]>
Commit | Line | Data |
---|---|---|
2f1a1aea | 1 | // recursive descent parser for systemtap scripts |
213bee8f | 2 | // Copyright (C) 2005-2006 Red Hat Inc. |
77a5c1f9 | 3 | // Copyright (C) 2006 Intel Corporation. |
69c68955 FCE |
4 | // |
5 | // This file is part of systemtap, and is free software. You can | |
6 | // redistribute it and/or modify it under the terms of the GNU General | |
7 | // Public License (GPL); either version 2, or (at your option) any | |
8 | // later version. | |
2f1a1aea | 9 | |
2b066ec1 | 10 | #include "config.h" |
2f1a1aea FCE |
11 | #include "staptree.h" |
12 | #include "parse.h" | |
177a8ead | 13 | #include "session.h" |
2b066ec1 FCE |
14 | #include <iostream> |
15 | #include <fstream> | |
2f1a1aea | 16 | #include <cctype> |
9c0c0e46 | 17 | #include <cstdlib> |
29e64872 | 18 | #include <cassert> |
9c0c0e46 FCE |
19 | #include <cerrno> |
20 | #include <climits> | |
57b73400 | 21 | #include <sstream> |
2f1a1aea FCE |
22 | |
23 | using namespace std; | |
24 | ||
25 | // ------------------------------------------------------------------------ | |
26 | ||
bb2e3076 FCE |
27 | |
28 | ||
177a8ead FCE |
29 | parser::parser (systemtap_session& s, istream& i, bool p): |
30 | session (s), | |
24cb178f | 31 | input_name ("<input>"), free_input (0), |
213bee8f | 32 | input (i, input_name, s), privileged (p), |
6e213f58 | 33 | context(con_unknown), last_t (0), next_t (0), num_errors (0) |
2f1a1aea FCE |
34 | { } |
35 | ||
177a8ead FCE |
36 | parser::parser (systemtap_session& s, const string& fn, bool p): |
37 | session (s), | |
2f1a1aea | 38 | input_name (fn), free_input (new ifstream (input_name.c_str(), ios::in)), |
213bee8f | 39 | input (* free_input, input_name, s), privileged (p), |
6e213f58 | 40 | context(con_unknown), last_t (0), next_t (0), num_errors (0) |
2f1a1aea FCE |
41 | { } |
42 | ||
43 | parser::~parser() | |
44 | { | |
45 | if (free_input) delete free_input; | |
46 | } | |
47 | ||
48 | ||
82919855 | 49 | stapfile* |
177a8ead | 50 | parser::parse (systemtap_session& s, std::istream& i, bool pr) |
82919855 | 51 | { |
177a8ead | 52 | parser p (s, i, pr); |
82919855 FCE |
53 | return p.parse (); |
54 | } | |
55 | ||
56 | ||
57 | stapfile* | |
177a8ead | 58 | parser::parse (systemtap_session& s, const std::string& n, bool pr) |
82919855 | 59 | { |
177a8ead | 60 | parser p (s, n, pr); |
82919855 FCE |
61 | return p.parse (); |
62 | } | |
63 | ||
d7f3e0c5 GH |
64 | static string |
65 | tt2str(token_type tt) | |
66 | { | |
67 | switch (tt) | |
68 | { | |
69 | case tok_junk: return "junk"; | |
70 | case tok_identifier: return "identifier"; | |
71 | case tok_operator: return "operator"; | |
72 | case tok_string: return "string"; | |
73 | case tok_number: return "number"; | |
74 | case tok_embedded: return "embedded-code"; | |
6e213f58 | 75 | case tok_keyword: return "keyword"; |
d7f3e0c5 GH |
76 | } |
77 | return "unknown token"; | |
78 | } | |
82919855 | 79 | |
0323ed4d WC |
80 | ostream& |
81 | operator << (ostream& o, const source_loc& loc) | |
82 | { | |
83 | o << loc.file << ":" | |
84 | << loc.line << ":" | |
85 | << loc.column; | |
86 | ||
87 | return o; | |
88 | } | |
89 | ||
56099f08 FCE |
90 | ostream& |
91 | operator << (ostream& o, const token& t) | |
92 | { | |
d7f3e0c5 | 93 | o << tt2str(t.type); |
56099f08 | 94 | |
6e213f58 | 95 | if (t.type != tok_embedded && t.type != tok_keyword) // XXX: other types? |
56099f08 | 96 | { |
24cb178f FCE |
97 | o << " '"; |
98 | for (unsigned i=0; i<t.content.length(); i++) | |
99 | { | |
100 | char c = t.content[i]; | |
101 | o << (isprint (c) ? c : '?'); | |
102 | } | |
103 | o << "'"; | |
56099f08 | 104 | } |
56099f08 FCE |
105 | |
106 | o << " at " | |
0323ed4d | 107 | << t.location; |
56099f08 FCE |
108 | |
109 | return o; | |
110 | } | |
111 | ||
112 | ||
2f1a1aea FCE |
113 | void |
114 | parser::print_error (const parse_error &pe) | |
115 | { | |
116 | cerr << "parse error: " << pe.what () << endl; | |
117 | ||
177a8ead FCE |
118 | if (pe.tok) |
119 | { | |
120 | cerr << "\tat: " << *pe.tok << endl; | |
121 | } | |
2f1a1aea | 122 | else |
177a8ead FCE |
123 | { |
124 | const token* t = last_t; | |
125 | if (t) | |
126 | cerr << "\tsaw: " << *t << endl; | |
127 | else | |
128 | cerr << "\tsaw: " << input_name << " EOF" << endl; | |
129 | } | |
2f1a1aea FCE |
130 | |
131 | // XXX: make it possible to print the last input line, | |
132 | // so as to line up an arrow with the specific error column | |
133 | ||
134 | num_errors ++; | |
135 | } | |
136 | ||
137 | ||
138 | const token* | |
139 | parser::last () | |
140 | { | |
141 | return last_t; | |
142 | } | |
143 | ||
144 | ||
177a8ead FCE |
145 | // Here, we perform on-the-fly preprocessing. |
146 | // The basic form is %( CONDITION %? THEN-TOKENS %: ELSE-TOKENS %) | |
44ce8ed5 FCE |
147 | // where CONDITION is: kernel_v[r] COMPARISON-OP "version-string" |
148 | // or: arch COMPARISON-OP "arch-string" | |
149 | // The %: ELSE-TOKENS part is optional. | |
177a8ead FCE |
150 | // |
151 | // e.g. %( kernel_v > "2.5" %? "foo" %: "baz" %) | |
44ce8ed5 | 152 | // e.g. %( arch != "i686" %? "foo" %: "baz" %) |
177a8ead FCE |
153 | // |
154 | // Up to an entire %( ... %) expression is processed by a single call | |
155 | // to this function. Tokens included by any nested conditions are | |
156 | // enqueued in a private vector. | |
157 | ||
158 | bool eval_pp_conditional (systemtap_session& s, | |
159 | const token* l, const token* op, const token* r) | |
160 | { | |
44ce8ed5 FCE |
161 | if (l->type == tok_identifier && (l->content == "kernel_v" || |
162 | l->content == "kernel_vr")) | |
163 | { | |
164 | string target_kernel_vr = s.kernel_release; | |
165 | string target_kernel_v = target_kernel_vr; | |
166 | // cut off any release code suffix | |
77a5c1f9 | 167 | string::size_type dr = target_kernel_vr.find ('-'); |
44ce8ed5 FCE |
168 | if (dr > 0 && dr != string::npos) |
169 | target_kernel_v = target_kernel_vr.substr (0, dr); | |
170 | ||
171 | if (! (r->type == tok_string)) | |
172 | throw parse_error ("expected string literal", r); | |
173 | string query_kernel_vr = r->content; | |
174 | ||
175 | // collect acceptable strverscmp results. | |
176 | int rvc_ok1, rvc_ok2; | |
177 | if (op->type == tok_operator && op->content == "<=") | |
178 | { rvc_ok1 = -1; rvc_ok2 = 0; } | |
179 | else if (op->type == tok_operator && op->content == ">=") | |
180 | { rvc_ok1 = 1; rvc_ok2 = 0; } | |
181 | else if (op->type == tok_operator && op->content == "<") | |
182 | { rvc_ok1 = -1; rvc_ok2 = -1; } | |
183 | else if (op->type == tok_operator && op->content == ">") | |
184 | { rvc_ok1 = 1; rvc_ok2 = 1; } | |
185 | else if (op->type == tok_operator && op->content == "==") | |
186 | { rvc_ok1 = 0; rvc_ok2 = 0; } | |
187 | else if (op->type == tok_operator && op->content == "!=") | |
188 | { rvc_ok1 = -1; rvc_ok2 = 1; } | |
189 | else | |
190 | throw parse_error ("expected comparison operator", op); | |
191 | ||
192 | int rvc_result = strverscmp ((l->content == "kernel_vr" ? | |
193 | target_kernel_vr.c_str() : | |
194 | target_kernel_v.c_str()), | |
195 | query_kernel_vr.c_str()); | |
196 | // normalize rvc_result | |
197 | if (rvc_result < 0) rvc_result = -1; | |
198 | if (rvc_result > 0) rvc_result = 1; | |
199 | ||
200 | return (rvc_result == rvc_ok1 || rvc_result == rvc_ok2); | |
201 | } | |
202 | else if (l->type == tok_identifier && l->content == "arch") | |
203 | { | |
204 | string target_architecture = s.architecture; | |
205 | if (! (r->type == tok_string)) | |
206 | throw parse_error ("expected string literal", r); | |
207 | string query_architecture = r->content; | |
208 | ||
209 | bool result; | |
210 | if (op->type == tok_operator && op->content == "==") | |
211 | result = target_architecture == query_architecture; | |
212 | else if (op->type == tok_operator && op->content == "!=") | |
213 | result = target_architecture != query_architecture; | |
214 | else | |
215 | throw parse_error ("expected '==' or '!='", op); | |
216 | ||
217 | return result; | |
218 | } | |
219 | // XXX: support other forms? "CONFIG_SMP" ? | |
177a8ead | 220 | else |
44ce8ed5 | 221 | throw parse_error ("expected 'arch' or 'kernel_v' or 'kernel_vr'", l); |
177a8ead FCE |
222 | } |
223 | ||
224 | ||
225 | const token* | |
226 | parser::scan_pp () | |
227 | { | |
228 | while (true) | |
229 | { | |
230 | if (enqueued_pp.size() > 0) | |
231 | { | |
232 | const token* t = enqueued_pp[0]; | |
233 | enqueued_pp.erase (enqueued_pp.begin()); | |
234 | return t; | |
235 | } | |
236 | ||
237 | const token* t = input.scan (); // NB: not recursive! | |
238 | if (t == 0) // EOF | |
239 | return t; | |
240 | ||
241 | if (! (t->type == tok_operator && t->content == "%(")) // ordinary token | |
242 | return t; | |
243 | ||
244 | // We have a %( - it's time to throw a preprocessing party! | |
245 | ||
246 | const token *l, *op, *r; | |
247 | l = input.scan (); // NB: not recursive, though perhaps could be | |
248 | op = input.scan (); | |
249 | r = input.scan (); | |
250 | if (l == 0 || op == 0 || r == 0) | |
251 | throw parse_error ("incomplete condition after '%('", t); | |
252 | // NB: consider generalizing to consume all tokens until %?, and | |
