiler, compiler->parser, compiler, true); |
3280 | |
3281 | // Allocate the instance. |
3282 | emitOp(&methodCompiler, compiler->enclosingClass->isForeign |
3283 | ? CODE_FOREIGN_CONSTRUCT : CODE_CONSTRUCT); |
3284 | |
3285 | // Run its initializer. |
3286 | emitShortArg(&methodCompiler, (Code)(CODE_CALL_0 + signature->arity), |
3287 | initializerSymbol); |
3288 | |
3289 | // Return the instance. |
3290 | emitOp(&methodCompiler, CODE_RETURN); |
3291 | |
3292 | endCompiler(&methodCompiler, "" , 0); |
3293 | } |
3294 | |
3295 | // Loads the enclosing class onto the stack and then binds the function already |
3296 | // on the stack as a method on that class. |
3297 | static void defineMethod(Compiler* compiler, Variable classVariable, |
3298 | bool isStatic, int methodSymbol) |
3299 | { |
3300 | // Load the class. We have to do this for each method because we can't |
3301 | // keep the class on top of the stack. If there are static fields, they |
3302 | // will be locals above the initial variable slot for the class on the |
3303 | // stack. To skip past those, we just load the class each time right before |
3304 | // defining a method. |
3305 | loadVariable(compiler, classVariable); |
3306 | |
3307 | // Define the method. |
3308 | Code instruction = isStatic ? CODE_METHOD_STATIC : CODE_METHOD_INSTANCE; |
3309 | emitShortArg(compiler, instruction, methodSymbol); |
3310 | } |
3311 | |
3312 | // Declares a method in the enclosing class with [signature]. |
3313 | // |
3314 | // Reports an error if a method with that signature is already declared. |
3315 | // Returns the symbol for the method. |
3316 | static int declareMethod(Compiler* compiler, Signature* signature, |
3317 | const char* name, int length) |
3318 | { |
3319 | int symbol = signatureSymbol(compiler, signature); |
3320 | |
3321 | // See if the class has already declared method with this signature. |
3322 | ClassInfo* classInfo = compiler->enclosingClass; |
3323 | IntBuffer* methods = classInfo->inStatic |
3324 | ? &classInfo->staticMethods : &classInfo->methods; |
3325 | for (int i = 0; i < methods->count; i++) |
3326 | { |
3327 | if (methods->data[i] == symbol) |
3328 | { |
3329 | const char* staticPrefix = classInfo->inStatic ? "static " : "" ; |
3330 | error(compiler, "Class %s already defines a %smethod '%s'." , |
3331 | &compiler->enclosingClass->name->value, staticPrefix, name); |
3332 | break; |
3333 | } |
3334 | } |
3335 | |
3336 | wrenIntBufferWrite(compiler->parser->vm, methods, symbol); |
3337 | return symbol; |
3338 | } |
3339 | |
3340 | static Value consumeLiteral(Compiler* compiler, const char* message) |
3341 | { |
3342 | if(match(compiler, TOKEN_FALSE)) return FALSE_VAL; |
3343 | if(match(compiler, TOKEN_TRUE)) return TRUE_VAL; |
3344 | if(match(compiler, TOKEN_NUMBER)) return compiler->parser->previous.value; |
3345 | if(match(compiler, TOKEN_STRING)) return compiler->parser->previous.value; |
3346 | if(match(compiler, TOKEN_NAME)) return compiler->parser->previous.value; |
3347 | |
3348 | error(compiler, message); |
3349 | nextToken(compiler->parser); |
3350 | return NULL_VAL; |
3351 | } |
3352 | |
3353 | static bool matchAttribute(Compiler* compiler) { |
3354 | |
3355 | if(match(compiler, TOKEN_HASH)) |
3356 | { |
3357 | compiler->numAttributes++; |
3358 | bool runtimeAccess = match(compiler, TOKEN_BANG); |
3359 | if(match(compiler, TOKEN_NAME)) |
3360 | { |
3361 | Value group = compiler->parser->previous.value; |
3362 | TokenType ahead = peek(compiler); |
3363 | if(ahead == TOKEN_EQ || ahead == TOKEN_LINE) |
3364 | { |
3365 | Value key = group; |
3366 | Value value = NULL_VAL; |
3367 | if(match(compiler, TOKEN_EQ)) |
3368 | { |
3369 | value = consumeLiteral(compiler, "Expect a Bool, Num, String or Identifier literal for an attribute value." ); |
3370 | } |
3371 | if(runtimeAccess) addToAttributeGroup(compiler, NULL_VAL, key, value); |
3372 | } |
3373 | else if(match(compiler, TOKEN_LEFT_PAREN)) |
3374 | { |
3375 | ignoreNewlines(compiler); |
3376 | if(match(compiler, TOKEN_RIGHT_PAREN)) |
3377 | { |
