]>
Commit | Line | Data |
---|---|---|
2f1a1aea | 1 | // recursive descent parser for systemtap scripts |
482fe2af | 2 | // Copyright (C) 2005-2009 Red Hat Inc. |
77a5c1f9 | 3 | // Copyright (C) 2006 Intel Corporation. |
5811366a | 4 | // Copyright (C) 2007 Bull S.A.S |
69c68955 FCE |
5 | // |
6 | // This file is part of systemtap, and is free software. You can | |
7 | // redistribute it and/or modify it under the terms of the GNU General | |
8 | // Public License (GPL); either version 2, or (at your option) any | |
9 | // later version. | |
2f1a1aea | 10 | |
2b066ec1 | 11 | #include "config.h" |
2f1a1aea FCE |
12 | #include "staptree.h" |
13 | #include "parse.h" | |
177a8ead | 14 | #include "session.h" |
3f99432c FCE |
15 | #include "util.h" |
16 | ||
2b066ec1 | 17 | #include <iostream> |
eacb10ce | 18 | |
2b066ec1 | 19 | #include <fstream> |
2f1a1aea | 20 | #include <cctype> |
9c0c0e46 | 21 | #include <cstdlib> |
29e64872 | 22 | #include <cassert> |
9c0c0e46 FCE |
23 | #include <cerrno> |
24 | #include <climits> | |
57b73400 | 25 | #include <sstream> |
f74fb737 | 26 | #include <cstring> |
3f99432c | 27 | #include <cctype> |
eacb10ce FCE |
28 | #include <iterator> |
29 | ||
7a468d68 FCE |
30 | extern "C" { |
31 | #include <fnmatch.h> | |
32 | } | |
2f1a1aea FCE |
33 | |
34 | using namespace std; | |
35 | ||
36 | // ------------------------------------------------------------------------ | |
37 | ||
bb2e3076 FCE |
38 | |
39 | ||
177a8ead FCE |
40 | parser::parser (systemtap_session& s, istream& i, bool p): |
41 | session (s), | |
24cb178f | 42 | input_name ("<input>"), free_input (0), |
213bee8f | 43 | input (i, input_name, s), privileged (p), |
6e213f58 | 44 | context(con_unknown), last_t (0), next_t (0), num_errors (0) |
2f1a1aea FCE |
45 | { } |
46 | ||
177a8ead FCE |
47 | parser::parser (systemtap_session& s, const string& fn, bool p): |
48 | session (s), | |
2f1a1aea | 49 | input_name (fn), free_input (new ifstream (input_name.c_str(), ios::in)), |
213bee8f | 50 | input (* free_input, input_name, s), privileged (p), |
6e213f58 | 51 | context(con_unknown), last_t (0), next_t (0), num_errors (0) |
2f1a1aea FCE |
52 | { } |
53 | ||
54 | parser::~parser() | |
55 | { | |
56 | if (free_input) delete free_input; | |
57 | } | |
58 | ||
59 | ||
82919855 | 60 | stapfile* |
177a8ead | 61 | parser::parse (systemtap_session& s, std::istream& i, bool pr) |
82919855 | 62 | { |
177a8ead | 63 | parser p (s, i, pr); |
82919855 FCE |
64 | return p.parse (); |
65 | } | |
66 | ||
67 | ||
68 | stapfile* | |
177a8ead | 69 | parser::parse (systemtap_session& s, const std::string& n, bool pr) |
82919855 | 70 | { |
177a8ead | 71 | parser p (s, n, pr); |
82919855 FCE |
72 | return p.parse (); |
73 | } | |
74 | ||
d7f3e0c5 GH |
75 | static string |
76 | tt2str(token_type tt) | |
77 | { | |
78 | switch (tt) | |
79 | { | |
80 | case tok_junk: return "junk"; | |
81 | case tok_identifier: return "identifier"; | |
82 | case tok_operator: return "operator"; | |
83 | case tok_string: return "string"; | |
84 | case tok_number: return "number"; | |
85 | case tok_embedded: return "embedded-code"; | |
6e213f58 | 86 | case tok_keyword: return "keyword"; |
d7f3e0c5 GH |
87 | } |
88 | return "unknown token"; | |
89 | } | |
82919855 | 90 | |
0323ed4d WC |
91 | ostream& |
92 | operator << (ostream& o, const source_loc& loc) | |
93 | { | |
a704a23b | 94 | o << loc.file->name << ":" |
0323ed4d WC |
95 | << loc.line << ":" |
96 | << loc.column; | |
97 | ||
98 | return o; | |
99 | } | |
100 | ||
56099f08 FCE |
101 | ostream& |
102 | operator << (ostream& o, const token& t) | |
103 | { | |
d7f3e0c5 | 104 | o << tt2str(t.type); |
56099f08 | 105 | |
6e213f58 | 106 | if (t.type != tok_embedded && t.type != tok_keyword) // XXX: other types? |
56099f08 | 107 | { |
24cb178f FCE |
108 | o << " '"; |
109 | for (unsigned i=0; i<t.content.length(); i++) | |
110 | { | |
111 | char c = t.content[i]; | |
112 | o << (isprint (c) ? c : '?'); | |
113 | } | |
114 | o << "'"; | |
56099f08 | 115 | } |
56099f08 | 116 | |
dff50e09 | 117 | o << " at " |
0323ed4d | 118 | << t.location; |
56099f08 FCE |
119 | |
120 | return o; | |
121 | } | |
122 | ||
123 | ||
dff50e09 | 124 | void |
2f1a1aea FCE |
125 | parser::print_error (const parse_error &pe) |
126 | { | |
1b1b4ceb | 127 | string align_parse_error (" "); |
2f1a1aea FCE |
128 | cerr << "parse error: " << pe.what () << endl; |
129 | ||
177a8ead FCE |
130 | if (pe.tok) |
131 | { | |
132 | cerr << "\tat: " << *pe.tok << endl; | |
1b1b4ceb | 133 | session.print_error_source (cerr, align_parse_error, pe.tok); |
177a8ead | 134 | } |
2f1a1aea | 135 | else |
177a8ead FCE |
136 | { |
137 | const token* t = last_t; | |
138 | if (t) | |
1b1b4ceb RA |
139 | { |
140 | cerr << "\tsaw: " << *t << endl; | |
141 | session.print_error_source (cerr, align_parse_error, t); | |
142 | } | |
177a8ead FCE |
143 | else |
144 | cerr << "\tsaw: " << input_name << " EOF" << endl; | |
145 | } | |
2f1a1aea FCE |
146 | |
147 | // XXX: make it possible to print the last input line, | |
148 | // so as to line up an arrow with the specific error column | |
149 | ||
150 | num_errors ++; | |
151 | } | |
152 | ||
153 | ||
dff50e09 | 154 | const token* |
2f1a1aea FCE |
155 | parser::last () |
156 | { | |
157 | return last_t; | |
158 | } | |
159 | ||
160 | ||
c434ec7e FCE |
161 | |