253 | // passing that as a vector to an evaluator. | |
254 | ||
255 | bool result = eval_pp_conditional (session, l, op, r); | |
70c743d8 JS |
256 | delete l; |
257 | delete op; | |
258 | delete r; | |
177a8ead FCE |
259 | |
260 | const token *m = input.scan (); // NB: not recursive | |
261 | if (! (m && m->type == tok_operator && m->content == "%?")) | |
262 | throw parse_error ("expected '%?' marker for conditional", t); | |
70c743d8 | 263 | delete m; // "%?" |
177a8ead FCE |
264 | |
265 | vector<const token*> my_enqueued_pp; | |
266 | ||
267 | while (true) // consume THEN tokens | |
268 | { | |
269 | m = scan_pp (); // NB: recursive | |
270 | if (m == 0) | |
271 | throw parse_error ("missing THEN tokens for conditional", t); | |
272 | ||
273 | if (m->type == tok_operator && (m->content == "%:" || // ELSE | |
274 | m->content == "%)")) // END | |
275 | break; | |
276 | // enqueue token | |
277 | if (result) | |
278 | my_enqueued_pp.push_back (m); | |
70c743d8 JS |
279 | else |
280 | delete m; // unused token | |
177a8ead FCE |
281 | // continue |
282 | } | |
283 | ||
284 | if (m && m->type == tok_operator && m->content == "%:") // ELSE | |
70c743d8 JS |
285 | { |
286 | delete m; // "%:" | |
287 | while (true) | |
288 | { | |
289 | m = scan_pp (); // NB: recursive | |
290 | if (m == 0) | |
291 | throw parse_error ("missing ELSE tokens for conditional", t); | |
292 | ||
293 | if (m->type == tok_operator && m->content == "%)") // END | |
294 | break; | |
295 | // enqueue token | |
296 | if (! result) | |
297 | my_enqueued_pp.push_back (m); | |
298 | else | |
299 | delete m; // unused token | |
300 | // continue | |
301 | } | |
302 | } | |
303 | delete t; // "%(" | |
304 | delete m; // "%)" | |
177a8ead FCE |
305 | |
306 | // NB: we transcribe the retained tokens here, and not inside | |
307 | // the THEN/ELSE while loops. If it were done there, each loop | |
308 | // would become infinite (each iteration consuming an ordinary | |
309 | // token the previous one just pushed there). Guess how I | |
310 | // figured that out. | |
311 | enqueued_pp.insert (enqueued_pp.end(), | |
312 | my_enqueued_pp.begin(), | |
313 | my_enqueued_pp.end()); | |
314 | ||
315 | // Go back to outermost while(true) loop. We hope that at least | |
316 | // some THEN or ELSE tokens were enqueued. If not, around we go | |
317 | // again, until EOF. | |
318 | } | |
319 | } | |
320 | ||
321 | ||
2f1a1aea FCE |
322 | const token* |
323 | parser::next () | |
324 | { | |
325 | if (! next_t) | |
177a8ead | 326 | next_t = scan_pp (); |
2f1a1aea FCE |
327 | if (! next_t) |
328 | throw parse_error ("unexpected end-of-file"); | |
329 | ||
2f1a1aea FCE |
330 | last_t = next_t; |
331 | // advance by zeroing next_t | |
332 | next_t = 0; | |
333 | return last_t; | |
334 | } | |
335 | ||
336 | ||
337 | const token* | |
338 | parser::peek () | |
339 | { | |
340 | if (! next_t) | |
177a8ead | 341 | next_t = scan_pp (); |
2f1a1aea FCE |
342 | |
343 | // don't advance by zeroing next_t | |
344 | last_t = next_t; | |
345 | return next_t; | |
346 | } | |
347 | ||
348 | ||
d7f3e0c5 GH |
349 | static inline bool |
350 | tok_is(token const * t, token_type tt, string const & expected) | |
351 | { | |
352 | return t && t->type == tt && t->content == expected; | |
353 | } | |
354 | ||
355 | ||
356 | const token* | |
357 | parser::expect_known (token_type tt, string const & expected) | |
358 | { | |
359 | const token *t = next(); | |
57b73400 | 360 | if (! (t && t->type == tt && t->content == expected)) |
d7f3e0c5 GH |
361 | throw parse_error ("expected '" + expected + "'"); |
362 | return t; | |
363 | } | |
364 | ||
365 | ||
366 | const token* | |
367 | parser::expect_unknown (token_type tt, string & target) | |
368 | { | |
369 | const token *t = next(); | |
370 | if (!(t && t->type == tt)) | |
371 | throw parse_error ("expected " + tt2str(tt)); | |
372 | target = t->content; | |
373 | return t; | |
374 | } | |
375 | ||
376 | ||
377 | const token* | |
378 | parser::expect_op (std::string const & expected) | |
379 | { | |
380 | return expect_known (tok_operator, expected); | |
381 | } | |
382 | ||
383 | ||
384 | const token* | |
385 | parser::expect_kw (std::string const & expected) | |
386 | { | |
387 | return expect_known (tok_identifier, expected); | |
388 | } | |
389 | ||
57b73400 GH |
390 | const token* |
391 | parser::expect_number (int64_t & expected) | |
392 | { | |
393 | std::string tmp; | |
394 | token const * tt = expect_unknown (tok_number, tmp); | |
395 | istringstream iss(tmp); | |
396 | iss >> expected; | |
397 | return tt; | |
398 | } | |
399 | ||
d7f3e0c5 GH |
400 | |
401 | const token* | |
402 | parser::expect_ident (std::string & target) | |
403 | { | |
404 | return expect_unknown (tok_identifier, target); | |
405 | } | |
406 | ||
407 | ||
408 | bool | |
409 | parser::peek_op (std::string const & op) | |
410 | { | |
411 | return tok_is (peek(), tok_operator, op); | |
412 | } | |
413 | ||
414 | ||
415 | bool | |
416 | parser::peek_kw (std::string const & kw) | |
417 | { | |
418 | return tok_is (peek(), tok_identifier, kw); | |
419 | } | |
420 | ||
421 | ||
422 | ||
213bee8f FCE |
423 | lexer::lexer (istream& i, const string& in, systemtap_session& s): |
424 | input (i), input_name (in), cursor_line (1), cursor_column (1), session(s) | |
2f1a1aea FCE |
425 | { } |
426 | ||
bb2e3076 FCE |
427 | |
428 | int | |
429 | lexer::input_peek (unsigned n) | |
430 | { | |
431 | while (lookahead.size() <= n) | |
432 | { | |
433 | int c = input.get (); | |
434 | lookahead.push_back (input ? c : -1); | |
435 | } | |
436 | return lookahead[n]; | |
437 | } | |
438 | ||
439 | ||
2f1a1aea FCE |
440 | int |
441 | lexer::input_get () | |
442 | { | |
bb2e3076 FCE |
443 | int c = input_peek (0); |
444 | lookahead.erase (lookahead.begin ()); | |
445 | ||
446 | if (c < 0) return c; // EOF | |
447 | ||
2f1a1aea FCE |
448 | // update source cursor |
449 | if (c == '\n') | |
450 | { | |
451 | cursor_line ++; | |
452 | cursor_column = 1; | |
453 | } | |
454 | else | |
455 | cursor_column ++; | |
456 | ||
457 | return c; | |
458 | } | |
459 | ||
460 | ||
461 | token* | |
462 | lexer::scan () | |
463 | { | |
464 | token* n = new token; | |
465 | n->location.file = input_name; | |
466 | ||
467 | skip: | |
468 | n->location.line = cursor_line; | |
469 | n->location.column = cursor_column; | |
470 | ||
471 | int c = input_get(); | |
472 | if (c < 0) | |
473 | { | |
474 | delete n; | |
475 | return 0; | |
476 | } | |
477 | ||
478 | if (isspace (c)) | |
479 | goto skip; | |
480 | ||
d02548c0 | 481 | else if (isalpha (c) || c == '$' || c == '@' || c == '_') |
2f1a1aea FCE |
482 | { |
483 | n->type = tok_identifier; | |
484 | n->content = (char) c; | |
485 | while (1) | |
486 | { | |
bb2e3076 | 487 | int c2 = input_peek (); |
2f1a1aea FCE |
488 | if (! input) |
489 | break; | |
0fefb486 | 490 | if ((isalnum(c2) || c2 == '_' || c2 == '$')) |
2f1a1aea FCE |
491 | { |
492 | n->content.push_back(c2); | |
493 | input_get (); | |
494 | } | |
495 | else | |
496 | break; | |
497 | } | |
213bee8f FCE |
498 | |
499 | // Expand command line arguments to literals. $1 .. $999 as | |
500 | // numbers and @1 .. @999 as strings. | |
501 | if (n->content[0] == '@' || n->content[0] == '$') | |
502 | { | |
503 | string idxstr = n->content.substr(1); | |
504 | const char* startp = idxstr.c_str(); | |
505 | char *endp; | |
506 | errno = 0; | |
507 | unsigned long idx = strtoul (startp, &endp, 10); | |
508 | if (endp == startp) | |
509 | ; // no numbers at all - leave alone as identifier | |
510 | else | |
511 | { | |
512 | // Use @1/$1 as the base, not @0/$0. Thus the idx-1. | |
513 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' || | |
514 | idx == 0 || idx-1 >= session.args.size ()) | |
515 | throw parse_error ("command line argument index invalid or out of range"); | |
516 | ||
517 | string arg = session.args[idx-1]; | |
518 | n->type = (n->content[0] == '@') ? tok_string : tok_number; | |
519 | n->content = arg; | |
520 | } | |
521 | } | |
6e213f58 DS |
522 | else |
523 | { | |
524 | if (n->content == "probe" | |
525 | || n->content == "global" | |
526 | || n->content == "function" | |
527 | || n->content == "if" | |
528 | || n->content == "else" | |
529 | || n->content == "for" | |
530 | || n->content == "foreach" | |
531 | || n->content == "in" | |
532 | || n->content == "return" | |
533 | || n->content == "delete" | |
534 | || n->content == "while" | |
535 | || n->content == "break" | |
536 | || n->content == "continue" | |
537 | || n->content == "next" | |
538 | || n->content == "string" | |
539 | || n->content == "long") | |
540 | n->type = tok_keyword; | |
541 | } | |
213bee8f | 542 | |
2f1a1aea FCE |
543 | return n; |
544 | } | |
545 | ||
3a20432b | 546 | else if (isdigit (c)) // positive literal |
2f1a1aea | 547 | { |
2f1a1aea | 548 | n->type = tok_number; |
9c0c0e46 FCE |
549 | n->content = (char) c; |
550 | ||
2f1a1aea FCE |
551 | while (1) |
552 | { | |
bb2e3076 | 553 | int c2 = input_peek (); |
2f1a1aea FCE |
554 | if (! input) |
555 | break; | |
9c0c0e46 FCE |
556 | |
557 | // NB: isalnum is very permissive. We rely on strtol, called in | |
558 | // parser::parse_literal below, to confirm that the number string | |