3378 | error(compiler, "Expected attributes in group, group cannot be empty." ); |
3379 | } |
3380 | else |
3381 | { |
3382 | while(peek(compiler) != TOKEN_RIGHT_PAREN) |
3383 | { |
3384 | consume(compiler, TOKEN_NAME, "Expect name for attribute key." ); |
3385 | Value key = compiler->parser->previous.value; |
3386 | Value value = NULL_VAL; |
3387 | if(match(compiler, TOKEN_EQ)) |
3388 | { |
3389 | value = consumeLiteral(compiler, "Expect a Bool, Num, String or Identifier literal for an attribute value." ); |
3390 | } |
3391 | if(runtimeAccess) addToAttributeGroup(compiler, group, key, value); |
3392 | ignoreNewlines(compiler); |
3393 | if(!match(compiler, TOKEN_COMMA)) break; |
3394 | ignoreNewlines(compiler); |
3395 | } |
3396 | |
3397 | ignoreNewlines(compiler); |
3398 | consume(compiler, TOKEN_RIGHT_PAREN, |
3399 | "Expected ')' after grouped attributes." ); |
3400 | } |
3401 | } |
3402 | else |
3403 | { |
3404 | error(compiler, "Expect an equal, newline or grouping after an attribute key." ); |
3405 | } |
3406 | } |
3407 | else |
3408 | { |
3409 | error(compiler, "Expect an attribute definition after #." ); |
3410 | } |
3411 | |
3412 | consumeLine(compiler, "Expect newline after attribute." ); |
3413 | return true; |
3414 | } |
3415 | |
3416 | return false; |
3417 | } |
3418 | |
3419 | // Compiles a method definition inside a class body. |
3420 | // |
3421 | // Returns `true` if it compiled successfully, or `false` if the method couldn't |
3422 | // be parsed. |
3423 | static bool method(Compiler* compiler, Variable classVariable) |
3424 | { |
3425 | // Parse any attributes before the method and store them |
3426 | if(matchAttribute(compiler)) { |
3427 | return method(compiler, classVariable); |
3428 | } |
3429 | |
3430 | // TODO: What about foreign constructors? |
3431 | bool isForeign = match(compiler, TOKEN_FOREIGN); |
3432 | bool isStatic = match(compiler, TOKEN_STATIC); |
3433 | compiler->enclosingClass->inStatic = isStatic; |
3434 | |
3435 | SignatureFn signatureFn = rules[compiler->parser->current.type].method; |
3436 | nextToken(compiler->parser); |
3437 | |
3438 | if (signatureFn == NULL) |
3439 | { |
3440 | error(compiler, "Expect method definition." ); |
3441 | return false; |
3442 | } |
3443 | |
3444 | // Build the method signature. |
3445 | Signature signature = signatureFromToken(compiler, SIG_GETTER); |
3446 | compiler->enclosingClass->signature = &signature; |
3447 | |
3448 | Compiler methodCompiler; |
3449 | initCompiler(&methodCompiler, compiler->parser, compiler, true); |
3450 | |
3451 | // Compile the method signature. |
3452 | signatureFn(&methodCompiler, &signature); |
3453 | |
3454 | methodCompiler.isInitializer = signature.type == SIG_INITIALIZER; |
3455 | |
3456 | if (isStatic && signature.type == SIG_INITIALIZER) |
3457 | { |
3458 | error(compiler, "A constructor cannot be static." ); |
3459 | } |
3460 | |
3461 | // Include the full signature in debug messages in stack traces. |
3462 | char fullSignature[MAX_METHOD_SIGNATURE]; |
3463 | int length; |
3464 | signatureToString(&signature, fullSignature, &length); |
3465 | |
3466 | // Copy any attributes the compiler collected into the enclosing class |
3467 | copyMethodAttributes(compiler, isForeign, isStatic, fullSignature, length); |
3468 | |
3469 | // Check for duplicate methods. Doesn't matter that it's already been |
3470 | // defined, error will discard bytecode anyway. |
3471 | // Check if the method table already contains this symbol |
3472 | int methodSymbol = declareMethod(compiler, &signature, fullSignature, length); |
3473 | |
3474 | if (isForeign) |
3475 | { |
3476 | // Define a constant for the signature. |
3477 | emitConstant(compiler, wrenNewStringLength(compiler->parser->vm, |
3478 | fullSignature, length)); |
3479 | |
3480 | // We don't need the function we started compiling in the parameter list |
3481 | // any more. |
3482 | methodCompiler.parser->vm->compiler = methodCompiler.parent; |
3483 | } |
3484 | else |
3485 | { |
3486 | consume(compiler, TOKEN_LEFT_BRACE, "Expect '{' to begin method body." ); |