162 | template <typename OPERAND> | |
163 | bool eval_comparison (const OPERAND& lhs, const token* op, const OPERAND& rhs) | |
164 | { | |
165 | if (op->type == tok_operator && op->content == "<=") | |
166 | { return lhs <= rhs; } | |
167 | else if (op->type == tok_operator && op->content == ">=") | |
168 | { return lhs >= rhs; } | |
169 | else if (op->type == tok_operator && op->content == "<") | |
170 | { return lhs < rhs; } | |
171 | else if (op->type == tok_operator && op->content == ">") | |
172 | { return lhs > rhs; } | |
173 | else if (op->type == tok_operator && op->content == "==") | |
174 | { return lhs == rhs; } | |
175 | else if (op->type == tok_operator && op->content == "!=") | |
176 | { return lhs != rhs; } | |
177 | else | |
178 | throw parse_error ("expected comparison operator", op); | |
179 | } | |
180 | ||
181 | ||
177a8ead FCE |
182 | // Here, we perform on-the-fly preprocessing. |
183 | // The basic form is %( CONDITION %? THEN-TOKENS %: ELSE-TOKENS %) | |
44ce8ed5 FCE |
184 | // where CONDITION is: kernel_v[r] COMPARISON-OP "version-string" |
185 | // or: arch COMPARISON-OP "arch-string" | |
561079c8 | 186 | // or: CONFIG_foo COMPARISON-OP "config-string" |
717a457b | 187 | // or: CONFIG_foo COMPARISON-OP number |
5811366a FCE |
188 | // or: "string1" COMPARISON-OP "string2" |
189 | // or: number1 COMPARISON-OP number2 | |
44ce8ed5 | 190 | // The %: ELSE-TOKENS part is optional. |
177a8ead FCE |
191 | // |
192 | // e.g. %( kernel_v > "2.5" %? "foo" %: "baz" %) | |
c434ec7e | 193 | // e.g. %( arch != "i?86" %? "foo" %: "baz" %) |
561079c8 | 194 | // e.g. %( CONFIG_foo %? "foo" %: "baz" %) |
177a8ead FCE |
195 | // |
196 | // Up to an entire %( ... %) expression is processed by a single call | |
197 | // to this function. Tokens included by any nested conditions are | |
198 | // enqueued in a private vector. | |
199 | ||
200 | bool eval_pp_conditional (systemtap_session& s, | |
201 | const token* l, const token* op, const token* r) | |
202 | { | |
44ce8ed5 FCE |
203 | if (l->type == tok_identifier && (l->content == "kernel_v" || |
204 | l->content == "kernel_vr")) | |
205 | { | |
206 | string target_kernel_vr = s.kernel_release; | |
197a4d62 | 207 | string target_kernel_v = s.kernel_base_release; |
dff50e09 | 208 | |
44ce8ed5 FCE |
209 | if (! (r->type == tok_string)) |
210 | throw parse_error ("expected string literal", r); | |
7a468d68 | 211 | |
dff50e09 | 212 | string target = (l->content == "kernel_vr" ? |
7a468d68 FCE |
213 | target_kernel_vr.c_str() : |
214 | target_kernel_v.c_str()); | |
215 | string query = r->content; | |
216 | bool rhs_wildcard = (strpbrk (query.c_str(), "*?[") != 0); | |
217 | ||
44ce8ed5 FCE |
218 | // collect acceptable strverscmp results. |
219 | int rvc_ok1, rvc_ok2; | |
7a468d68 | 220 | bool wc_ok = false; |
44ce8ed5 FCE |
221 | if (op->type == tok_operator && op->content == "<=") |
222 | { rvc_ok1 = -1; rvc_ok2 = 0; } | |
223 | else if (op->type == tok_operator && op->content == ">=") | |
224 | { rvc_ok1 = 1; rvc_ok2 = 0; } | |
225 | else if (op->type == tok_operator && op->content == "<") | |
226 | { rvc_ok1 = -1; rvc_ok2 = -1; } | |
227 | else if (op->type == tok_operator && op->content == ">") | |
228 | { rvc_ok1 = 1; rvc_ok2 = 1; } | |
229 | else if (op->type == tok_operator && op->content == "==") | |
7a468d68 | 230 | { rvc_ok1 = 0; rvc_ok2 = 0; wc_ok = true; } |
44ce8ed5 | 231 | else if (op->type == tok_operator && op->content == "!=") |
7a468d68 | 232 | { rvc_ok1 = -1; rvc_ok2 = 1; wc_ok = true; } |
44ce8ed5 FCE |
233 | else |
234 | throw parse_error ("expected comparison operator", op); | |
7a468d68 FCE |
235 | |
236 | if ((!wc_ok) && rhs_wildcard) | |
237 | throw parse_error ("wildcard not allowed with order comparison operators", op); | |
238 | ||
239 | if (rhs_wildcard) | |
240 | { | |
241 | int rvc_result = fnmatch (query.c_str(), target.c_str(), | |
242 | FNM_NOESCAPE); // spooky | |
243 | bool badness = (rvc_result == 0) ^ (op->content == "=="); | |
244 | return !badness; | |
245 | } | |
246 | else | |
247 | { | |
248 | int rvc_result = strverscmp (target.c_str(), query.c_str()); | |
249 | // normalize rvc_result | |
250 | if (rvc_result < 0) rvc_result = -1; | |
251 | if (rvc_result > 0) rvc_result = 1; | |
252 | return (rvc_result == rvc_ok1 || rvc_result == rvc_ok2); | |
253 | } | |
44ce8ed5 FCE |
254 | } |
255 | else if (l->type == tok_identifier && l->content == "arch") | |
256 | { | |
257 | string target_architecture = s.architecture; | |
258 | if (! (r->type == tok_string)) | |
259 | throw parse_error ("expected string literal", r); | |
260 | string query_architecture = r->content; | |
dff50e09 | 261 | |
7a468d68 FCE |
262 | int nomatch = fnmatch (query_architecture.c_str(), |
263 | target_architecture.c_str(), | |
264 | FNM_NOESCAPE); // still spooky | |
265 | ||
561079c8 FCE |
266 | bool result; |
267 | if (op->type == tok_operator && op->content == "==") | |
268 | result = !nomatch; | |
269 | else if (op->type == tok_operator && op->content == "!=") | |
270 | result = nomatch; | |
271 | else | |
272 | throw parse_error ("expected '==' or '!='", op); | |
273 | ||
274 | return result; | |
275 | } | |
717a457b | 276 | else if (l->type == tok_identifier && l->content.substr(0,7) == "CONFIG_") |
561079c8 | 277 | { |
717a457b MW |
278 | if (r->type == tok_string) |
279 | { | |