559 | // is correctly formatted and in range. | |
560 | ||
561 | if (isalnum (c2)) | |
2f1a1aea | 562 | { |
9c0c0e46 | 563 | n->content.push_back (c2); |
2f1a1aea FCE |
564 | input_get (); |
565 | } | |
566 | else | |
567 | break; | |
568 | } | |
569 | return n; | |
570 | } | |
571 | ||
572 | else if (c == '\"') | |
573 | { | |
574 | n->type = tok_string; | |
575 | while (1) | |
576 | { | |
577 | c = input_get (); | |
578 | ||
579 | if (! input || c == '\n') | |
580 | { | |
581 | n->type = tok_junk; | |
582 | break; | |
583 | } | |
584 | if (c == '\"') // closing double-quotes | |
585 | break; | |
586 | else if (c == '\\') | |
7d46afb8 GH |
587 | { |
588 | c = input_get (); | |
589 | switch (c) | |
590 | { | |
591 | case 'a': | |
592 | case 'b': | |
593 | case 't': | |
594 | case 'n': | |
595 | case 'v': | |
596 | case 'f': | |
597 | case 'r': | |
f03954fd | 598 | case '0' ... '7': // NB: need only match the first digit |
7d46afb8 GH |
599 | case '\\': |
600 | ||
601 | // Pass these escapes through to the string value | |
f03954fd | 602 | // beign parsed; it will be emitted into a C literal. |
7d46afb8 GH |
603 | |
604 | n->content.push_back('\\'); | |
605 | ||
606 | default: | |
607 | ||
608 | n->content.push_back(c); | |
609 | break; | |
610 | } | |
2f1a1aea FCE |
611 | } |
612 | else | |
613 | n->content.push_back(c); | |
614 | } | |
615 | return n; | |
616 | } | |
617 | ||
618 | else if (ispunct (c)) | |
619 | { | |
bb2e3076 FCE |
620 | int c2 = input_peek (); |
621 | int c3 = input_peek (1); | |
622 | string s1 = string("") + (char) c; | |
623 | string s2 = (c2 > 0 ? s1 + (char) c2 : s1); | |
624 | string s3 = (c3 > 0 ? s2 + (char) c3 : s2); | |
2f1a1aea | 625 | |
3a20432b FCE |
626 | // NB: if we were to recognize negative numeric literals here, |
627 | // we'd introduce another grammar ambiguity: | |
628 | // 1-1 would be parsed as tok_number(1) and tok_number(-1) | |
629 | // instead of tok_number(1) tok_operator('-') tok_number(1) | |
630 | ||
bb2e3076 | 631 | if (s1 == "#") // shell comment |
2f1a1aea FCE |
632 | { |
633 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
634 | do { c = input_get (); } |
635 | while (c >= 0 && cursor_line == this_line); | |
2f1a1aea FCE |
636 | goto skip; |
637 | } | |
bb2e3076 | 638 | else if (s2 == "//") // C++ comment |
63a7c90e FCE |
639 | { |
640 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
641 | do { c = input_get (); } |
642 | while (c >= 0 && cursor_line == this_line); | |
63a7c90e FCE |
643 | goto skip; |
644 | } | |
645 | else if (c == '/' && c2 == '*') // C comment | |
646 | { | |
647 | c2 = input_get (); | |
648 | unsigned chars = 0; | |
bb2e3076 | 649 | while (c2 >= 0) |
63a7c90e FCE |
650 | { |
651 | chars ++; // track this to prevent "/*/" from being accepted | |
652 | c = c2; | |
653 | c2 = input_get (); | |
654 | if (chars > 1 && c == '*' && c2 == '/') | |
bb2e3076 | 655 | break; |
63a7c90e | 656 | } |
bb2e3076 | 657 | goto skip; |
63a7c90e | 658 | } |
54dfabe9 FCE |
659 | else if (c == '%' && c2 == '{') // embedded code |
660 | { | |
661 | n->type = tok_embedded; | |
662 | (void) input_get (); // swallow '{' already in c2 | |
663 | while (true) | |
664 | { | |
665 | c = input_get (); | |
2ba0c474 | 666 | if (c < 0) // EOF |
54dfabe9 FCE |
667 | { |
668 | n->type = tok_junk; | |
669 | break; | |
670 | } | |
671 | if (c == '%') | |
672 | { | |
673 | c2 = input_peek (); | |
674 | if (c2 == '}') | |
675 | { | |
676 | (void) input_get (); // swallow '}' too | |
677 | break; | |
678 | } | |
679 | } | |
680 | n->content += c; | |
681 | } | |
682 | return n; | |
683 | } | |
2f1a1aea | 684 | |
bb2e3076 FCE |
685 | // We're committed to recognizing at least the first character |
686 | // as an operator. | |
2f1a1aea | 687 | n->type = tok_operator; |
2f1a1aea | 688 | |
bb2e3076 FCE |
689 | // match all valid operators, in decreasing size order |
690 | if (s3 == "<<<" || | |
691 | s3 == "<<=" || | |
692 | s3 == ">>=") | |
82919855 | 693 | { |
bb2e3076 FCE |
694 | n->content = s3; |
695 | input_get (); input_get (); // swallow other two characters | |
696 | } | |
697 | else if (s2 == "==" || | |
698 | s2 == "!=" || | |
699 | s2 == "<=" || | |
700 | s2 == ">=" || | |
701 | s2 == "+=" || | |
702 | s2 == "-=" || | |
703 | s2 == "*=" || | |
704 | s2 == "/=" || | |
705 | s2 == "%=" || | |
706 | s2 == "&=" || | |
707 | s2 == "^=" || | |
708 | s2 == "|=" || | |
d5d7c2cc | 709 | s2 == ".=" || |
bb2e3076 FCE |
710 | s2 == "&&" || |
711 | s2 == "||" || | |
712 | s2 == "++" || | |
713 | s2 == "--" || | |
714 | s2 == "->" || | |
715 | s2 == "<<" || | |
177a8ead FCE |
716 | s2 == ">>" || |
717 | // preprocessor tokens | |
718 | s2 == "%(" || | |
719 | s2 == "%?" || | |
720 | s2 == "%:" || | |
721 | s2 == "%)") | |
bb2e3076 FCE |
722 | { |
723 | n->content = s2; | |
724 | input_get (); // swallow other character | |
725 | } | |
726 | else | |
727 | { | |
728 | n->content = s1; | |
82919855 | 729 | } |
2f1a1aea FCE |
730 | |
731 | return n; | |
732 | } | |
733 | ||
734 | else | |
735 | { | |
736 | n->type = tok_junk; | |
737 | n->content = (char) c; | |
738 | return n; | |
739 | } | |
740 | } | |
741 | ||
742 | ||
743 | // ------------------------------------------------------------------------ | |
744 | ||
745 | stapfile* | |
746 | parser::parse () | |
747 | { | |
748 | stapfile* f = new stapfile; | |
749 | f->name = input_name; | |
56099f08 FCE |
750 | |
751 | bool empty = true; | |
752 | ||
2f1a1aea FCE |
753 | while (1) |
754 | { | |
755 | try | |
756 | { | |
757 | const token* t = peek (); | |
56099f08 | 758 | if (! t) // nice clean EOF |
2f1a1aea FCE |
759 | break; |
760 | ||
56099f08 | 761 | empty = false; |
6e213f58 DS |
762 | if (t->type == tok_keyword && t->content == "probe") |
763 | { | |
764 | context = con_probe; | |
765 | parse_probe (f->probes, f->aliases); | |
766 | } | |
767 | else if (t->type == tok_keyword && t->content == "global") | |
768 | { | |
769 | context = con_global; | |
770 | parse_global (f->globals); | |
771 | } | |
772 | else if (t->type == tok_keyword && t->content == "function") | |
773 | { | |
774 | context = con_function; | |
775 | parse_functiondecl (f->functions); | |
776 | } | |
54dfabe9 | 777 | else if (t->type == tok_embedded) |
6e213f58 DS |
778 | { |
779 | context = con_embedded; | |
780 | f->embeds.push_back (parse_embeddedcode ()); | |
781 | } | |
2f1a1aea | 782 | else |
6e213f58 DS |
783 | { |
784 | context = con_unknown; | |
785 | throw parse_error ("expected 'probe', 'global', 'function', or '%{'"); | |
786 | } | |
2f1a1aea FCE |
787 | } |
788 | catch (parse_error& pe) | |
789 | { | |
790 | print_error (pe); | |
177a8ead FCE |
791 | try |
792 | { | |
793 | // Quietly swallow all tokens until the next '}'. | |
794 | while (1) | |
795 | { | |
796 | const token* t = peek (); | |
797 | if (! t) | |
798 | break; | |
799 | next (); | |
800 | if (t->type == tok_operator && t->content == "}") | |
801 | break; | |
802 | } | |
803 | } | |
804 | catch (parse_error& pe2) | |
805 | { | |
806 | // parse error during recovery ... ugh | |
807 | print_error (pe2); | |
808 | } | |
809 | } | |
2f1a1aea FCE |
810 | } |
811 | ||
56099f08 FCE |
812 | if (empty) |
813 | { | |
814 | cerr << "Input file '" << input_name << "' is empty or missing." << endl; | |
815 | delete f; | |
816 | return 0; | |
817 | } | |
818 | else if (num_errors > 0) | |
2f1a1aea FCE |
819 | { |
820 | cerr << num_errors << " parse error(s)." << endl; | |
821 | delete f; | |
56099f08 | 822 | return 0; |
2f1a1aea FCE |
823 | } |
824 | ||
825 | return f; | |
826 | } | |
827 | ||
828 | ||
20c6c071 | 829 | void |
54dfabe9 FCE |
830 | parser::parse_probe (std::vector<probe *> & probe_ret, |
831 | std::vector<probe_alias *> & alias_ret) | |
2f1a1aea | 832 | { |
82919855 | 833 | const token* t0 = next (); |
6e213f58 | 834 | if (! (t0->type == tok_keyword && t0->content == "probe")) |
82919855 FCE |
835 | throw parse_error ("expected 'probe'"); |
836 | ||
20c6c071 GH |
837 | vector<probe_point *> aliases; |
838 | vector<probe_point *> locations; | |
839 | ||
840 | bool equals_ok = true; | |
82919855 | 841 | |
97266278 LG |
842 | int epilogue_alias = 0; |
843 | ||
2f1a1aea FCE |
844 | while (1) |
845 | { | |
b4ceace2 FCE |
846 | probe_point * pp = parse_probe_point (); |
847 | ||
848 | const token* t = peek (); | |
849 | if (equals_ok && t | |
850 | && t->type == tok_operator && t->content == "=") | |
851 | { | |
852 | aliases.push_back(pp); | |
853 | next (); | |
854 | continue; | |
855 | } | |
97266278 LG |
856 | else if (equals_ok && t |
857 | && t->type == tok_operator && t->content == "+=") | |
858 | { | |
859 | aliases.push_back(pp); | |
860 | epilogue_alias = 1; | |
861 | next (); | |
862 | continue; | |
863 | } | |
864 | ||
b4ceace2 FCE |
865 | else if (t && t->type == tok_operator && t->content == ",") |
866 | { | |
867 | locations.push_back(pp); | |
868 | equals_ok = false; | |
869 | next (); | |
870 | continue; | |
871 | } | |
872 | else if (t && t->type == tok_operator && t->content == "{") | |
873 | { | |
874 | locations.push_back(pp); | |
875 | break; | |
876 | } | |
2f1a1aea | 877 | else |
9c0c0e46 | 878 | throw parse_error ("expected probe point specifier"); |