3487 | finishBody(&methodCompiler); |
3488 | endCompiler(&methodCompiler, fullSignature, length); |
3489 | } |
3490 | |
3491 | // Define the method. For a constructor, this defines the instance |
3492 | // initializer method. |
3493 | defineMethod(compiler, classVariable, isStatic, methodSymbol); |
3494 | |
3495 | if (signature.type == SIG_INITIALIZER) |
3496 | { |
3497 | // Also define a matching constructor method on the metaclass. |
3498 | signature.type = SIG_METHOD; |
3499 | int constructorSymbol = signatureSymbol(compiler, &signature); |
3500 | |
3501 | createConstructor(compiler, &signature, methodSymbol); |
3502 | defineMethod(compiler, classVariable, true, constructorSymbol); |
3503 | } |
3504 | |
3505 | return true; |
3506 | } |
3507 | |
3508 | // Compiles a class definition. Assumes the "class" token has already been |
3509 | // consumed (along with a possibly preceding "foreign" token). |
3510 | static void classDefinition(Compiler* compiler, bool isForeign) |
3511 | { |
3512 | // Create a variable to store the class in. |
3513 | Variable classVariable; |
3514 | classVariable.scope = compiler->scopeDepth == -1 ? SCOPE_MODULE : SCOPE_LOCAL; |
3515 | classVariable.index = declareNamedVariable(compiler); |
3516 | |
3517 | // Create shared class name value |
3518 | Value classNameString = wrenNewStringLength(compiler->parser->vm, |
3519 | compiler->parser->previous.start, compiler->parser->previous.length); |
3520 | |
3521 | // Create class name string to track method duplicates |
3522 | ObjString* className = AS_STRING(classNameString); |
3523 | |
3524 | // Make a string constant for the name. |
3525 | emitConstant(compiler, classNameString); |
3526 | |
3527 | // Load the superclass (if there is one). |
3528 | if (match(compiler, TOKEN_IS)) |
3529 | { |
3530 | parsePrecedence(compiler, PREC_CALL); |
3531 | } |
3532 | else |
3533 | { |
3534 | // Implicitly inherit from Object. |
3535 | loadCoreVariable(compiler, "Object" ); |
3536 | } |
3537 | |
3538 | // Store a placeholder for the number of fields argument. We don't know the |
3539 | // count until we've compiled all the methods to see which fields are used. |
3540 | int numFieldsInstruction = -1; |
3541 | if (isForeign) |
3542 | { |
3543 | emitOp(compiler, CODE_FOREIGN_CLASS); |
3544 | } |
3545 | else |
3546 | { |
3547 | numFieldsInstruction = emitByteArg(compiler, CODE_CLASS, 255); |
3548 | } |
3549 | |
3550 | // Store it in its name. |
3551 | defineVariable(compiler, classVariable.index); |
3552 | |
3553 | // Push a local variable scope. Static fields in a class body are hoisted out |
3554 | // into local variables declared in this scope. Methods that use them will |
3555 | // have upvalues referencing them. |
3556 | pushScope(compiler); |
3557 | |
3558 | ClassInfo classInfo; |
3559 | classInfo.isForeign = isForeign; |
3560 | classInfo.name = className; |
3561 | |
3562 | // Allocate attribute maps if necessary. |
3563 | // A method will allocate the methods one if needed |
3564 | classInfo.classAttributes = compiler->attributes->count > 0 |
3565 | ? wrenNewMap(compiler->parser->vm) |
3566 | : NULL; |
3567 | classInfo.methodAttributes = NULL; |
3568 | // Copy any existing attributes into the class |
3569 | copyAttributes(compiler, classInfo.classAttributes); |
3570 | |
3571 | // Set up a symbol table for the class's fields. We'll initially compile |
3572 | // them to slots starting at zero. When the method is bound to the class, the |
3573 | // bytecode will be adjusted by [wrenBindMethod] to take inherited fields |
3574 | // into account. |
3575 | wrenSymbolTableInit(&classInfo.fields); |
3576 | |
3577 | // Set up symbol buffers to track duplicate static and instance methods. |
3578 | wrenIntBufferInit(&classInfo.methods); |
3579 | wrenIntBufferInit(&classInfo.staticMethods); |
3580 | compiler->enclosingClass = &classInfo; |
3581 | |
3582 | // Compile the method definitions. |
3583 | consume(compiler, TOKEN_LEFT_BRACE, "Expect '{' after class declaration." ); |
3584 | matchLine(compiler); |
3585 | |