280 | string lhs = s.kernel_config[l->content]; // may be empty | |
281 | string rhs = r->content; | |
561079c8 | 282 | |
717a457b | 283 | int nomatch = fnmatch (rhs.c_str(), lhs.c_str(), FNM_NOESCAPE); // still spooky |
561079c8 | 284 | |
717a457b MW |
285 | bool result; |
286 | if (op->type == tok_operator && op->content == "==") | |
287 | result = !nomatch; | |
288 | else if (op->type == tok_operator && op->content == "!=") | |
289 | result = nomatch; | |
290 | else | |
291 | throw parse_error ("expected '==' or '!='", op); | |
dff50e09 | 292 | |
717a457b MW |
293 | return result; |
294 | } | |
295 | else if (r->type == tok_number) | |
296 | { | |
297 | const char* startp = s.kernel_config[l->content].c_str (); | |
298 | char* endp = (char*) startp; | |
299 | errno = 0; | |
300 | int64_t lhs = (int64_t) strtoll (startp, & endp, 0); | |
301 | if (errno == ERANGE || errno == EINVAL || *endp != '\0') | |
302 | throw parse_error ("Config option value not a number", l); | |
303 | ||
304 | int64_t rhs = lex_cast<int64_t>(r->content); | |
305 | return eval_comparison (lhs, op, rhs); | |
306 | } | |
307 | else | |
308 | throw parse_error ("expected string or number literal as right value", | |
309 | r); | |
dff50e09 | 310 | } |
c434ec7e | 311 | else if (l->type == tok_string && r->type == tok_string) |
5811366a | 312 | { |
c434ec7e FCE |
313 | string lhs = l->content; |
314 | string rhs = r->content; | |
315 | return eval_comparison (lhs, op, rhs); | |
316 | // NB: no wildcarding option here | |
317 | } | |
318 | else if (l->type == tok_number && r->type == tok_number) | |
319 | { | |
320 | int64_t lhs = lex_cast<int64_t>(l->content); | |
321 | int64_t rhs = lex_cast<int64_t>(r->content); | |
322 | return eval_comparison (lhs, op, rhs); | |
7a468d68 | 323 | // NB: no wildcarding option here |
5811366a FCE |
324 | } |
325 | else if (l->type == tok_string && r->type == tok_number | |
326 | && op->type == tok_operator) | |
327 | throw parse_error ("expected string literal as right value", r); | |
328 | else if (l->type == tok_number && r->type == tok_string | |
329 | && op->type == tok_operator) | |
330 | throw parse_error ("expected number literal as right value", r); | |
c434ec7e | 331 | |
177a8ead | 332 | else |
561079c8 | 333 | throw parse_error ("expected 'arch' or 'kernel_v' or 'kernel_vr' or 'CONFIG_...'\n" |
5811366a | 334 | " or comparison between strings or integers", l); |
177a8ead FCE |
335 | } |
336 | ||
337 | ||
5811366a | 338 | // Only tokens corresponding to the TRUE statement must be expanded |
177a8ead | 339 | const token* |
3f847830 | 340 | parser::scan_pp (bool wildcard) |
177a8ead FCE |
341 | { |
342 | while (true) | |
343 | { | |
344 | if (enqueued_pp.size() > 0) | |
345 | { | |
346 | const token* t = enqueued_pp[0]; | |
347 | enqueued_pp.erase (enqueued_pp.begin()); | |
348 | return t; | |
349 | } | |
350 | ||
3f847830 | 351 | const token* t = input.scan (wildcard); // NB: not recursive! |
177a8ead FCE |
352 | if (t == 0) // EOF |
353 | return t; | |
dff50e09 | 354 | |
177a8ead FCE |
355 | if (! (t->type == tok_operator && t->content == "%(")) // ordinary token |
356 | return t; | |
357 | ||
358 | // We have a %( - it's time to throw a preprocessing party! | |
359 | ||
2d7881bf PP |
360 | bool result = false; |
361 | bool and_result = true; | |
362 | const token *n = NULL; | |
363 | do { | |
364 | const token *l, *op, *r; | |
365 | l = input.scan (false); // NB: not recursive, though perhaps could be | |
366 | op = input.scan (false); | |
367 | r = input.scan (false); | |
368 | if (l == 0 || op == 0 || r == 0) | |
369 | throw parse_error ("incomplete condition after '%('", t); | |
370 | // NB: consider generalizing to consume all tokens until %?, and | |
371 | // passing that as a vector to an evaluator. | |
372 | ||
373 | // Do not evaluate the condition if we haven't expanded everything. | |
374 | // This may occur when having several recursive conditionals. | |
375 | and_result &= eval_pp_conditional (session, l, op, r); | |
376 | delete l; | |
377 | delete op; | |
378 | delete r; | |
379 | delete n; | |
380 | ||
381 | n = input.scan (); | |
382 | if (n && n->type == tok_operator && n->content == "&&") | |
383 | continue; | |
384 | result |= and_result; | |
385 | and_result = true; | |
386 | if (! (n && n->type == tok_operator && n->content == "||")) | |
387 | break; | |
388 | } while (true); | |
3f847830 FCE |
389 | |
390 | /* | |
391 | clog << "PP eval (" << *t << ") == " << result << endl; | |
392 | */ | |
393 | ||
2d7881bf | 394 | const token *m = n; // NB: not recursive |
177a8ead FCE |
395 | if (! (m && m->type == tok_operator && m->content == "%?")) |
396 | throw parse_error ("expected '%?' marker for conditional", t); | |
70c743d8 | 397 | delete m; // "%?" |
177a8ead FCE |
398 | |
399 | vector<const token*> my_enqueued_pp; | |
3f847830 FCE |
400 | |
401 | int nesting = 0; | |
c28668ea | 402 | int then = 0; |
177a8ead FCE |
403 | while (true) // consume THEN tokens |
404 | { | |
3f847830 FCE |
405 | try |
406 | { | |
407 | m = result ? scan_pp (wildcard) : input.scan (wildcard); | |
408 | } | |
409 | catch (const parse_error &e) | |
410 | { | |
dff50e09 | 411 | if (result) throw e; // propagate errors if THEN branch taken |
d57671d3 | 412 | continue; |
3f847830 FCE |
413 | } |
414 | ||