2f1a1aea | 879 | } |
20c6c071 | 880 | |
20c6c071 GH |
881 | if (aliases.empty()) |
882 | { | |
54dfabe9 FCE |
883 | probe* p = new probe; |
884 | p->tok = t0; | |
885 | p->locations = locations; | |
886 | p->body = parse_stmt_block (); | |
887 | probe_ret.push_back (p); | |
20c6c071 GH |
888 | } |
889 | else | |
890 | { | |
54dfabe9 | 891 | probe_alias* p = new probe_alias (aliases); |
97266278 LG |
892 | if(epilogue_alias) |
893 | p->epilogue_style = true; | |
894 | else | |
895 | p->epilogue_style = false; | |
54dfabe9 FCE |
896 | p->tok = t0; |
897 | p->locations = locations; | |
898 | p->body = parse_stmt_block (); | |
899 | alias_ret.push_back (p); | |
20c6c071 | 900 | } |
54dfabe9 | 901 | } |
20c6c071 | 902 | |
54dfabe9 FCE |
903 | |
904 | embeddedcode* | |
905 | parser::parse_embeddedcode () | |
906 | { | |
907 | embeddedcode* e = new embeddedcode; | |
908 | const token* t = next (); | |
909 | if (t->type != tok_embedded) | |
24cb178f FCE |
910 | throw parse_error ("expected '%{'"); |
911 | ||
912 | if (! privileged) | |
913 | throw parse_error ("embedded code in unprivileged script"); | |
54dfabe9 FCE |
914 | |
915 | e->tok = t; | |
916 | e->code = t->content; | |
917 | return e; | |
2f1a1aea FCE |
918 | } |
919 | ||
920 | ||
921 | block* | |
56099f08 | 922 | parser::parse_stmt_block () |
2f1a1aea FCE |
923 | { |
924 | block* pb = new block; | |
925 | ||
56099f08 FCE |
926 | const token* t = next (); |
927 | if (! (t->type == tok_operator && t->content == "{")) | |
928 | throw parse_error ("expected '{'"); | |
929 | ||
930 | pb->tok = t; | |
2b066ec1 | 931 | |
2f1a1aea FCE |
932 | while (1) |
933 | { | |
934 | try | |
935 | { | |
2b066ec1 FCE |
936 | t = peek (); |
937 | if (t && t->type == tok_operator && t->content == "}") | |
938 | { | |
939 | next (); | |
940 | break; | |
941 | } | |
942 | ||
2f1a1aea | 943 | pb->statements.push_back (parse_statement ()); |
2f1a1aea FCE |
944 | } |
945 | catch (parse_error& pe) | |
946 | { | |
947 | print_error (pe); | |
54dfabe9 | 948 | |
2f1a1aea FCE |
949 | // Quietly swallow all tokens until the next ';' or '}'. |
950 | while (1) | |
951 | { | |
952 | const token* t = peek (); | |
54dfabe9 | 953 | if (! t) return 0; |
2f1a1aea | 954 | next (); |
54dfabe9 FCE |
955 | if (t->type == tok_operator |
956 | && (t->content == "}" || t->content == ";")) | |
2f1a1aea FCE |
957 | break; |
958 | } | |
959 | } | |
960 | } | |
961 | ||
962 | return pb; | |
963 | } | |
964 | ||
965 | ||
966 | statement* | |
967 | parser::parse_statement () | |
968 | { | |
969 | const token* t = peek (); | |
970 | if (t && t->type == tok_operator && t->content == ";") | |
971 | { | |
69c68955 FCE |
972 | null_statement* n = new null_statement (); |
973 | n->tok = next (); | |
974 | return n; | |
2f1a1aea FCE |
975 | } |
976 | else if (t && t->type == tok_operator && t->content == "{") | |
56099f08 | 977 | return parse_stmt_block (); |
6e213f58 | 978 | else if (t && t->type == tok_keyword && t->content == "if") |
56099f08 | 979 | return parse_if_statement (); |
6e213f58 | 980 | else if (t && t->type == tok_keyword && t->content == "for") |
69c68955 | 981 | return parse_for_loop (); |
6e213f58 | 982 | else if (t && t->type == tok_keyword && t->content == "foreach") |
69c68955 | 983 | return parse_foreach_loop (); |
6e213f58 | 984 | else if (t && t->type == tok_keyword && t->content == "return") |
56099f08 | 985 | return parse_return_statement (); |
6e213f58 | 986 | else if (t && t->type == tok_keyword && t->content == "delete") |
56099f08 | 987 | return parse_delete_statement (); |
6e213f58 | 988 | else if (t && t->type == tok_keyword && t->content == "while") |
f3c26ea5 | 989 | return parse_while_loop (); |
6e213f58 | 990 | else if (t && t->type == tok_keyword && t->content == "break") |
f3c26ea5 | 991 | return parse_break_statement (); |
6e213f58 | 992 | else if (t && t->type == tok_keyword && t->content == "continue") |
f3c26ea5 | 993 | return parse_continue_statement (); |
6e213f58 | 994 | else if (t && t->type == tok_keyword && t->content == "next") |
f3c26ea5 FCE |
995 | return parse_next_statement (); |
996 | // XXX: "do/while" statement? | |
2f1a1aea FCE |
997 | else if (t && (t->type == tok_operator || // expressions are flexible |
998 | t->type == tok_identifier || | |
999 | t->type == tok_number || | |
1000 | t->type == tok_string)) | |
69c68955 | 1001 | return parse_expr_statement (); |
54dfabe9 | 1002 | // XXX: consider generally accepting tok_embedded here too |
2f1a1aea FCE |
1003 | else |
1004 | throw parse_error ("expected statement"); | |
1005 | } | |
1006 | ||
1007 | ||
56099f08 | 1008 | void |
07c17d67 | 1009 | parser::parse_global (vector <vardecl*>& globals) |
2f1a1aea | 1010 | { |
82919855 | 1011 | const token* t0 = next (); |
6e213f58 | 1012 | if (! (t0->type == tok_keyword && t0->content == "global")) |
82919855 FCE |
1013 | throw parse_error ("expected 'global'"); |
1014 | ||
56099f08 FCE |
1015 | while (1) |
1016 | { | |
1017 | const token* t = next (); | |
1018 | if (! (t->type == tok_identifier)) | |
1019 | throw parse_error ("expected identifier"); | |
1020 | ||
2b066ec1 FCE |
1021 | for (unsigned i=0; i<globals.size(); i++) |
1022 | if (globals[i]->name == t->content) | |
57b73400 GH |
1023 | throw parse_error ("duplicate global name"); |
1024 | ||
24cb178f FCE |
1025 | vardecl* d = new vardecl; |
1026 | d->name = t->content; | |
1027 | d->tok = t; | |
1028 | globals.push_back (d); | |
56099f08 | 1029 | |
82919855 FCE |
1030 | t = peek (); |
1031 | if (t && t->type == tok_operator && t->content == ",") | |
1032 | { | |
1033 | next (); | |
1034 | continue; | |
1035 | } | |
56099f08 | 1036 | else |
82919855 | 1037 | break; |
56099f08 FCE |
1038 | } |
1039 | } | |
1040 | ||
1041 | ||
24cb178f FCE |
1042 | void |
1043 | parser::parse_functiondecl (std::vector<functiondecl*>& functions) | |
56099f08 | 1044 | { |
82919855 | 1045 | const token* t = next (); |
6e213f58 | 1046 | if (! (t->type == tok_keyword && t->content == "function")) |
82919855 FCE |
1047 | throw parse_error ("expected 'function'"); |
1048 | ||
56099f08 | 1049 | |
82919855 | 1050 | t = next (); |
6e213f58 DS |
1051 | if (! (t->type == tok_identifier) |
1052 | && ! (t->type == tok_keyword | |
1053 | && (t->content == "string" || t->content == "long"))) | |
56099f08 | 1054 | throw parse_error ("expected identifier"); |
24cb178f FCE |
1055 | |
1056 | for (unsigned i=0; i<functions.size(); i++) | |
1057 | if (functions[i]->name == t->content) | |
1058 | throw parse_error ("duplicate function name"); | |
1059 | ||
1060 | functiondecl *fd = new functiondecl (); | |
56099f08 FCE |
1061 | fd->name = t->content; |
1062 | fd->tok = t; | |
1063 | ||
1064 | t = next (); | |
6a505121 FCE |
1065 | if (t->type == tok_operator && t->content == ":") |
1066 | { | |
1067 | t = next (); | |
6e213f58 | 1068 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 1069 | fd->type = pe_string; |
6e213f58 | 1070 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 FCE |
1071 | fd->type = pe_long; |
1072 | else throw parse_error ("expected 'string' or 'long'"); | |
1073 | ||
1074 | t = next (); | |
1075 | } | |
1076 | ||
56099f08 FCE |
1077 | if (! (t->type == tok_operator && t->content == "(")) |
1078 | throw parse_error ("expected '('"); | |
1079 | ||
1080 | while (1) | |
1081 | { | |
1082 | t = next (); | |
1083 | ||
1084 | // permit zero-argument fuctions | |
1085 | if (t->type == tok_operator && t->content == ")") | |
1086 | break; | |
1087 | else if (! (t->type == tok_identifier)) | |
1088 | throw parse_error ("expected identifier"); | |
1089 | vardecl* vd = new vardecl; | |
1090 | vd->name = t->content; | |
1091 | vd->tok = t; | |
1092 | fd->formal_args.push_back (vd); | |
1093 | ||
1094 | t = next (); | |
6a505121 FCE |
1095 | if (t->type == tok_operator && t->content == ":") |
1096 | { | |
1097 | t = next (); | |
6e213f58 | 1098 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 1099 | vd->type = pe_string; |
6e213f58 | 1100 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 FCE |
1101 | vd->type = pe_long; |
1102 | else throw parse_error ("expected 'string' or 'long'"); | |
1103 | ||
1104 | t = next (); | |
1105 | } | |
56099f08 FCE |
1106 | if (t->type == tok_operator && t->content == ")") |
1107 | break; | |
1108 | if (t->type == tok_operator && t->content == ",") | |
1109 | continue; | |
1110 | else | |
1111 | throw parse_error ("expected ',' or ')'"); | |
1112 | } | |
1113 | ||
54dfabe9 FCE |
1114 | t = peek (); |
1115 | if (t && t->type == tok_embedded) | |
1116 | fd->body = parse_embeddedcode (); | |
1117 | else | |
1118 | fd->body = parse_stmt_block (); | |
24cb178f FCE |
1119 | |
1120 | functions.push_back (fd); | |
2f1a1aea FCE |
1121 | } |
1122 | ||
1123 | ||
9c0c0e46 FCE |
1124 | probe_point* |
1125 | parser::parse_probe_point () | |
2f1a1aea | 1126 | { |
9c0c0e46 | 1127 | probe_point* pl = new probe_point; |
2f1a1aea | 1128 | |
9c0c0e46 | 1129 | while (1) |
2f1a1aea | 1130 | { |
9c0c0e46 | 1131 | const token* t = next (); |
6e213f58 DS |
1132 | if (! (t->type == tok_identifier |
1133 | // we must allow ".return" and ".function", which are keywords | |
1134 | || t->type == tok_keyword | |