3586 | while (!match(compiler, TOKEN_RIGHT_BRACE)) |
3587 | { |
3588 | if (!method(compiler, classVariable)) break; |
3589 | |
3590 | // Don't require a newline after the last definition. |
3591 | if (match(compiler, TOKEN_RIGHT_BRACE)) break; |
3592 | |
3593 | consumeLine(compiler, "Expect newline after definition in class." ); |
3594 | } |
3595 | |
3596 | // If any attributes are present, |
3597 | // instantiate a ClassAttributes instance for the class |
3598 | // and send it over to CODE_END_CLASS |
3599 | bool hasAttr = classInfo.classAttributes != NULL || |
3600 | classInfo.methodAttributes != NULL; |
3601 | if(hasAttr) { |
3602 | emitClassAttributes(compiler, &classInfo); |
3603 | loadVariable(compiler, classVariable); |
3604 | // At the moment, we don't have other uses for CODE_END_CLASS, |
3605 | // so we put it inside this condition. Later, we can always |
3606 | // emit it and use it as needed. |
3607 | emitOp(compiler, CODE_END_CLASS); |
3608 | } |
3609 | |
3610 | // Update the class with the number of fields. |
3611 | if (!isForeign) |
3612 | { |
3613 | compiler->fn->code.data[numFieldsInstruction] = |
3614 | (uint8_t)classInfo.fields.count; |
3615 | } |
3616 | |
3617 | // Clear symbol tables for tracking field and method names. |
3618 | wrenSymbolTableClear(compiler->parser->vm, &classInfo.fields); |
3619 | wrenIntBufferClear(compiler->parser->vm, &classInfo.methods); |
3620 | wrenIntBufferClear(compiler->parser->vm, &classInfo.staticMethods); |
3621 | compiler->enclosingClass = NULL; |
3622 | popScope(compiler); |
3623 | } |
3624 | |
3625 | // Compiles an "import" statement. |
3626 | // |
3627 | // An import compiles to a series of instructions. Given: |
3628 | // |
3629 | // import "foo" for Bar, Baz |
3630 | // |
3631 | // We compile a single IMPORT_MODULE "foo" instruction to load the module |
3632 | // itself. When that finishes executing the imported module, it leaves the |
3633 | // ObjModule in vm->lastModule. Then, for Bar and Baz, we: |
3634 | // |
3635 | // * Declare a variable in the current scope with that name. |
3636 | // * Emit an IMPORT_VARIABLE instruction to load the variable's value from the |
3637 | // other module. |
3638 | // * Compile the code to store that value in the variable in this scope. |
3639 | static void import(Compiler* compiler) |
3640 | { |
3641 | ignoreNewlines(compiler); |
3642 | consume(compiler, TOKEN_STRING, "Expect a string after 'import'." ); |
3643 | int moduleConstant = addConstant(compiler, compiler->parser->previous.value); |
3644 | |
3645 | // Load the module. |
3646 | emitShortArg(compiler, CODE_IMPORT_MODULE, moduleConstant); |
3647 | |
3648 | // Discard the unused result value from calling the module body's closure. |
3649 | emitOp(compiler, CODE_POP); |
3650 | |
3651 | // The for clause is optional. |
3652 | if (!match(compiler, TOKEN_FOR)) return; |
3653 | |
3654 | // Compile the comma-separated list of variables to import. |
3655 | do |
3656 | { |
3657 | ignoreNewlines(compiler); |
3658 | |
3659 | consume(compiler, TOKEN_NAME, "Expect variable name." ); |
3660 | |
3661 | // We need to hold onto the source variable, |
3662 | // in order to reference it in the import later |
3663 | Token sourceVariableToken = compiler->parser->previous; |
3664 | |
3665 | // Define a string constant for the original variable name. |
3666 | int sourceVariableConstant = addConstant(compiler, |
3667 | wrenNewStringLength(compiler->parser->vm, |
3668 | sourceVariableToken.start, |
3669 | sourceVariableToken.length)); |
3670 | |
3671 | // Store the symbol we care about for the variable |
3672 | int slot = -1; |
3673 | if(match(compiler, TOKEN_AS)) |
3674 | { |
3675 | //import "module" for Source as Dest |
3676 | //Use 'Dest' as the name by declaring a new variable for it. |
3677 | //This parses a name after the 'as' and defines it. |
3678 | slot = declareNamedVariable(compiler); |
3679 | } |
3680 | else |
3681 | { |
3682 | //import "module" for Source |
3683 | //Uses 'Source' as the name directly |
3684 | slot = declareVariable(compiler, &sourceVariableToken); |