415 | if (m && m->type == tok_operator && m->content == "%(") // nested %( | |
416 | nesting ++; | |
c28668ea WH |
417 | if (m && m->type == tok_operator && m->content == "%?") { |
418 | then ++; | |
419 | if (nesting != then) | |
420 | throw parse_error ("incomplete conditional - missing '%('", m); | |
421 | } | |
3f847830 FCE |
422 | if (nesting == 0 && m && (m->type == tok_operator && (m->content == "%:" || // ELSE |
423 | m->content == "%)"))) // END | |
177a8ead | 424 | break; |
c28668ea | 425 | if (nesting && m && m->type == tok_operator && m->content == "%)") { // nested %) |
3f847830 | 426 | nesting --; |
c28668ea WH |
427 | then --; |
428 | } | |
3f847830 | 429 | |
d57671d3 FCE |
430 | if (!m) |
431 | throw parse_error ("incomplete conditional - missing '%:' or '%)'", t); | |
432 | if (result) | |
177a8ead | 433 | my_enqueued_pp.push_back (m); |
d57671d3 | 434 | if (!result) |
3f847830 FCE |
435 | delete m; // do nothing, just dispose of unkept THEN token |
436 | ||
437 | continue; | |
177a8ead | 438 | } |
dff50e09 | 439 | |
177a8ead | 440 | if (m && m->type == tok_operator && m->content == "%:") // ELSE |
70c743d8 JS |
441 | { |
442 | delete m; // "%:" | |
3f847830 | 443 | int nesting = 0; |
c28668ea | 444 | int then = 0; |
70c743d8 JS |
445 | while (true) |
446 | { | |
3f847830 FCE |
447 | try |
448 | { | |
449 | m = result ? input.scan (wildcard) : scan_pp (wildcard); | |
dff50e09 | 450 | } |
3f847830 FCE |
451 | catch (const parse_error& e) |
452 | { | |
dff50e09 | 453 | if (!result) throw e; // propagate errors if ELSE branch taken |
d57671d3 | 454 | continue; |
3f847830 FCE |
455 | } |
456 | ||
457 | if (m && m->type == tok_operator && m->content == "%(") // nested %( | |
458 | nesting ++; | |
c28668ea WH |
459 | if (m && m->type == tok_operator && m->content == "%?") { |
460 | then ++; | |
461 | if (nesting != then) | |
462 | throw parse_error ("incomplete conditional - missing '%('", m); | |
463 | } | |
3f847830 | 464 | if (nesting == 0 && m && m->type == tok_operator && m->content == "%)") // END |
70c743d8 | 465 | break; |
c28668ea | 466 | if (nesting && m && m->type == tok_operator && m->content == "%)") { // nested %) |
3f847830 | 467 | nesting --; |
c28668ea WH |
468 | then --; |
469 | } | |
3f847830 | 470 | |
d57671d3 | 471 | if (!m) |
3f847830 | 472 | throw parse_error ("incomplete conditional - missing %)", t); |
d57671d3 | 473 | if (!result) |
dff50e09 | 474 | my_enqueued_pp.push_back (m); |
d57671d3 | 475 | if (result) |
3f847830 FCE |
476 | delete m; // do nothing, just dispose of unkept ELSE token |
477 | ||
478 | continue; | |
70c743d8 JS |
479 | } |
480 | } | |
3f847830 FCE |
481 | |
482 | /* | |
483 | clog << "PP eval (" << *t << ") == " << result << " tokens: " << endl; | |
484 | for (unsigned k=0; k<my_enqueued_pp.size(); k++) | |
485 | clog << * my_enqueued_pp[k] << endl; | |
486 | clog << endl; | |
487 | */ | |
488 | ||
70c743d8 JS |
489 | delete t; // "%(" |
490 | delete m; // "%)" | |
177a8ead | 491 | |
3f847830 | 492 | |
177a8ead FCE |
493 | // NB: we transcribe the retained tokens here, and not inside |
494 | // the THEN/ELSE while loops. If it were done there, each loop | |
495 | // would become infinite (each iteration consuming an ordinary | |
496 | // token the previous one just pushed there). Guess how I | |
497 | // figured that out. | |
498 | enqueued_pp.insert (enqueued_pp.end(), | |
499 | my_enqueued_pp.begin(), | |
500 | my_enqueued_pp.end()); | |
501 | ||
502 | // Go back to outermost while(true) loop. We hope that at least | |
503 | // some THEN or ELSE tokens were enqueued. If not, around we go | |
504 | // again, until EOF. | |
505 | } | |
506 | } | |
507 | ||
508 | ||
2f1a1aea | 509 | const token* |
0c218afb | 510 | parser::next (bool wildcard) |
2f1a1aea FCE |
511 | { |
512 | if (! next_t) | |
0c218afb | 513 | next_t = scan_pp (wildcard); |
2f1a1aea FCE |
514 | if (! next_t) |
515 | throw parse_error ("unexpected end-of-file"); | |
516 | ||
2f1a1aea FCE |
517 | last_t = next_t; |
518 | // advance by zeroing next_t | |
519 | next_t = 0; | |
520 | return last_t; | |
521 | } | |
522 | ||
523 | ||
524 | const token* | |
0c218afb | 525 | parser::peek (bool wildcard) |
2f1a1aea FCE |
526 | { |
527 | if (! next_t) | |
0c218afb | 528 | next_t = scan_pp (wildcard); |
2f1a1aea FCE |
529 | |
530 | // don't advance by zeroing next_t | |
531 | last_t = next_t; | |
532 | return next_t; | |
533 | } | |
534 | ||
535 | ||
d7f3e0c5 GH |
536 | static inline bool |
537 | tok_is(token const * t, token_type tt, string const & expected) | |
538 | { | |
539 | return t && t->type == tt && t->content == expected; | |
540 | } | |
541 | ||
542 | ||
dff50e09 | 543 | const token* |
d7f3e0c5 GH |
544 | parser::expect_known (token_type tt, string const & expected) |
545 | { | |
546 | const token *t = next(); | |
57b73400 | 547 | if (! (t && t->type == tt && t->content == expected)) |
d7f3e0c5 GH |
548 | throw parse_error ("expected '" + expected + "'"); |
549 | return t; | |
550 | } | |
551 | ||
552 | ||
dff50e09 | 553 | const token* |
d7f3e0c5 GH |
554 | parser::expect_unknown (token_type tt, string & target) |
555 | { | |
556 | const token *t = next(); | |
557 | if (!(t && t->type == tt)) | |
558 | throw parse_error ("expected " + tt2str(tt)); | |
559 | target = t->content; | |
560 | return t; | |
561 | } | |
562 | ||
563 | ||
dff50e09 | 564 | const token* |