1135 | || (t->type == tok_operator && t->content == "*"))) | |
b4ceace2 | 1136 | throw parse_error ("expected identifier or '*'"); |
9c0c0e46 FCE |
1137 | |
1138 | if (pl->tok == 0) pl->tok = t; | |
1139 | ||
1140 | probe_point::component* c = new probe_point::component; | |
1141 | c->functor = t->content; | |
1142 | pl->components.push_back (c); | |
1143 | // NB though we still may add c->arg soon | |
1144 | ||
a477f3f1 | 1145 | const token* last_t = t; |
9c0c0e46 | 1146 | t = peek (); |
a477f3f1 DS |
1147 | |
1148 | // We need to keep going until we find something other than a | |
1149 | // '*' or identifier, since a probe point wildcard can be | |
1150 | // something like "*a", "*a*", "a*b", "a*b*", etc. | |
1151 | while (t && | |
1152 | // case 1: '*{identifier}' | |
1153 | ((last_t->type == tok_operator && last_t->content == "*" | |
1154 | && (t->type == tok_identifier || t->type == tok_keyword)) | |
1155 | // case 2: '{identifier}*' | |
1156 | || ((last_t->type == tok_identifier | |
1157 | || last_t->type == tok_keyword) | |
1158 | && t->type == tok_operator && t->content == "*"))) | |
1159 | { | |
1160 | c->functor += t->content; | |
1161 | next (); // consume the identifier or '*' | |
1162 | ||
1163 | last_t = t; | |
1164 | t = peek (); | |
1165 | } | |
1166 | ||
9c0c0e46 | 1167 | if (t && t->type == tok_operator |
97266278 LG |
1168 | && (t->content == "{" || t->content == "," || t->content == "=" |
1169 | || t->content == "+=" )) | |
9c0c0e46 FCE |
1170 | break; |
1171 | ||
1172 | if (t && t->type == tok_operator && t->content == "(") | |
1173 | { | |
1174 | next (); // consume "(" | |
1175 | c->arg = parse_literal (); | |
1176 | ||
1177 | t = next (); | |
1178 | if (! (t->type == tok_operator && t->content == ")")) | |
1179 | throw parse_error ("expected ')'"); | |
1180 | ||
1181 | t = peek (); | |
1182 | if (t && t->type == tok_operator | |
20c6c071 | 1183 | && (t->content == "{" || t->content == "," || t->content == "=")) |
9c0c0e46 | 1184 | break; |
2b066ec1 FCE |
1185 | else if (t && t->type == tok_operator && |
1186 | t->content == "(") | |
1187 | throw parse_error ("unexpected '.' or ',' or '{'"); | |
9c0c0e46 FCE |
1188 | } |
1189 | // fall through | |
1190 | ||
1191 | if (t && t->type == tok_operator && t->content == ".") | |
1192 | next (); | |
1193 | else | |
97266278 | 1194 | throw parse_error ("expected '.' or ',' or '(' or '{' or '=' or '+='"); |
2f1a1aea FCE |
1195 | } |
1196 | ||
1197 | return pl; | |
1198 | } | |
1199 | ||
1200 | ||
1201 | literal* | |
1202 | parser::parse_literal () | |
1203 | { | |
1204 | const token* t = next (); | |
56099f08 | 1205 | literal* l; |
2f1a1aea | 1206 | if (t->type == tok_string) |
56099f08 | 1207 | l = new literal_string (t->content); |
2f1a1aea | 1208 | else if (t->type == tok_number) |
9c0c0e46 FCE |
1209 | { |
1210 | const char* startp = t->content.c_str (); | |
1211 | char* endp = (char*) startp; | |
1212 | ||
3a20432b FCE |
1213 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX |
1214 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
1215 | // since the lexer only gives us positive digit strings. | |
9c0c0e46 | 1216 | errno = 0; |
3a20432b | 1217 | long long value = (long long) strtoull (startp, & endp, 0); |
9c0c0e46 | 1218 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' |
3a20432b FCE |
1219 | || (unsigned long long) value > 18446744073709551615ULL |
1220 | || value < -9223372036854775807LL-1) | |
9c0c0e46 FCE |
1221 | throw parse_error ("number invalid or out of range"); |
1222 | ||
3a20432b | 1223 | l = new literal_number (value); |
9c0c0e46 | 1224 | } |
2f1a1aea FCE |
1225 | else |
1226 | throw parse_error ("expected literal string or number"); | |
56099f08 FCE |
1227 | |
1228 | l->tok = t; | |
1229 | return l; | |
2f1a1aea FCE |
1230 | } |
1231 | ||
1232 | ||
1233 | if_statement* | |
1234 | parser::parse_if_statement () | |
1235 | { | |
1236 | const token* t = next (); | |
6e213f58 | 1237 | if (! (t->type == tok_keyword && t->content == "if")) |
56099f08 FCE |
1238 | throw parse_error ("expected 'if'"); |
1239 | if_statement* s = new if_statement; | |
1240 | s->tok = t; | |
1241 | ||
1242 | t = next (); | |
2f1a1aea FCE |
1243 | if (! (t->type == tok_operator && t->content == "(")) |
1244 | throw parse_error ("expected '('"); | |
1245 | ||
2f1a1aea FCE |
1246 | s->condition = parse_expression (); |
1247 | ||
1248 | t = next (); | |
1249 | if (! (t->type == tok_operator && t->content == ")")) | |
1250 | throw parse_error ("expected ')'"); | |
1251 | ||
1252 | s->thenblock = parse_statement (); | |
1253 | ||
1254 | t = peek (); | |
6e213f58 | 1255 | if (t && t->type == tok_keyword && t->content == "else") |
2f1a1aea FCE |
1256 | { |
1257 | next (); | |
1258 | s->elseblock = parse_statement (); | |
1259 | } | |
ed10c639 FCE |
1260 | else |
1261 | s->elseblock = 0; // in case not otherwise initialized | |
2f1a1aea FCE |
1262 | |
1263 | return s; | |
1264 | } | |
1265 | ||
1266 | ||
69c68955 FCE |
1267 | expr_statement* |
1268 | parser::parse_expr_statement () | |
1269 | { | |
1270 | expr_statement *es = new expr_statement; | |
1271 | const token* t = peek (); | |
1272 | es->tok = t; | |
1273 | es->value = parse_expression (); | |
1274 | return es; | |
1275 | } | |
1276 | ||
1277 | ||
56099f08 FCE |
1278 | return_statement* |
1279 | parser::parse_return_statement () | |
1280 | { | |
1281 | const token* t = next (); | |
6e213f58 | 1282 | if (! (t->type == tok_keyword && t->content == "return")) |
56099f08 | 1283 | throw parse_error ("expected 'return'"); |
6e213f58 DS |
1284 | if (context != con_function) |
1285 | throw parse_error ("found 'return' not in function context"); | |
56099f08 FCE |
1286 | return_statement* s = new return_statement; |
1287 | s->tok = t; | |
1288 | s->value = parse_expression (); | |
1289 | return s; | |
1290 | } | |
1291 | ||
1292 | ||
1293 | delete_statement* | |
1294 | parser::parse_delete_statement () | |
1295 | { | |
1296 | const token* t = next (); | |
6e213f58 | 1297 | if (! (t->type == tok_keyword && t->content == "delete")) |
56099f08 FCE |
1298 | throw parse_error ("expected 'delete'"); |
1299 | delete_statement* s = new delete_statement; | |
1300 | s->tok = t; | |
1301 | s->value = parse_expression (); | |
1302 | return s; | |
1303 | } | |
1304 | ||
1305 | ||
f3c26ea5 FCE |
1306 | next_statement* |
1307 | parser::parse_next_statement () | |
1308 | { | |
1309 | const token* t = next (); | |
6e213f58 | 1310 | if (! (t->type == tok_keyword && t->content == "next")) |
f3c26ea5 | 1311 | throw parse_error ("expected 'next'"); |
6e213f58 DS |
1312 | if (context != con_probe) |
1313 | throw parse_error ("found 'next' not in probe context"); | |
f3c26ea5 FCE |
1314 | next_statement* s = new next_statement; |
1315 | s->tok = t; | |
1316 | return s; | |
1317 | } | |
1318 | ||
1319 | ||
1320 | break_statement* | |
1321 | parser::parse_break_statement () | |
1322 | { | |
1323 | const token* t = next (); | |
6e213f58 | 1324 | if (! (t->type == tok_keyword && t->content == "break")) |
f3c26ea5 FCE |
1325 | throw parse_error ("expected 'break'"); |
1326 | break_statement* s = new break_statement; | |
1327 | s->tok = t; | |
1328 | return s; | |
1329 | } | |
1330 | ||
1331 | ||
1332 | continue_statement* | |
1333 | parser::parse_continue_statement () | |
1334 | { | |
1335 | const token* t = next (); | |
6e213f58 | 1336 | if (! (t->type == tok_keyword && t->content == "continue")) |
f3c26ea5 FCE |
1337 | throw parse_error ("expected 'continue'"); |
1338 | continue_statement* s = new continue_statement; | |
1339 | s->tok = t; | |
1340 | return s; | |
1341 | } | |
1342 | ||
1343 | ||
69c68955 FCE |
1344 | for_loop* |
1345 | parser::parse_for_loop () | |
1346 | { | |
f3c26ea5 | 1347 | const token* t = next (); |
6e213f58 | 1348 | if (! (t->type == tok_keyword && t->content == "for")) |
f3c26ea5 FCE |
1349 | throw parse_error ("expected 'for'"); |
1350 | for_loop* s = new for_loop; | |
1351 | s->tok = t; | |
1352 | ||
1353 | t = next (); | |
1354 | if (! (t->type == tok_operator && t->content == "(")) | |
1355 | throw parse_error ("expected '('"); | |
1356 | ||
1357 | // initializer + ";" | |
1358 | t = peek (); | |
1359 | if (t && t->type == tok_operator && t->content == ";") | |
1360 | { | |
cbfbbf69 FCE |
1361 | s->init = 0; |
1362 | next (); | |
f3c26ea5 FCE |
1363 | } |
1364 | else | |
1365 | { | |
1366 | s->init = parse_expr_statement (); | |
1367 | t = next (); | |
1368 | if (! (t->type == tok_operator && t->content == ";")) | |
1369 | throw parse_error ("expected ';'"); | |
1370 | } | |
1371 | ||
1372 | // condition + ";" | |
1373 | t = peek (); | |
1374 | if (t && t->type == tok_operator && t->content == ";") | |
1375 | { | |
1376 | literal_number* l = new literal_number(1); | |
1377 | s->cond = l; | |
1378 | s->cond->tok = next (); | |
1379 | } | |
1380 | else | |
1381 | { | |
1382 | s->cond = parse_expression (); | |
1383 | t = next (); | |
1384 | if (! (t->type == tok_operator && t->content == ";")) | |
1385 | throw parse_error ("expected ';'"); | |
1386 | } | |
1387 | ||
1388 | // increment + ")" | |
1389 | t = peek (); | |
1390 | if (t && t->type == tok_operator && t->content == ")") | |
1391 | { | |
cbfbbf69 FCE |
1392 | s->incr = 0; |