3685 | } |
3686 | |
3687 | // Load the variable from the other module. |
3688 | emitShortArg(compiler, CODE_IMPORT_VARIABLE, sourceVariableConstant); |
3689 | |
3690 | // Store the result in the variable here. |
3691 | defineVariable(compiler, slot); |
3692 | } while (match(compiler, TOKEN_COMMA)); |
3693 | } |
3694 | |
3695 | // Compiles a "var" variable definition statement. |
3696 | static void variableDefinition(Compiler* compiler) |
3697 | { |
3698 | // Grab its name, but don't declare it yet. A (local) variable shouldn't be |
3699 | // in scope in its own initializer. |
3700 | consume(compiler, TOKEN_NAME, "Expect variable name." ); |
3701 | Token nameToken = compiler->parser->previous; |
3702 | |
3703 | // Compile the initializer. |
3704 | if (match(compiler, TOKEN_EQ)) |
3705 | { |
3706 | ignoreNewlines(compiler); |
3707 | expression(compiler); |
3708 | } |
3709 | else |
3710 | { |
3711 | // Default initialize it to null. |
3712 | null(compiler, false); |
3713 | } |
3714 | |
3715 | // Now put it in scope. |
3716 | int symbol = declareVariable(compiler, &nameToken); |
3717 | defineVariable(compiler, symbol); |
3718 | } |
3719 | |
3720 | // Compiles a "definition". These are the statements that bind new variables. |
3721 | // They can only appear at the top level of a block and are prohibited in places |
3722 | // like the non-curly body of an if or while. |
3723 | void definition(Compiler* compiler) |
3724 | { |
3725 | if(matchAttribute(compiler)) { |
3726 | definition(compiler); |
3727 | return; |
3728 | } |
3729 | |
3730 | if (match(compiler, TOKEN_CLASS)) |
3731 | { |
3732 | classDefinition(compiler, false); |
3733 | return; |
3734 | } |
3735 | else if (match(compiler, TOKEN_FOREIGN)) |
3736 | { |
3737 | consume(compiler, TOKEN_CLASS, "Expect 'class' after 'foreign'." ); |
3738 | classDefinition(compiler, true); |
3739 | return; |
3740 | } |
3741 | |
3742 | disallowAttributes(compiler); |
3743 | |
3744 | if (match(compiler, TOKEN_IMPORT)) |
3745 | { |
3746 | import(compiler); |
3747 | } |
3748 | else if (match(compiler, TOKEN_VAR)) |
3749 | { |
3750 | variableDefinition(compiler); |
3751 | } |
3752 | else |
3753 | { |
3754 | statement(compiler); |
3755 | } |
3756 | } |
3757 | |
3758 | ObjFn* wrenCompile(WrenVM* vm, ObjModule* module, const char* source, |
3759 | bool isExpression, bool printErrors) |
3760 | { |
3761 | // Skip the UTF-8 BOM if there is one. |
3762 | if (strncmp(source, "\xEF\xBB\xBF" , 3) == 0) source += 3; |
3763 | |
3764 | Parser parser; |
3765 | parser.vm = vm; |
3766 | parser.module = module; |
3767 | parser.source = source; |
3768 | |
3769 | parser.tokenStart = source; |
3770 | parser.currentChar = source; |
3771 | parser.currentLine = 1; |
3772 | parser.numParens = 0; |
3773 | |
3774 | // Zero-init the current token. This will get copied to previous when |
3775 | // nextToken() is called below. |
3776 | parser.next.type = TOKEN_ERROR; |
3777 | parser.next.start = source; |
3778 | parser.next.length = 0; |
3779 | parser.next.line = 0; |
3780 | parser.next.value = UNDEFINED_VAL; |
3781 | |
3782 | parser.printErrors = printErrors; |
3783 | parser.hasError = false; |
3784 | |
3785 | // Read the first token into next |
3786 | nextToken(&parser); |
3787 | // Copy next -> current |
3788 | nextToken(&parser); |
3789 | |
3790 | int numExistingVariables = module->variables.count; |
3791 | |
3792 | Compiler compiler; |
3793 | initCompiler(&compiler, &parser, NULL, false); |
3794 | ignoreNewlines(&compiler); |
3795 | |
3796 | if (isExpression) |
3797 | { |
3798 | expression(&compiler); |
3799 | consume(&compiler, TOKEN_EOF, "Expect end of expression." ); |
3800 | } |
3801 | else |
3802 | { |
3803 | while (!match(&compiler, TOKEN_EOF)) |
3804 | { |
3805 | definition(&compiler); |
3806 | |
3807 | // If there is no newline, it must be the end of file on the same line. |
3808 | if (!matchLine(&compiler)) |
3809 | { |
3810 | consume(&compiler, TOKEN_EOF, "Expect end of file." ); |
3811 | break; |
3812 | } |
3813 | } |
3814 | |
3815 | emitOp(&compiler, CODE_END_MODULE); |