493ee224 DS |
565 | parser::expect_unknown2 (token_type tt1, token_type tt2, string & target) |
566 | { | |
567 | const token *t = next(); | |
568 | if (!(t && (t->type == tt1 || t->type == tt2))) | |
569 | throw parse_error ("expected " + tt2str(tt1) + " or " + tt2str(tt2)); | |
570 | target = t->content; | |
571 | return t; | |
572 | } | |
573 | ||
574 | ||
dff50e09 | 575 | const token* |
d7f3e0c5 GH |
576 | parser::expect_op (std::string const & expected) |
577 | { | |
578 | return expect_known (tok_operator, expected); | |
579 | } | |
580 | ||
581 | ||
dff50e09 | 582 | const token* |
d7f3e0c5 GH |
583 | parser::expect_kw (std::string const & expected) |
584 | { | |
585 | return expect_known (tok_identifier, expected); | |
586 | } | |
587 | ||
dff50e09 | 588 | const token* |
e38723d2 | 589 | parser::expect_number (int64_t & value) |
57b73400 | 590 | { |
e38723d2 MH |
591 | bool neg = false; |
592 | const token *t = next(); | |
593 | if (t->type == tok_operator && t->content == "-") | |
594 | { | |
595 | neg = true; | |
596 | t = next (); | |
597 | } | |
598 | if (!(t && t->type == tok_number)) | |
599 | throw parse_error ("expected number"); | |
600 | ||
601 | const char* startp = t->content.c_str (); | |
602 | char* endp = (char*) startp; | |
603 | ||
604 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
605 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
606 | // since the lexer only gives us positive digit strings, but we'll | |
607 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
608 | errno = 0; | |
609 | value = (int64_t) strtoull (startp, & endp, 0); | |
610 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' | |
611 | || (neg && (unsigned long long) value > 9223372036854775808ULL) | |
612 | || (unsigned long long) value > 18446744073709551615ULL | |
613 | || value < -9223372036854775807LL-1) | |
dff50e09 FCE |
614 | throw parse_error ("number invalid or out of range"); |
615 | ||
e38723d2 MH |
616 | if (neg) |
617 | value = -value; | |
618 | ||
619 | return t; | |
57b73400 GH |
620 | } |
621 | ||
d7f3e0c5 | 622 | |
dff50e09 | 623 | const token* |
d7f3e0c5 GH |
624 | parser::expect_ident (std::string & target) |
625 | { | |
626 | return expect_unknown (tok_identifier, target); | |
627 | } | |
628 | ||
629 | ||
dff50e09 | 630 | const token* |
493ee224 DS |
631 | parser::expect_ident_or_keyword (std::string & target) |
632 | { | |
633 | return expect_unknown2 (tok_identifier, tok_keyword, target); | |
634 | } | |
635 | ||
636 | ||
dff50e09 | 637 | bool |
d7f3e0c5 GH |
638 | parser::peek_op (std::string const & op) |
639 | { | |
640 | return tok_is (peek(), tok_operator, op); | |
641 | } | |
642 | ||
643 | ||
dff50e09 | 644 | bool |
d7f3e0c5 GH |
645 | parser::peek_kw (std::string const & kw) |
646 | { | |
647 | return tok_is (peek(), tok_identifier, kw); | |
648 | } | |
649 | ||
650 | ||
651 | ||
66c7d4c1 | 652 | lexer::lexer (istream& input, const string& in, systemtap_session& s): |
2203b032 | 653 | input_name (in), input_pointer (0), input_end (0), |
9300f661 JS |
654 | cursor_suspend_count(0), cursor_suspend_line (1), cursor_suspend_column (1), |
655 | cursor_line (1), cursor_column (1), | |
66c7d4c1 | 656 | session(s), current_file (0) |
eacb10ce | 657 | { |
66c7d4c1 | 658 | getline(input, input_contents, '\0'); |
2203b032 | 659 | |
66c7d4c1 JS |
660 | input_pointer = input_contents.data(); |
661 | input_end = input_contents.data() + input_contents.size(); | |
662 | ||
663 | if (keywords.empty()) | |
664 | { | |
665 | keywords.insert("probe"); | |
666 | keywords.insert("global"); | |
667 | keywords.insert("function"); | |
668 | keywords.insert("if"); | |
669 | keywords.insert("else"); | |
670 | keywords.insert("for"); | |
671 | keywords.insert("foreach"); | |
672 | keywords.insert("in"); | |
673 | keywords.insert("limit"); | |
674 | keywords.insert("return"); | |
675 | keywords.insert("delete"); | |
676 | keywords.insert("while"); | |
677 | keywords.insert("break"); | |
678 | keywords.insert("continue"); | |
679 | keywords.insert("next"); | |
680 | keywords.insert("string"); | |
681 | keywords.insert("long"); | |
682 | } | |
eacb10ce | 683 | } |
2f1a1aea | 684 | |
66c7d4c1 JS |
685 | set<string> lexer::keywords; |
686 | ||
1b1b4ceb RA |
687 | void |
688 | lexer::set_current_file (stapfile* f) | |
689 | { | |
690 | current_file = f; | |
2203b032 JS |
691 | if (f) |
692 | { | |
693 | f->file_contents = input_contents; | |
694 | f->name = input_name; | |
695 | } | |
1b1b4ceb | 696 | } |
bb2e3076 FCE |
697 | |
698 | int | |
699 | lexer::input_peek (unsigned n) | |
700 | { | |
66c7d4c1 JS |
701 | if (input_pointer + n >= input_end) |
702 | return -1; // EOF | |
703 | return (unsigned char)*(input_pointer + n); | |
bb2e3076 FCE |
704 | } |
705 | ||
706 | ||
dff50e09 | 707 | int |
2f1a1aea FCE |
708 | lexer::input_get () |
709 | { | |
66c7d4c1 | 710 | int c = input_peek(); |
bb2e3076 FCE |
711 | if (c < 0) return c; // EOF |
712 | ||
66c7d4c1 JS |
713 | ++input_pointer; |
714 | ||
3f99432c | 715 | if (cursor_suspend_count) |
9300f661 JS |
716 | { |
717 | // Track effect of input_put: preserve previous cursor/line_column | |
718 | // until all of its characters are consumed. | |
719 | if (--cursor_suspend_count == 0) | |
720 | { | |
721 | cursor_line = cursor_suspend_line; | |
722 | cursor_column = cursor_suspend_column; | |