1393 | next (); | |
f3c26ea5 FCE |
1394 | } |
1395 | else | |
1396 | { | |
1397 | s->incr = parse_expr_statement (); | |
1398 | t = next (); | |
1399 | if (! (t->type == tok_operator && t->content == ")")) | |
c958a431 | 1400 | throw parse_error ("expected ')'"); |
f3c26ea5 FCE |
1401 | } |
1402 | ||
1403 | // block | |
1404 | s->block = parse_statement (); | |
1405 | ||
1406 | return s; | |
1407 | } | |
1408 | ||
1409 | ||
1410 | for_loop* | |
1411 | parser::parse_while_loop () | |
1412 | { | |
1413 | const token* t = next (); | |
6e213f58 | 1414 | if (! (t->type == tok_keyword && t->content == "while")) |
f3c26ea5 FCE |
1415 | throw parse_error ("expected 'while'"); |
1416 | for_loop* s = new for_loop; | |
1417 | s->tok = t; | |
1418 | ||
1419 | t = next (); | |
1420 | if (! (t->type == tok_operator && t->content == "(")) | |
1421 | throw parse_error ("expected '('"); | |
1422 | ||
1423 | // dummy init and incr fields | |
cbfbbf69 FCE |
1424 | s->init = 0; |
1425 | s->incr = 0; | |
f3c26ea5 FCE |
1426 | |
1427 | // condition | |
1428 | s->cond = parse_expression (); | |
1429 | ||
f3c26ea5 FCE |
1430 | t = next (); |
1431 | if (! (t->type == tok_operator && t->content == ")")) | |
1432 | throw parse_error ("expected ')'"); | |
1433 | ||
1434 | // block | |
1435 | s->block = parse_statement (); | |
1436 | ||
1437 | return s; | |
69c68955 FCE |
1438 | } |
1439 | ||
1440 | ||
1441 | foreach_loop* | |
1442 | parser::parse_foreach_loop () | |
1443 | { | |
1444 | const token* t = next (); | |
6e213f58 | 1445 | if (! (t->type == tok_keyword && t->content == "foreach")) |
69c68955 FCE |
1446 | throw parse_error ("expected 'foreach'"); |
1447 | foreach_loop* s = new foreach_loop; | |
1448 | s->tok = t; | |
93484556 | 1449 | s->sort_direction = 0; |
69c68955 FCE |
1450 | |
1451 | t = next (); | |
1452 | if (! (t->type == tok_operator && t->content == "(")) | |
1453 | throw parse_error ("expected '('"); | |
1454 | ||
1455 | // see also parse_array_in | |
1456 | ||
1457 | bool parenthesized = false; | |
1458 | t = peek (); | |
1459 | if (t && t->type == tok_operator && t->content == "[") | |
1460 | { | |
1461 | next (); | |
1462 | parenthesized = true; | |
1463 | } | |
1464 | ||
1465 | while (1) | |
1466 | { | |
1467 | t = next (); | |
1468 | if (! (t->type == tok_identifier)) | |
1469 | throw parse_error ("expected identifier"); | |
1470 | symbol* sym = new symbol; | |
1471 | sym->tok = t; | |
1472 | sym->name = t->content; | |
1473 | s->indexes.push_back (sym); | |
1474 | ||
93484556 FCE |
1475 | t = peek (); |
1476 | if (t && t->type == tok_operator && | |
1477 | (t->content == "+" || t->content == "-")) | |
1478 | { | |
1479 | if (s->sort_direction) | |
1480 | throw parse_error ("multiple sort directives"); | |
1481 | s->sort_direction = (t->content == "+") ? 1 : -1; | |
1482 | s->sort_column = s->indexes.size(); | |
1483 | next(); | |
1484 | } | |
1485 | ||
69c68955 FCE |
1486 | if (parenthesized) |
1487 | { | |
93484556 | 1488 | t = peek (); |
69c68955 FCE |
1489 | if (t && t->type == tok_operator && t->content == ",") |
1490 | { | |
1491 | next (); | |
1492 | continue; | |
1493 | } | |
1494 | else if (t && t->type == tok_operator && t->content == "]") | |
1495 | { | |
1496 | next (); | |
1497 | break; | |
1498 | } | |
1499 | else | |
1500 | throw parse_error ("expected ',' or ']'"); | |
1501 | } | |
1502 | else | |
1503 | break; // expecting only one expression | |
1504 | } | |
1505 | ||
1506 | t = next (); | |
6e213f58 | 1507 | if (! (t->type == tok_keyword && t->content == "in")) |
69c68955 | 1508 | throw parse_error ("expected 'in'"); |
d02548c0 GH |
1509 | |
1510 | s->base = parse_indexable(); | |
69c68955 | 1511 | |
93484556 FCE |
1512 | t = peek (); |
1513 | if (t && t->type == tok_operator && | |
1514 | (t->content == "+" || t->content == "-")) | |
1515 | { | |
1516 | if (s->sort_direction) | |
1517 | throw parse_error ("multiple sort directives"); | |
1518 | s->sort_direction = (t->content == "+") ? 1 : -1; | |
1519 | s->sort_column = 0; | |
1520 | next(); | |
1521 | } | |
1522 | ||
69c68955 FCE |
1523 | t = next (); |
1524 | if (! (t->type == tok_operator && t->content == ")")) | |
1525 | throw parse_error ("expected ')'"); | |
1526 | ||
1527 | s->block = parse_statement (); | |
1528 | return s; | |
1529 | } | |
1530 | ||
1531 | ||
2f1a1aea FCE |
1532 | expression* |
1533 | parser::parse_expression () | |
1534 | { | |
1535 | return parse_assignment (); | |
1536 | } | |
1537 | ||
2f1a1aea FCE |
1538 | |
1539 | expression* | |
1540 | parser::parse_assignment () | |
1541 | { | |
1542 | expression* op1 = parse_ternary (); | |
1543 | ||
1544 | const token* t = peek (); | |
82919855 FCE |
1545 | // right-associative operators |
1546 | if (t && t->type == tok_operator | |
2f1a1aea | 1547 | && (t->content == "=" || |
82919855 | 1548 | t->content == "<<<" || |
2f1a1aea | 1549 | t->content == "+=" || |
bb2e3076 FCE |
1550 | t->content == "-=" || |
1551 | t->content == "*=" || | |
1552 | t->content == "/=" || | |
1553 | t->content == "%=" || | |
1554 | t->content == "<<=" || | |
1555 | t->content == ">>=" || | |
1556 | t->content == "&=" || | |
1557 | t->content == "^=" || | |
1558 | t->content == "|=" || | |
d5d7c2cc | 1559 | t->content == ".=" || |
bb2e3076 | 1560 | false)) |
2f1a1aea | 1561 | { |
bb2e3076 | 1562 | // NB: lvalueness is checked during elaboration / translation |
2f1a1aea | 1563 | assignment* e = new assignment; |
56099f08 | 1564 | e->left = op1; |
2f1a1aea | 1565 | e->op = t->content; |
56099f08 | 1566 | e->tok = t; |
2f1a1aea | 1567 | next (); |
82919855 | 1568 | e->right = parse_expression (); |
56099f08 | 1569 | op1 = e; |
2f1a1aea | 1570 | } |
56099f08 FCE |
1571 | |
1572 | return op1; | |
2f1a1aea FCE |
1573 | } |
1574 | ||
1575 | ||
1576 | expression* | |
1577 | parser::parse_ternary () | |
1578 | { | |
1579 | expression* op1 = parse_logical_or (); | |
1580 | ||
1581 | const token* t = peek (); | |
1582 | if (t && t->type == tok_operator && t->content == "?") | |
1583 | { | |
2f1a1aea | 1584 | ternary_expression* e = new ternary_expression; |
56099f08 | 1585 | e->tok = t; |
2f1a1aea | 1586 | e->cond = op1; |
56099f08 FCE |
1587 | next (); |
1588 | e->truevalue = parse_expression (); // XXX | |
2f1a1aea FCE |
1589 | |
1590 | t = next (); | |
1591 | if (! (t->type == tok_operator && t->content == ":")) | |
1592 | throw parse_error ("expected ':'"); | |
1593 | ||
56099f08 | 1594 | e->falsevalue = parse_expression (); // XXX |
2f1a1aea FCE |
1595 | return e; |
1596 | } | |
1597 | else | |
1598 | return op1; | |
1599 | } | |
1600 | ||
1601 | ||
1602 | expression* | |
1603 | parser::parse_logical_or () | |
1604 | { | |
1605 | expression* op1 = parse_logical_and (); | |
1606 | ||
1607 | const token* t = peek (); | |
56099f08 | 1608 | while (t && t->type == tok_operator && t->content == "||") |
2f1a1aea | 1609 | { |
2f1a1aea | 1610 | logical_or_expr* e = new logical_or_expr; |
56099f08 FCE |
1611 | e->tok = t; |
1612 | e->op = t->content; | |
2f1a1aea | 1613 | e->left = op1; |
56099f08 FCE |
1614 | next (); |
1615 | e->right = parse_logical_and (); | |
1616 | op1 = e; | |
1617 | t = peek (); | |
2f1a1aea | 1618 | } |
56099f08 FCE |
1619 | |
1620 | return op1; | |
2f1a1aea FCE |
1621 | } |
1622 | ||
1623 | ||
1624 | expression* | |
1625 | parser::parse_logical_and () | |
1626 | { | |
bb2e3076 | 1627 | expression* op1 = parse_boolean_or (); |
2f1a1aea FCE |
1628 | |
1629 | const token* t = peek (); | |
56099f08 | 1630 | while (t && t->type == tok_operator && t->content == "&&") |
2f1a1aea | 1631 | { |
2f1a1aea FCE |
1632 | logical_and_expr *e = new logical_and_expr; |
1633 | e->left = op1; | |
56099f08 FCE |
1634 | e->op = t->content; |
1635 | e->tok = t; | |
1636 | next (); | |
bb2e3076 FCE |
1637 | e->right = parse_boolean_or (); |
1638 | op1 = e; | |
1639 | t = peek (); | |
1640 | } | |
1641 | ||
1642 | return op1; | |
1643 | } | |
1644 | ||
1645 | ||
1646 | expression* | |
1647 | parser::parse_boolean_or () | |
1648 | { | |
1649 | expression* op1 = parse_boolean_xor (); | |
1650 | ||
1651 | const token* t = peek (); | |
1652 | while (t && t->type == tok_operator && t->content == "|") | |
1653 | { | |
1654 | binary_expression* e = new binary_expression; | |
1655 | e->left = op1; | |
1656 | e->op = t->content; | |
1657 | e->tok = t; | |
1658 | next (); | |
1659 | e->right = parse_boolean_xor (); | |
1660 | op1 = e; | |
1661 | t = peek (); | |
1662 | } | |
1663 | ||
1664 | return op1; | |
1665 | } | |
1666 | ||
1667 | ||
1668 | expression* | |
1669 | parser::parse_boolean_xor () | |
1670 | { | |
1671 | expression* op1 = parse_boolean_and (); | |
1672 | ||
1673 | const token* t = peek (); | |
1674 | while (t && t->type == tok_operator && t->content == "^") | |
1675 | { | |
1676 | binary_expression* e = new binary_expression; | |
1677 | e->left = op1; | |
1678 | e->op = t->content; | |
1679 | e->tok = t; | |
1680 | next (); | |
1681 | e->right = parse_boolean_and (); | |
1682 | op1 = e; | |
1683 | t = peek (); | |
1684 | } | |
1685 | ||
1686 | return op1; | |
1687 | } | |
1688 | ||
1689 | ||
1690 | expression* | |
1691 | parser::parse_boolean_and () | |
1692 | { | |
1693 | expression* op1 = parse_array_in (); | |
1694 | ||