3816 | } |
3817 | |
3818 | emitOp(&compiler, CODE_RETURN); |
3819 | |
3820 | // See if there are any implicitly declared module-level variables that never |
3821 | // got an explicit definition. They will have values that are numbers |
3822 | // indicating the line where the variable was first used. |
3823 | for (int i = numExistingVariables; i < parser.module->variables.count; i++) |
3824 | { |
3825 | if (IS_NUM(parser.module->variables.data[i])) |
3826 | { |
3827 | // Synthesize a token for the original use site. |
3828 | parser.previous.type = TOKEN_NAME; |
3829 | parser.previous.start = parser.module->variableNames.data[i]->value; |
3830 | parser.previous.length = parser.module->variableNames.data[i]->length; |
3831 | parser.previous.line = (int)AS_NUM(parser.module->variables.data[i]); |
3832 | error(&compiler, "Variable is used but not defined." ); |
3833 | } |
3834 | } |
3835 | |
3836 | return endCompiler(&compiler, "(script)" , 8); |
3837 | } |
3838 | |
3839 | void wrenBindMethodCode(ObjClass* classObj, ObjFn* fn) |
3840 | { |
3841 | int ip = 0; |
3842 | for (;;) |
3843 | { |
3844 | Code instruction = (Code)fn->code.data[ip]; |
3845 | switch (instruction) |
3846 | { |
3847 | case CODE_LOAD_FIELD: |
3848 | case CODE_STORE_FIELD: |
3849 | case CODE_LOAD_FIELD_THIS: |
3850 | case CODE_STORE_FIELD_THIS: |
3851 | // Shift this class's fields down past the inherited ones. We don't |
3852 | // check for overflow here because we'll see if the number of fields |
3853 | // overflows when the subclass is created. |
3854 | fn->code.data[ip + 1] += classObj->superclass->numFields; |
3855 | break; |
3856 | |
3857 | case CODE_SUPER_0: |
3858 | case CODE_SUPER_1: |
3859 | case CODE_SUPER_2: |
3860 | case CODE_SUPER_3: |
3861 | case CODE_SUPER_4: |
3862 | case CODE_SUPER_5: |
3863 | case CODE_SUPER_6: |
3864 | case CODE_SUPER_7: |
3865 | case CODE_SUPER_8: |
3866 | case CODE_SUPER_9: |
3867 | case CODE_SUPER_10: |
3868 | case CODE_SUPER_11: |
3869 | case CODE_SUPER_12: |
3870 | case CODE_SUPER_13: |
3871 | case CODE_SUPER_14: |
3872 | case CODE_SUPER_15: |
3873 | case CODE_SUPER_16: |
3874 | { |
3875 | // Fill in the constant slot with a reference to the superclass. |
3876 | int constant = (fn->code.data[ip + 3] << 8) | fn->code.data[ip + 4]; |
3877 | fn->constants.data[constant] = OBJ_VAL(classObj->superclass); |
3878 | break; |
3879 | } |
3880 | |
3881 | case CODE_CLOSURE: |
3882 | { |
3883 | // Bind the nested closure too. |
3884 | int constant = (fn->code.data[ip + 1] << 8) | fn->code.data[ip + 2]; |
3885 | wrenBindMethodCode(classObj, AS_FN(fn->constants.data[constant])); |
3886 | break; |
3887 | } |
3888 | |
3889 | case CODE_END: |
3890 | return; |
3891 | |
3892 | default: |
3893 | // Other instructions are unaffected, so just skip over them. |
3894 | break; |
3895 | } |
3896 | ip += 1 + getByteCountForArguments(fn->code.data, fn->constants.data, ip); |
3897 | } |
3898 | } |
3899 | |
3900 | void wrenMarkCompiler(WrenVM* vm, Compiler* compiler) |
3901 | { |
3902 | wrenGrayValue(vm, compiler->parser->current.value); |
3903 | wrenGrayValue(vm, compiler->parser->previous.value); |
3904 | wrenGrayValue(vm, compiler->parser->next.value); |
3905 | |
3906 | // Walk up the parent chain to mark the outer compilers too. The VM only |
3907 | // tracks the innermost one. |
3908 | do |
3909 | { |
3910 | wrenGrayObj(vm, (Obj*)compiler->fn); |
3911 | wrenGrayObj(vm, (Obj*)compiler->constants); |
3912 | wrenGrayObj(vm, (Obj*)compiler->attributes); |
3913 | |
3914 | if (compiler->enclosingClass != NULL) |
3915 | { |
3916 | wrenBlackenSymbolTable(vm, &compiler->enclosingClass->fields); |
3917 | |
3918 | if(compiler->enclosingClass->methodAttributes != NULL) |
3919 | { |
3920 | wrenGrayObj(vm, (Obj*)compiler->enclosingClass->methodAttributes); |
3921 | } |
3922 | if(compiler->enclosingClass->classAttributes != NULL) |
3923 | { |
3924 | wrenGrayObj(vm, (Obj*)compiler->enclosingClass->classAttributes); |
3925 | } |
3926 | } |
3927 | |
3928 | compiler = compiler->parent; |
3929 | } |