723 | } | |
724 | } | |
3f99432c | 725 | else |
2f1a1aea | 726 | { |
3f99432c FCE |
727 | // update source cursor |
728 | if (c == '\n') | |
729 | { | |
730 | cursor_line ++; | |
731 | cursor_column = 1; | |
732 | } | |
733 | else | |
734 | cursor_column ++; | |
2f1a1aea | 735 | } |
2f1a1aea | 736 | |
eacb10ce | 737 | // clog << "[" << (char)c << "]"; |
2f1a1aea FCE |
738 | return c; |
739 | } | |
740 | ||
741 | ||
3f99432c | 742 | void |
9300f661 | 743 | lexer::input_put (const string& chars, const token* t) |
3f99432c | 744 | { |
66c7d4c1 JS |
745 | size_t pos = input_pointer - input_contents.data(); |
746 | // clog << "[put:" << chars << " @" << pos << "]"; | |
747 | input_contents.insert (pos, chars); | |
eacb10ce | 748 | cursor_suspend_count += chars.size(); |
9300f661 JS |
749 | cursor_suspend_line = cursor_line; |
750 | cursor_suspend_column = cursor_column; | |
751 | cursor_line = t->location.line; | |
752 | cursor_column = t->location.column; | |
66c7d4c1 JS |
753 | input_pointer = input_contents.data() + pos; |
754 | input_end = input_contents.data() + input_contents.size(); | |
3f99432c FCE |
755 | } |
756 | ||
757 | ||
2f1a1aea | 758 | token* |
3f847830 | 759 | lexer::scan (bool wildcard) |
2f1a1aea FCE |
760 | { |
761 | token* n = new token; | |
2203b032 | 762 | n->location.file = current_file; |
2f1a1aea | 763 | |
9300f661 JS |
764 | skip: |
765 | bool suspended = (cursor_suspend_count > 0); | |
2f1a1aea FCE |
766 | n->location.line = cursor_line; |
767 | n->location.column = cursor_column; | |
768 | ||
769 | int c = input_get(); | |
3f99432c | 770 | // clog << "{" << (char)c << (char)c2 << "}"; |
2f1a1aea FCE |
771 | if (c < 0) |
772 | { | |
773 | delete n; | |
774 | return 0; | |
775 | } | |
776 | ||
777 | if (isspace (c)) | |
778 | goto skip; | |
779 | ||
66c7d4c1 JS |
780 | int c2 = input_peek (); |
781 | ||
3f99432c FCE |
782 | // Paste command line arguments as character streams into |
783 | // the beginning of a token. $1..$999 go through as raw | |
784 | // characters; @1..@999 are quoted/escaped as strings. | |
785 | // $# and @# expand to the number of arguments, similarly | |
786 | // raw or quoted. | |
9300f661 | 787 | if ((c == '$' || c == '@') && (c2 == '#')) |
3f99432c | 788 | { |
9300f661 JS |
789 | n->content.push_back (c); |
790 | n->content.push_back (c2); | |
3f99432c | 791 | input_get(); // swallow '#' |
9300f661 JS |
792 | if (suspended) |
793 | throw parse_error ("invalid nested substitution of command line arguments", n); | |
794 | size_t num_args = session.args.size (); | |
795 | input_put ((c == '$') ? lex_cast (num_args) : lex_cast_qstring (num_args), n); | |
796 | n->content.clear(); | |
797 | goto skip; | |
3f99432c | 798 | } |
9300f661 | 799 | else if ((c == '$' || c == '@') && (isdigit (c2))) |
3f99432c | 800 | { |
9300f661 | 801 | n->content.push_back (c); |
3f99432c FCE |
802 | unsigned idx = 0; |
803 | do | |
804 | { | |
805 | input_get (); | |
806 | idx = (idx * 10) + (c2 - '0'); | |
9300f661 | 807 | n->content.push_back (c2); |
3f99432c FCE |
808 | c2 = input_peek (); |
809 | } while (c2 > 0 && | |
dff50e09 | 810 | isdigit (c2) && |
3f99432c | 811 | idx <= session.args.size()); // prevent overflow |
9300f661 JS |
812 | if (suspended) |
813 | throw parse_error ("invalid nested substitution of command line arguments", n); | |
3f99432c FCE |
814 | if (idx == 0 || |
815 | idx-1 >= session.args.size()) | |
aca66a36 JS |
816 | throw parse_error ("command line argument index " + lex_cast(idx) |
817 | + " out of range [1-" + lex_cast(session.args.size()) + "]", n); | |
9300f661 JS |
818 | const string& arg = session.args[idx-1]; |
819 | input_put ((c == '$') ? arg : lex_cast_qstring (arg), n); | |
820 | n->content.clear(); | |
821 | goto skip; | |
3f99432c FCE |
822 | } |
823 | ||
0c218afb MH |
824 | else if (isalpha (c) || c == '$' || c == '@' || c == '_' || |
825 | (wildcard && c == '*')) | |
2f1a1aea FCE |
826 | { |
827 | n->type = tok_identifier; | |
828 | n->content = (char) c; | |
0c218afb MH |
829 | while (isalnum (c2) || c2 == '_' || c2 == '$' || |
830 | (wildcard && c2 == '*')) | |
2f1a1aea | 831 | { |
3f99432c FCE |
832 | input_get (); |
833 | n->content.push_back (c2); | |
834 | c2 = input_peek (); | |
6e213f58 | 835 | } |
213bee8f | 836 | |
66c7d4c1 | 837 | if (keywords.count(n->content)) |
3f99432c | 838 | n->type = tok_keyword; |
dff50e09 | 839 | |
2f1a1aea FCE |
840 | return n; |
841 | } | |
842 | ||
3a20432b | 843 | else if (isdigit (c)) // positive literal |
2f1a1aea | 844 | { |
2f1a1aea | 845 | n->type = tok_number; |
9c0c0e46 FCE |
846 | n->content = (char) c; |
847 | ||
66c7d4c1 | 848 | while (isalnum (c2)) |
2f1a1aea | 849 | { |
9c0c0e46 FCE |
850 | // NB: isalnum is very permissive. We rely on strtol, called in |
851 | // parser::parse_literal below, to confirm that the number string | |
852 | // is correctly formatted and in range. | |
853 | ||
66c7d4c1 JS |
854 | input_get (); |
855 | n->content.push_back (c2); | |
856 | c2 = input_peek (); | |
2f1a1aea FCE |
857 | } |
858 | return n; | |
859 | } | |
860 | ||
861 | else if (c == '\"') | |
862 | { | |
863 | n->type = tok_string; | |
864 | while (1) | |
865 | { | |
866 | c = input_get (); | |
867 | ||
3f99432c | 868 | if (c < 0 || c == '\n') |
2f1a1aea FCE |
869 | { |
870 | n->type = tok_junk; | |
871 | break; | |