1695 | const token* t = peek (); | |
1696 | while (t && t->type == tok_operator && t->content == "&") | |
1697 | { | |
1698 | binary_expression* e = new binary_expression; | |
1699 | e->left = op1; | |
1700 | e->op = t->content; | |
1701 | e->tok = t; | |
1702 | next (); | |
56099f08 FCE |
1703 | e->right = parse_array_in (); |
1704 | op1 = e; | |
1705 | t = peek (); | |
2f1a1aea | 1706 | } |
56099f08 FCE |
1707 | |
1708 | return op1; | |
2f1a1aea FCE |
1709 | } |
1710 | ||
1711 | ||
1712 | expression* | |
1713 | parser::parse_array_in () | |
1714 | { | |
ce10591c | 1715 | // This is a very tricky case. All these are legit expressions: |
69c68955 | 1716 | // "a in b" "a+0 in b" "[a,b] in c" "[c,(d+0)] in b" |
ce10591c FCE |
1717 | vector<expression*> indexes; |
1718 | bool parenthesized = false; | |
2f1a1aea FCE |
1719 | |
1720 | const token* t = peek (); | |
69c68955 | 1721 | if (t && t->type == tok_operator && t->content == "[") |
ce10591c FCE |
1722 | { |
1723 | next (); | |
1724 | parenthesized = true; | |
1725 | } | |
1726 | ||
1727 | while (1) | |
1728 | { | |
1729 | expression* op1 = parse_comparison (); | |
1730 | indexes.push_back (op1); | |
1731 | ||
1732 | if (parenthesized) | |
1733 | { | |
1734 | const token* t = peek (); | |
1735 | if (t && t->type == tok_operator && t->content == ",") | |
1736 | { | |
1737 | next (); | |
1738 | continue; | |
1739 | } | |
69c68955 | 1740 | else if (t && t->type == tok_operator && t->content == "]") |
ce10591c FCE |
1741 | { |
1742 | next (); | |
1743 | break; | |
1744 | } | |
1745 | else | |
69c68955 | 1746 | throw parse_error ("expected ',' or ']'"); |
ce10591c FCE |
1747 | } |
1748 | else | |
1749 | break; // expecting only one expression | |
1750 | } | |
1751 | ||
1752 | t = peek (); | |
6e213f58 | 1753 | if (t && t->type == tok_keyword && t->content == "in") |
2f1a1aea | 1754 | { |
2f1a1aea | 1755 | array_in *e = new array_in; |
56099f08 | 1756 | e->tok = t; |
ce10591c FCE |
1757 | next (); // swallow "in" |
1758 | ||
1759 | arrayindex* a = new arrayindex; | |
1760 | a->indexes = indexes; | |
d02548c0 GH |
1761 | a->base = parse_indexable(); |
1762 | a->tok = a->base->get_tok(); | |
ce10591c | 1763 | e->operand = a; |
2f1a1aea FCE |
1764 | return e; |
1765 | } | |
ce10591c FCE |
1766 | else if (indexes.size() == 1) // no "in" - need one expression only |
1767 | return indexes[0]; | |
2f1a1aea | 1768 | else |
ce10591c | 1769 | throw parse_error ("unexpected comma-separated expression list"); |
2f1a1aea FCE |
1770 | } |
1771 | ||
1772 | ||
1773 | expression* | |
1774 | parser::parse_comparison () | |
1775 | { | |
bb2e3076 | 1776 | expression* op1 = parse_shift (); |
2f1a1aea FCE |
1777 | |
1778 | const token* t = peek (); | |
56099f08 | 1779 | while (t && t->type == tok_operator |
553d27a5 FCE |
1780 | && (t->content == ">" || |
1781 | t->content == "<" || | |
1782 | t->content == "==" || | |
1783 | t->content == "!=" || | |
1784 | t->content == "<=" || | |
bb2e3076 | 1785 | t->content == ">=")) |
2f1a1aea FCE |
1786 | { |
1787 | comparison* e = new comparison; | |
1788 | e->left = op1; | |
1789 | e->op = t->content; | |
56099f08 | 1790 | e->tok = t; |
2f1a1aea | 1791 | next (); |
bb2e3076 FCE |
1792 | e->right = parse_shift (); |
1793 | op1 = e; | |
1794 | t = peek (); | |
1795 | } | |
1796 | ||
1797 | return op1; | |
1798 | } | |
1799 | ||
1800 | ||
1801 | expression* | |
1802 | parser::parse_shift () | |
1803 | { | |
1804 | expression* op1 = parse_concatenation (); | |
1805 | ||
1806 | const token* t = peek (); | |
1807 | while (t && t->type == tok_operator && | |
1808 | (t->content == "<<" || t->content == ">>")) | |
1809 | { | |
1810 | binary_expression* e = new binary_expression; | |
1811 | e->left = op1; | |
1812 | e->op = t->content; | |
1813 | e->tok = t; | |
1814 | next (); | |
56099f08 FCE |
1815 | e->right = parse_concatenation (); |
1816 | op1 = e; | |
1817 | t = peek (); | |
2f1a1aea | 1818 | } |
56099f08 FCE |
1819 | |
1820 | return op1; | |
2f1a1aea FCE |
1821 | } |
1822 | ||
1823 | ||
1824 | expression* | |
1825 | parser::parse_concatenation () | |
1826 | { | |
1827 | expression* op1 = parse_additive (); | |
1828 | ||
1829 | const token* t = peek (); | |
1830 | // XXX: the actual awk string-concatenation operator is *whitespace*. | |
1831 | // I don't know how to easily to model that here. | |
56099f08 | 1832 | while (t && t->type == tok_operator && t->content == ".") |
2f1a1aea FCE |
1833 | { |
1834 | concatenation* e = new concatenation; | |
1835 | e->left = op1; | |
1836 | e->op = t->content; | |
56099f08 | 1837 | e->tok = t; |
2f1a1aea | 1838 | next (); |
56099f08 FCE |
1839 | e->right = parse_additive (); |
1840 | op1 = e; | |
1841 | t = peek (); | |
2f1a1aea | 1842 | } |
56099f08 FCE |
1843 | |
1844 | return op1; | |
2f1a1aea FCE |
1845 | } |
1846 | ||
1847 | ||
1848 | expression* | |
1849 | parser::parse_additive () | |
1850 | { | |
1851 | expression* op1 = parse_multiplicative (); | |
1852 | ||
1853 | const token* t = peek (); | |
56099f08 | 1854 | while (t && t->type == tok_operator |
2f1a1aea FCE |
1855 | && (t->content == "+" || t->content == "-")) |
1856 | { | |
1857 | binary_expression* e = new binary_expression; | |
1858 | e->op = t->content; | |
1859 | e->left = op1; | |
56099f08 | 1860 | e->tok = t; |
2f1a1aea | 1861 | next (); |
56099f08 FCE |
1862 | e->right = parse_multiplicative (); |
1863 | op1 = e; | |
1864 | t = peek (); | |
2f1a1aea | 1865 | } |
56099f08 FCE |
1866 | |
1867 | return op1; | |
2f1a1aea FCE |
1868 | } |
1869 | ||
1870 | ||
1871 | expression* | |
1872 | parser::parse_multiplicative () | |
1873 | { | |
1874 | expression* op1 = parse_unary (); | |
1875 | ||
1876 | const token* t = peek (); | |
56099f08 | 1877 | while (t && t->type == tok_operator |
2f1a1aea FCE |
1878 | && (t->content == "*" || t->content == "/" || t->content == "%")) |
1879 | { | |
1880 | binary_expression* e = new binary_expression; | |
1881 | e->op = t->content; | |
1882 | e->left = op1; | |
56099f08 | 1883 | e->tok = t; |
2f1a1aea | 1884 | next (); |
56099f08 FCE |
1885 | e->right = parse_unary (); |
1886 | op1 = e; | |
1887 | t = peek (); | |
2f1a1aea | 1888 | } |
56099f08 FCE |
1889 | |
1890 | return op1; | |
2f1a1aea FCE |
1891 | } |
1892 | ||
1893 | ||
1894 | expression* | |
1895 | parser::parse_unary () | |
1896 | { | |
1897 | const token* t = peek (); | |
1898 | if (t && t->type == tok_operator | |
bb2e3076 FCE |
1899 | && (t->content == "+" || |
1900 | t->content == "-" || | |
1901 | t->content == "!" || | |
1902 | t->content == "~" || | |
1903 | false)) | |
2f1a1aea FCE |
1904 | { |
1905 | unary_expression* e = new unary_expression; | |
1906 | e->op = t->content; | |
56099f08 | 1907 | e->tok = t; |
2f1a1aea | 1908 | next (); |
3a20432b | 1909 | e->operand = parse_crement (); |
2f1a1aea FCE |
1910 | return e; |
1911 | } | |
1912 | else | |
bb2e3076 | 1913 | return parse_crement (); |
2f1a1aea FCE |
1914 | } |
1915 | ||
1916 | ||
1917 | expression* | |
1918 | parser::parse_crement () // as in "increment" / "decrement" | |
1919 | { | |
cbfbbf69 FCE |
1920 | // NB: Ideally, we'd parse only a symbol as an operand to the |
1921 | // *crement operators, instead of a general expression value. We'd | |
1922 | // need more complex lookahead code to tell apart the postfix cases. | |
1923 | // So we just punt, and leave it to pass-3 to signal errors on | |
1924 | // cases like "4++". | |
1925 | ||
2f1a1aea FCE |
1926 | const token* t = peek (); |
1927 | if (t && t->type == tok_operator | |
1928 | && (t->content == "++" || t->content == "--")) | |
1929 | { | |
1930 | pre_crement* e = new pre_crement; | |
1931 | e->op = t->content; | |
56099f08 | 1932 | e->tok = t; |
2f1a1aea FCE |
1933 | next (); |
1934 | e->operand = parse_value (); | |
1935 | return e; | |
1936 | } | |
1937 | ||
1938 | // post-crement or non-crement | |
1939 | expression *op1 = parse_value (); | |
1940 | ||
1941 | t = peek (); | |
1942 | if (t && t->type == tok_operator | |
1943 | && (t->content == "++" || t->content == "--")) | |
1944 | { | |
1945 | post_crement* e = new post_crement; | |
1946 | e->op = t->content; | |
56099f08 | 1947 | e->tok = t; |
2f1a1aea FCE |
1948 | next (); |
1949 | e->operand = op1; | |
1950 | return e; | |
1951 | } | |
1952 | else | |
1953 | return op1; | |
1954 | } | |
1955 | ||
1956 | ||
1957 | expression* | |
1958 | parser::parse_value () | |
1959 | { | |
1960 | const token* t = peek (); | |
1961 | if (! t) | |
1962 | throw parse_error ("expected value"); | |
1963 | ||
1964 | if (t->type == tok_operator && t->content == "(") | |
1965 | { | |
1966 | next (); | |
1967 | expression* e = parse_expression (); | |
1968 | t = next (); | |
1969 | if (! (t->type == tok_operator && t->content == ")")) | |
1970 | throw parse_error ("expected ')'"); | |
1971 | return e; | |
1972 | } | |
1973 | else if (t->type == tok_identifier) | |
1974 | return parse_symbol (); | |
1975 | else | |
1976 | return parse_literal (); | |
1977 | } | |
1978 | ||
1979 | ||
d02548c0 GH |
1980 | const token * |
1981 | parser::parse_hist_op_or_bare_name (hist_op *&hop, string &name) | |
1982 | { | |
1983 | hop = NULL; | |
1984 | const token* t = expect_ident (name); | |