3930 | while (compiler != NULL); |
3931 | } |
3932 | |
3933 | // Helpers for Attributes |
3934 | |
3935 | // Throw an error if any attributes were found preceding, |
3936 | // and clear the attributes so the error doesn't keep happening. |
3937 | static void disallowAttributes(Compiler* compiler) |
3938 | { |
3939 | if (compiler->numAttributes > 0) |
3940 | { |
3941 | error(compiler, "Attributes can only specified before a class or a method" ); |
3942 | wrenMapClear(compiler->parser->vm, compiler->attributes); |
3943 | compiler->numAttributes = 0; |
3944 | } |
3945 | } |
3946 | |
3947 | // Add an attribute to a given group in the compiler attribues map |
3948 | static void addToAttributeGroup(Compiler* compiler, |
3949 | Value group, Value key, Value value) |
3950 | { |
3951 | WrenVM* vm = compiler->parser->vm; |
3952 | |
3953 | if(IS_OBJ(group)) wrenPushRoot(vm, AS_OBJ(group)); |
3954 | if(IS_OBJ(key)) wrenPushRoot(vm, AS_OBJ(key)); |
3955 | if(IS_OBJ(value)) wrenPushRoot(vm, AS_OBJ(value)); |
3956 | |
3957 | Value groupMapValue = wrenMapGet(compiler->attributes, group); |
3958 | if(IS_UNDEFINED(groupMapValue)) |
3959 | { |
3960 | groupMapValue = OBJ_VAL(wrenNewMap(vm)); |
3961 | wrenMapSet(vm, compiler->attributes, group, groupMapValue); |
3962 | } |
3963 | |
3964 | //we store them as a map per so we can maintain duplicate keys |
3965 | //group = { key:[value, ...], } |
3966 | ObjMap* groupMap = AS_MAP(groupMapValue); |
3967 | |
3968 | //var keyItems = group[key] |
3969 | //if(!keyItems) keyItems = group[key] = [] |
3970 | Value keyItemsValue = wrenMapGet(groupMap, key); |
3971 | if(IS_UNDEFINED(keyItemsValue)) |
3972 | { |
3973 | keyItemsValue = OBJ_VAL(wrenNewList(vm, 0)); |
3974 | wrenMapSet(vm, groupMap, key, keyItemsValue); |
3975 | } |
3976 | |
3977 | //keyItems.add(value) |
3978 | ObjList* keyItems = AS_LIST(keyItemsValue); |
3979 | wrenValueBufferWrite(vm, &keyItems->elements, value); |
3980 | |
3981 | if(IS_OBJ(group)) wrenPopRoot(vm); |
3982 | if(IS_OBJ(key)) wrenPopRoot(vm); |
3983 | if(IS_OBJ(value)) wrenPopRoot(vm); |
3984 | } |
3985 | |
3986 | |
3987 | // Emit the attributes in the give map onto the stack |
3988 | static void emitAttributes(Compiler* compiler, ObjMap* attributes) |
3989 | { |
3990 | // Instantiate a new map for the attributes |
3991 | loadCoreVariable(compiler, "Map" ); |
3992 | callMethod(compiler, 0, "new()" , 5); |
3993 | |
3994 | // The attributes are stored as group = { key:[value, value, ...] } |
3995 | // so our first level is the group map |
3996 | for(uint32_t groupIdx = 0; groupIdx < attributes->capacity; groupIdx++) |
3997 | { |
3998 | const MapEntry* groupEntry = &attributes->entries[groupIdx]; |
3999 | if(IS_UNDEFINED(groupEntry->key)) continue; |
4000 | //group key |
4001 | emitConstant(compiler, groupEntry->key); |
4002 | |
4003 | //group value is gonna be a map |
4004 | loadCoreVariable(compiler, "Map" ); |
4005 | callMethod(compiler, 0, "new()" , 5); |
4006 | |
4007 | ObjMap* groupItems = AS_MAP(groupEntry->value); |
4008 | for(uint32_t itemIdx = 0; itemIdx < groupItems->capacity; itemIdx++) |
4009 | { |
4010 | const MapEntry* itemEntry = &groupItems->entries[itemIdx]; |
4011 | if(IS_UNDEFINED(itemEntry->key)) continue; |
4012 | |
4013 | emitConstant(compiler, itemEntry->key); |
4014 | // Attribute key value, key = [] |
4015 | loadCoreVariable(compiler, "List" ); |
4016 | callMethod(compiler, 0, "new()" , 5); |
4017 | // Add the items to the key list |
4018 | ObjList* items = AS_LIST(itemEntry->value); |
4019 | for(int itemIdx = 0; itemIdx < items->elements.count; ++itemIdx) |
4020 | { |
4021 | emitConstant(compiler, items->elements.data[itemIdx]); |
4022 | callMethod(compiler, 1, "addCore_(_)" , 11); |
4023 | } |
4024 | // Add the list to the map |
4025 | callMethod(compiler, 2, "addCore_(_,_)" , 13); |
4026 | } |
4027 | |
4028 | // Add the key/value to the map |
4029 | callMethod(compiler, 2, "addCore_(_,_)" , 13); |
4030 | } |
4031 | |
4032 | } |
4033 | |
4034 | // Methods are stored as method <-> attributes, so we have to have |