872 | } | |
873 | if (c == '\"') // closing double-quotes | |
874 | break; | |
3f99432c | 875 | else if (c == '\\') // see also input_put |
dff50e09 | 876 | { |
7d46afb8 GH |
877 | c = input_get (); |
878 | switch (c) | |
879 | { | |
880 | case 'a': | |
881 | case 'b': | |
882 | case 't': | |
883 | case 'n': | |
884 | case 'v': | |
885 | case 'f': | |
886 | case 'r': | |
f03954fd | 887 | case '0' ... '7': // NB: need only match the first digit |
7d46afb8 | 888 | case '\\': |
7d46afb8 | 889 | // Pass these escapes through to the string value |
dff50e09 | 890 | // being parsed; it will be emitted into a C literal. |
7d46afb8 GH |
891 | |
892 | n->content.push_back('\\'); | |
893 | ||
3f99432c | 894 | // fall through |
7d46afb8 | 895 | default: |
7d46afb8 GH |
896 | n->content.push_back(c); |
897 | break; | |
898 | } | |
2f1a1aea FCE |
899 | } |
900 | else | |
901 | n->content.push_back(c); | |
902 | } | |
903 | return n; | |
904 | } | |
905 | ||
906 | else if (ispunct (c)) | |
907 | { | |
bb2e3076 | 908 | int c3 = input_peek (1); |
2f1a1aea | 909 | |
3a20432b FCE |
910 | // NB: if we were to recognize negative numeric literals here, |
911 | // we'd introduce another grammar ambiguity: | |
912 | // 1-1 would be parsed as tok_number(1) and tok_number(-1) | |
913 | // instead of tok_number(1) tok_operator('-') tok_number(1) | |
914 | ||
66c7d4c1 | 915 | if (c == '#') // shell comment |
2f1a1aea FCE |
916 | { |
917 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
918 | do { c = input_get (); } |
919 | while (c >= 0 && cursor_line == this_line); | |
2f1a1aea FCE |
920 | goto skip; |
921 | } | |
66c7d4c1 | 922 | else if ((c == '/' && c2 == '/')) // C++ comment |
63a7c90e FCE |
923 | { |
924 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
925 | do { c = input_get (); } |
926 | while (c >= 0 && cursor_line == this_line); | |
63a7c90e FCE |
927 | goto skip; |
928 | } | |
929 | else if (c == '/' && c2 == '*') // C comment | |
930 | { | |
66c7d4c1 JS |
931 | (void) input_get (); // swallow '*' already in c2 |
932 | c = input_get (); | |
63a7c90e | 933 | c2 = input_get (); |
bb2e3076 | 934 | while (c2 >= 0) |
63a7c90e | 935 | { |
66c7d4c1 JS |
936 | if (c == '*' && c2 == '/') |
937 | break; | |
63a7c90e FCE |
938 | c = c2; |
939 | c2 = input_get (); | |
63a7c90e | 940 | } |
bb2e3076 | 941 | goto skip; |
63a7c90e | 942 | } |
54dfabe9 FCE |
943 | else if (c == '%' && c2 == '{') // embedded code |
944 | { | |
945 | n->type = tok_embedded; | |
946 | (void) input_get (); // swallow '{' already in c2 | |
66c7d4c1 JS |
947 | c = input_get (); |
948 | c2 = input_get (); | |
949 | while (c2 >= 0) | |
54dfabe9 | 950 | { |
66c7d4c1 JS |
951 | if (c == '%' && c2 == '}') |
952 | return n; | |
54dfabe9 | 953 | n->content += c; |
66c7d4c1 JS |
954 | c = c2; |
955 | c2 = input_get (); | |
54dfabe9 | 956 | } |
66c7d4c1 | 957 | n->type = tok_junk; |
54dfabe9 FCE |
958 | return n; |
959 | } | |
2f1a1aea | 960 | |
bb2e3076 FCE |
961 | // We're committed to recognizing at least the first character |
962 | // as an operator. | |
2f1a1aea | 963 | n->type = tok_operator; |
66c7d4c1 | 964 | n->content = c; |
2f1a1aea | 965 | |
bb2e3076 | 966 | // match all valid operators, in decreasing size order |
66c7d4c1 JS |
967 | if ((c == '<' && c2 == '<' && c3 == '<') || |
968 | (c == '<' && c2 == '<' && c3 == '=') || | |
969 | (c == '>' && c2 == '>' && c3 == '=')) | |
82919855 | 970 | { |
66c7d4c1 JS |
971 | n->content += c2; |
972 | n->content += c3; | |
bb2e3076 FCE |
973 | input_get (); input_get (); // swallow other two characters |
974 | } | |
66c7d4c1 JS |
975 | else if ((c == '=' && c2 == '=') || |
976 | (c == '!' && c2 == '=') || | |
977 | (c == '<' && c2 == '=') || | |
978 | (c == '>' && c2 == '=') || | |
979 | (c == '+' && c2 == '=') || | |
980 | (c == '-' && c2 == '=') || | |
981 | (c == '*' && c2 == '=') || | |
982 | (c == '/' && c2 == '=') || | |
983 | (c == '%' && c2 == '=') || | |
984 | (c == '&' && c2 == '=') || | |
985 | (c == '^' && c2 == '=') || | |
986 | (c == '|' && c2 == '=') || | |
987 | (c == '.' && c2 == '=') || | |
988 | (c == '&' && c2 == '&') || | |
989 | (c == '|' && c2 == '|') || | |
990 | (c == '+' && c2 == '+') || | |
991 | (c == '-' && c2 == '-') || | |
992 | (c == '-' && c2 == '>') || | |
993 | (c == '<' && c2 == '<') || | |
994 | (c == '>' && c2 == '>') || | |
177a8ead | 995 | // preprocessor tokens |
66c7d4c1 JS |
996 | (c == '%' && c2 == '(') || |
997 | (c == '%' && c2 == '?') || | |
998 | (c == '%' && c2 == ':') || | |
999 | (c == '%' && c2 == ')')) | |
bb2e3076 | 1000 | { |
66c7d4c1 | 1001 | n->content += c2; |
bb2e3076 | 1002 | input_get (); // swallow other character |
dff50e09 | 1003 | } |
2f1a1aea FCE |
1004 | |
1005 | return n; | |
1006 | } | |
1007 | ||
1008 | else | |
1009 | { | |
1010 | n->type = tok_junk; | |
1011 | n->content = (char) c; | |
1012 | return n; | |
1013 | } | |
1014 | } | |
1015 | ||
1016 | ||
1017 | // ------------------------------------------------------------------------ | |
1018 | ||
1019 | stapfile* | |
1020 | parser::parse () | |
1021 | { | |
1022 | stapfile* f = new stapfile; | |
1b1b4ceb | 1023 | input.set_current_file (f); |
56099f08 FCE |
1024 | |
1025 | bool empty = true; | |
1026 | ||
2f1a1aea FCE |
1027 | while (1) |
1028 | { | |
1029 | try | |
1030 | { | |
1031 | const token* t = peek (); | |