1985 | if (name == "@hist_linear" || name == "@hist_log") | |
1986 | { | |
1987 | hop = new hist_op; | |
1988 | if (name == "@hist_linear") | |
1989 | hop->htype = hist_linear; | |
1990 | else if (name == "@hist_log") | |
1991 | hop->htype = hist_log; | |
1992 | hop->tok = t; | |
1993 | expect_op("("); | |
1994 | hop->stat = parse_expression (); | |
1995 | int64_t tnum; | |
1996 | if (hop->htype == hist_linear) | |
1997 | { | |
1998 | for (size_t i = 0; i < 3; ++i) | |
1999 | { | |
2000 | expect_op (","); | |
2001 | expect_number (tnum); | |
2002 | hop->params.push_back (tnum); | |
2003 | } | |
2004 | } | |
2005 | else | |
2006 | { | |
2007 | assert(hop->htype == hist_log); | |
2008 | if (peek_op (",")) | |
2009 | { | |
2010 | expect_op (","); | |
2011 | expect_number (tnum); | |
2012 | hop->params.push_back (tnum); | |
2013 | } | |
2014 | else | |
2015 | { | |
2016 | // FIXME (magic value): Logarithmic histograms get 64 | |
2017 | // buckets by default. | |
2018 | hop->params.push_back (64); | |
2019 | } | |
2020 | } | |
2021 | expect_op(")"); | |
2022 | } | |
2023 | return t; | |
2024 | } | |
2025 | ||
2026 | ||
2027 | indexable* | |
2028 | parser::parse_indexable () | |
2029 | { | |
2030 | hist_op *hop = NULL; | |
2031 | string name; | |
2032 | const token *tok = parse_hist_op_or_bare_name(hop, name); | |
2033 | if (hop) | |
2034 | return hop; | |
2035 | else | |
2036 | { | |
2037 | symbol* sym = new symbol; | |
2038 | sym->name = name; | |
2039 | sym->tok = tok; | |
2040 | return sym; | |
2041 | } | |
2042 | } | |
2043 | ||
2044 | ||
2045 | // var, indexable[index], func(parms), printf("...", ...), $var, $var->member, @stat_op(stat) | |
2f1a1aea | 2046 | expression* |
0fefb486 | 2047 | parser::parse_symbol () |
2f1a1aea | 2048 | { |
d02548c0 GH |
2049 | hist_op *hop = NULL; |
2050 | symbol *sym = NULL; | |
d7f3e0c5 | 2051 | string name; |
d02548c0 GH |
2052 | const token *t = parse_hist_op_or_bare_name(hop, name); |
2053 | ||
2054 | if (!hop) | |
0fefb486 | 2055 | { |
d02548c0 GH |
2056 | // If we didn't get a hist_op, then we did get an identifier. We can |
2057 | // now scrutinize this identifier for the various magic forms of identifier | |
2058 | // (printf, @stat_op, and $var...) | |
2059 | ||
2060 | if (name.size() > 0 && name[0] == '@') | |
d7f3e0c5 | 2061 | { |
d02548c0 GH |
2062 | stat_op *sop = new stat_op; |
2063 | if (name == "@avg") | |
2064 | sop->ctype = sc_average; | |
2065 | else if (name == "@count") | |
2066 | sop->ctype = sc_count; | |
2067 | else if (name == "@sum") | |
2068 | sop->ctype = sc_sum; | |
2069 | else if (name == "@min") | |
2070 | sop->ctype = sc_min; | |
2071 | else if (name == "@max") | |
2072 | sop->ctype = sc_max; | |
2073 | else | |
2074 | throw parse_error("unknown statistic operator " + name); | |
2075 | expect_op("("); | |
2076 | sop->tok = t; | |
2077 | sop->stat = parse_expression (); | |
2078 | expect_op(")"); | |
2079 | return sop; | |
2080 | } | |
2081 | ||
2082 | else if (name.size() > 0 && (name == "print" | |
2083 | || name == "sprint" | |
2084 | || name == "printf" | |
01133ccb LG |
2085 | || name == "sprintf" |
2086 | || name == "lket_trace_extra")) | |
d02548c0 GH |
2087 | { |
2088 | print_format *fmt = new print_format; | |
2089 | fmt->tok = t; | |
2090 | fmt->print_with_format = (name[name.size() - 1] == 'f'); | |
2091 | fmt->print_to_stream = (name[0] == 'p'); | |
01133ccb LG |
2092 | |
2093 | fmt->lket_trace_extra = false; | |
2094 | ||
2095 | if(name == "lket_trace_extra") | |
2096 | { | |
2097 | fmt->print_with_format = true; | |
2098 | fmt->print_to_stream = true; | |
2099 | fmt->lket_trace_extra = true; | |
2100 | } | |
2101 | ||
d02548c0 GH |
2102 | expect_op("("); |
2103 | if (fmt->print_with_format) | |
2104 | { | |
2105 | // Consume and convert a format string, and any subsequent | |
2106 | // arguments. Agreement between the format string and the | |
2107 | // arguments is postponed to the typechecking phase. | |
2108 | string tmp; | |
2109 | expect_unknown (tok_string, tmp); | |
a9c62ac9 | 2110 | fmt->raw_components = tmp; |
d02548c0 GH |
2111 | fmt->components = print_format::string_to_components (tmp); |
2112 | while (!peek_op (")")) | |
2113 | { | |
2114 | expect_op(","); | |
2115 | expression *e = parse_expression (); | |
2116 | fmt->args.push_back(e); | |
2117 | } | |
d7f3e0c5 | 2118 | } |
a4636912 GH |
2119 | else if (name == "print" && |
2120 | (peek_kw("@hist_linear") || | |
2121 | peek_kw("@hist_log"))) | |
2122 | { | |
2123 | // We have a special case where we recognize | |
2124 | // print(@hist_foo(bar)) as a magic print-the-histogram | |
2125 | // construct. This is sort of gross but it avoids | |
2126 | // promoting histogram references to typeful | |
2127 | // expressions. | |
1bbeef03 GH |
2128 | |
2129 | hop = NULL; | |
2130 | t = parse_hist_op_or_bare_name(hop, name); | |
2131 | assert(hop); | |
2132 | ||
2133 | // It is, sadly, possible that even while parsing a | |
2134 | // hist_op, we *mis-guessed* and the user wishes to | |
2135 | // print(@hist_op(foo)[bucket]), a scalar. In that case | |
2136 | // we must parse the arrayindex and print an expression. | |
2137 | ||
2138 | if (!peek_op ("[")) | |
2139 | fmt->hist = hop; | |
2140 | else | |
2141 | { | |
2142 | // This is simplified version of the | |
2143 | // multi-array-index parser below, because we can | |
2144 | // only ever have one index on a histogram anyways. | |
2145 | expect_op("["); | |
2146 | struct arrayindex* ai = new arrayindex; | |
2147 | ai->tok = t; | |
2148 | ai->base = hop; | |
2149 | ai->indexes.push_back (parse_expression ()); | |
2150 | expect_op("]"); | |
2151 | fmt->args.push_back(ai); | |
2152 | } | |
a4636912 | 2153 | } |
d7f3e0c5 | 2154 | else |
d02548c0 GH |
2155 | { |
2156 | // If we are not printing with a format string, we permit | |
2157 | // exactly one argument (of any type). | |
2158 | expression *e = parse_expression (); | |
2159 | fmt->args.push_back(e); | |
2160 | } | |
2161 | expect_op(")"); | |
2162 | return fmt; | |
2163 | } | |
2164 | ||
2165 | else if (name.size() > 0 && name[0] == '$') | |
2166 | { | |
2167 | // target_symbol time | |
2168 | target_symbol *tsym = new target_symbol; | |
2169 | tsym->tok = t; | |
2170 | tsym->base_name = name; | |
2171 | while (true) | |
2172 | { | |
2173 | string c; | |
2174 | if (peek_op ("->")) | |
2175 | { | |
2176 | next(); | |
2177 | expect_ident (c); | |
2178 | tsym->components.push_back | |
2179 | (make_pair (target_symbol::comp_struct_member, c)); | |
2180 | } | |
2181 | else if (peek_op ("[")) | |
2182 | { | |
2183 | next(); | |
2184 | expect_unknown (tok_number, c); | |
2185 | expect_op ("]"); | |
2186 | tsym->components.push_back | |
2187 | (make_pair (target_symbol::comp_literal_array_index, c)); | |
2188 | } | |
2189 | else | |
2190 | break; | |
2191 | } | |
2192 | return tsym; | |
2193 | } | |
2194 | ||
2195 | else if (peek_op ("(")) // function call | |
2196 | { | |
2197 | next (); | |
2198 | struct functioncall* f = new functioncall; | |
2199 | f->tok = t; | |
2200 | f->function = name; | |
2201 | // Allow empty actual parameter list | |
2202 | if (peek_op (")")) | |
2203 | { | |
2204 | next (); | |
2205 | return f; | |
2206 | } | |
2207 | while (1) | |
2208 | { | |
2209 | f->args.push_back (parse_expression ()); | |
2210 | if (peek_op (")")) | |
2211 | { | |
2212 | next(); | |
2213 | break; | |
2214 | } | |
2215 | else if (peek_op (",")) | |
2216 | { | |
2217 | next(); | |
2218 | continue; | |
2219 | } | |
2220 | else | |
2221 | throw parse_error ("expected ',' or ')'"); | |
2222 | } | |
2223 | return f; | |
2224 | } | |
2225 | ||
2226 | else | |
2227 | { | |
2228 | sym = new symbol; | |
2229 | sym->name = name; | |
2230 | sym->tok = t; | |
d7f3e0c5 | 2231 | } |
0fefb486 | 2232 | } |
d7f3e0c5 | 2233 | |
d02548c0 GH |
2234 | // By now, either we had a hist_op in the first place, or else |
2235 | // we had a plain word and it was converted to a symbol. | |
2236 | ||
70c743d8 | 2237 | assert (!hop != !sym); // logical XOR |
d02548c0 GH |
2238 | |
2239 | // All that remains is to check for array indexing | |
2240 | ||
d7f3e0c5 | 2241 | if (peek_op ("[")) // array |
2f1a1aea FCE |
2242 | { |
2243 | next (); | |
2244 | struct arrayindex* ai = new arrayindex; | |
d02548c0 GH |
2245 | ai->tok = t; |
2246 | ||
2247 | if (hop) | |
2248 | ai->base = hop; | |
2249 | else | |
2250 | ai->base = sym; | |
2251 | ||
2f1a1aea FCE |
2252 | while (1) |
2253 | { | |
2254 | ai->indexes.push_back (parse_expression ()); | |
d7f3e0c5 GH |
2255 | if (peek_op ("]")) |
2256 | { | |
2257 | next(); | |
2258 | break; | |
2259 | } | |
2260 | else if (peek_op (",")) | |
2261 | { | |
2262 | next(); | |
2263 | continue; | |
2264 | } | |
2f1a1aea FCE |
2265 | else |
2266 | throw parse_error ("expected ',' or ']'"); | |
2267 | } | |
2268 | return ai; | |
2269 | } | |
d02548c0 GH |
2270 | |
2271 | // If we got to here, we *should* have a symbol; if we have | |
2272 | // a hist_op on its own, it doesn't count as an expression, | |
2273 | // so we throw a parse error. | |
2274 | ||
2275 | if (hop) | |
2276 | throw parse_error("base histogram operator where expression expected", t); | |
2277 | ||
2278 | return sym; | |
2f1a1aea | 2279 | } |
56099f08 | 2280 |