4035 | // an indirection to resolve for methods |
4036 | static void emitAttributeMethods(Compiler* compiler, ObjMap* attributes) |
4037 | { |
4038 | // Instantiate a new map for the attributes |
4039 | loadCoreVariable(compiler, "Map" ); |
4040 | callMethod(compiler, 0, "new()" , 5); |
4041 | |
4042 | for(uint32_t methodIdx = 0; methodIdx < attributes->capacity; methodIdx++) |
4043 | { |
4044 | const MapEntry* methodEntry = &attributes->entries[methodIdx]; |
4045 | if(IS_UNDEFINED(methodEntry->key)) continue; |
4046 | emitConstant(compiler, methodEntry->key); |
4047 | ObjMap* attributeMap = AS_MAP(methodEntry->value); |
4048 | emitAttributes(compiler, attributeMap); |
4049 | callMethod(compiler, 2, "addCore_(_,_)" , 13); |
4050 | } |
4051 | } |
4052 | |
4053 | |
4054 | // Emit the final ClassAttributes that exists at runtime |
4055 | static void emitClassAttributes(Compiler* compiler, ClassInfo* classInfo) |
4056 | { |
4057 | loadCoreVariable(compiler, "ClassAttributes" ); |
4058 | |
4059 | classInfo->classAttributes |
4060 | ? emitAttributes(compiler, classInfo->classAttributes) |
4061 | : null(compiler, false); |
4062 | |
4063 | classInfo->methodAttributes |
4064 | ? emitAttributeMethods(compiler, classInfo->methodAttributes) |
4065 | : null(compiler, false); |
4066 | |
4067 | callMethod(compiler, 2, "new(_,_)" , 8); |
4068 | } |
4069 | |
4070 | // Copy the current attributes stored in the compiler into a destination map |
4071 | // This also resets the counter, since the intent is to consume the attributes |
4072 | static void copyAttributes(Compiler* compiler, ObjMap* into) |
4073 | { |
4074 | compiler->numAttributes = 0; |
4075 | |
4076 | if(compiler->attributes->count == 0) return; |
4077 | if(into == NULL) return; |
4078 | |
4079 | WrenVM* vm = compiler->parser->vm; |
4080 | |
4081 | // Note we copy the actual values as is since we'll take ownership |
4082 | // and clear the original map |
4083 | for(uint32_t attrIdx = 0; attrIdx < compiler->attributes->capacity; attrIdx++) |
4084 | { |
4085 | const MapEntry* attrEntry = &compiler->attributes->entries[attrIdx]; |
4086 | if(IS_UNDEFINED(attrEntry->key)) continue; |
4087 | wrenMapSet(vm, into, attrEntry->key, attrEntry->value); |
4088 | } |
4089 | |
4090 | wrenMapClear(vm, compiler->attributes); |
4091 | } |
4092 | |
4093 | // Copy the current attributes stored in the compiler into the method specific |
4094 | // attributes for the current enclosingClass. |
4095 | // This also resets the counter, since the intent is to consume the attributes |
4096 | static void copyMethodAttributes(Compiler* compiler, bool isForeign, |
4097 | bool isStatic, const char* fullSignature, int32_t length) |
4098 | { |
4099 | compiler->numAttributes = 0; |
4100 | |
4101 | if(compiler->attributes->count == 0) return; |
4102 | |
4103 | WrenVM* vm = compiler->parser->vm; |
4104 | |
4105 | // Make a map for this method to copy into |
4106 | ObjMap* methodAttr = wrenNewMap(vm); |
4107 | wrenPushRoot(vm, (Obj*)methodAttr); |
4108 | copyAttributes(compiler, methodAttr); |
4109 | |
4110 | // Include 'foreign static ' in front as needed |
4111 | int32_t fullLength = length; |
4112 | if(isForeign) fullLength += 8; |
4113 | if(isStatic) fullLength += 7; |
4114 | char fullSignatureWithPrefix[MAX_METHOD_SIGNATURE + 8 + 7]; |
4115 | const char* foreignPrefix = isForeign ? "foreign " : "" ; |
4116 | const char* staticPrefix = isStatic ? "static " : "" ; |
4117 | sprintf(fullSignatureWithPrefix, "%s%s%.*s" , foreignPrefix, staticPrefix, |
4118 | length, fullSignature); |
4119 | fullSignatureWithPrefix[fullLength] = '\0'; |
4120 | |
4121 | if(compiler->enclosingClass->methodAttributes == NULL) { |
4122 | compiler->enclosingClass->methodAttributes = wrenNewMap(vm); |
4123 | } |
4124 | |
4125 | // Store the method attributes in the class map |
4126 | Value key = wrenNewStringLength(vm, fullSignatureWithPrefix, fullLength); |
4127 | wrenMapSet(vm, compiler->enclosingClass->methodAttributes, key, OBJ_VAL(methodAttr)); |
4128 | |
4129 | wrenPopRoot(vm); |
4130 | } |
4131 | |