56099f08 | 1032 | if (! t) // nice clean EOF |
2f1a1aea FCE |
1033 | break; |
1034 | ||
56099f08 | 1035 | empty = false; |
6e213f58 DS |
1036 | if (t->type == tok_keyword && t->content == "probe") |
1037 | { | |
1038 | context = con_probe; | |
1039 | parse_probe (f->probes, f->aliases); | |
1040 | } | |
1041 | else if (t->type == tok_keyword && t->content == "global") | |
1042 | { | |
1043 | context = con_global; | |
4b5f3e45 | 1044 | parse_global (f->globals, f->probes); |
6e213f58 DS |
1045 | } |
1046 | else if (t->type == tok_keyword && t->content == "function") | |
1047 | { | |
1048 | context = con_function; | |
1049 | parse_functiondecl (f->functions); | |
1050 | } | |
54dfabe9 | 1051 | else if (t->type == tok_embedded) |
6e213f58 DS |
1052 | { |
1053 | context = con_embedded; | |
1054 | f->embeds.push_back (parse_embeddedcode ()); | |
1055 | } | |
2f1a1aea | 1056 | else |
6e213f58 DS |
1057 | { |
1058 | context = con_unknown; | |
1059 | throw parse_error ("expected 'probe', 'global', 'function', or '%{'"); | |
1060 | } | |
2f1a1aea FCE |
1061 | } |
1062 | catch (parse_error& pe) | |
1063 | { | |
1064 | print_error (pe); | |
cd7116b8 | 1065 | if (pe.skip_some) // for recovery |
dff50e09 | 1066 | try |
cd7116b8 FCE |
1067 | { |
1068 | // Quietly swallow all tokens until the next '}'. | |
1069 | while (1) | |
1070 | { | |
1071 | const token* t = peek (); | |
1072 | if (! t) | |
1073 | break; | |
1074 | next (); | |
1075 | if (t->type == tok_operator && t->content == "}") | |
1076 | break; | |
1077 | } | |
1078 | } | |
1079 | catch (parse_error& pe2) | |
1080 | { | |
1081 | // parse error during recovery ... ugh | |
1082 | print_error (pe2); | |
1083 | } | |
177a8ead | 1084 | } |
2f1a1aea FCE |
1085 | } |
1086 | ||
56099f08 FCE |
1087 | if (empty) |
1088 | { | |
1089 | cerr << "Input file '" << input_name << "' is empty or missing." << endl; | |
1090 | delete f; | |
2203b032 | 1091 | f = 0; |
56099f08 FCE |
1092 | } |
1093 | else if (num_errors > 0) | |
2f1a1aea FCE |
1094 | { |
1095 | cerr << num_errors << " parse error(s)." << endl; | |
1096 | delete f; | |
2203b032 | 1097 | f = 0; |
2f1a1aea | 1098 | } |
dff50e09 | 1099 | |
2203b032 | 1100 | input.set_current_file(0); |
2f1a1aea FCE |
1101 | return f; |
1102 | } | |
1103 | ||
1104 | ||
20c6c071 | 1105 | void |
54dfabe9 FCE |
1106 | parser::parse_probe (std::vector<probe *> & probe_ret, |
1107 | std::vector<probe_alias *> & alias_ret) | |
2f1a1aea | 1108 | { |
82919855 | 1109 | const token* t0 = next (); |
6e213f58 | 1110 | if (! (t0->type == tok_keyword && t0->content == "probe")) |
82919855 FCE |
1111 | throw parse_error ("expected 'probe'"); |
1112 | ||
20c6c071 GH |
1113 | vector<probe_point *> aliases; |
1114 | vector<probe_point *> locations; | |
1115 | ||
1116 | bool equals_ok = true; | |
82919855 | 1117 | |
97266278 LG |
1118 | int epilogue_alias = 0; |
1119 | ||
2f1a1aea FCE |
1120 | while (1) |
1121 | { | |
b4ceace2 | 1122 | probe_point * pp = parse_probe_point (); |
dff50e09 | 1123 | |
b4ceace2 | 1124 | const token* t = peek (); |
dff50e09 | 1125 | if (equals_ok && t |
b4ceace2 FCE |
1126 | && t->type == tok_operator && t->content == "=") |
1127 | { | |
1ad820e3 | 1128 | if (pp->optional || pp->sufficient) |
f1a0157a | 1129 | throw parse_error ("probe point alias name cannot be optional nor sufficient", pp->components.front()->tok); |
b4ceace2 FCE |
1130 | aliases.push_back(pp); |
1131 | next (); | |
1132 | continue; | |
1133 | } | |
dff50e09 | 1134 | else if (equals_ok && t |
97266278 LG |
1135 | && t->type == tok_operator && t->content == "+=") |
1136 | { | |
1ad820e3 | 1137 | if (pp->optional || pp->sufficient) |
f1a0157a | 1138 | throw parse_error ("probe point alias name cannot be optional nor sufficient", pp->components.front()->tok); |
97266278 LG |
1139 | aliases.push_back(pp); |
1140 | epilogue_alias = 1; | |
1141 | next (); | |
1142 | continue; | |
1143 | } | |
b4ceace2 FCE |
1144 | else if (t && t->type == tok_operator && t->content == ",") |
1145 | { | |
1146 | locations.push_back(pp); | |
1147 | equals_ok = false; | |
1148 | next (); | |
1149 | continue; | |
1150 | } | |
1151 | else if (t && t->type == tok_operator && t->content == "{") | |
1152 | { | |
1153 | locations.push_back(pp); | |
1154 | break; | |
1155 | } | |
2f1a1aea | 1156 | else |
9c0c0e46 | 1157 | throw parse_error ("expected probe point specifier"); |
2f1a1aea | 1158 | } |
20c6c071 | 1159 | |
20c6c071 GH |
1160 | if (aliases.empty()) |
1161 | { | |
54dfabe9 FCE |
1162 | probe* p = new probe; |
1163 | p->tok = t0; | |
1164 | p->locations = locations; | |
1165 | p->body = parse_stmt_block (); | |
37ebca01 | 1166 | p->privileged = privileged; |
54dfabe9 | 1167 | probe_ret.push_back (p); |
20c6c071 GH |
1168 | } |
1169 | else | |
1170 | { | |
54dfabe9 | 1171 | probe_alias* p = new probe_alias (aliases); |
97266278 LG |
1172 | if(epilogue_alias) |
1173 | p->epilogue_style = true; | |
1174 | else | |
1175 | p->epilogue_style = false; | |
54dfabe9 FCE |
1176 | p->tok = t0; |
1177 | p->locations = locations; | |
1178 | p->body = parse_stmt_block (); | |
37ebca01 | 1179 | p->privileged = privileged; |
54dfabe9 | 1180 | alias_ret.push_back (p); |
20c6c071 | 1181 | } |
54dfabe9 | 1182 | } |
20c6c071 | 1183 | |
54dfabe9 FCE |
1184 | |
1185 | embeddedcode* | |
1186 | parser::parse_embeddedcode () | |
1187 | { | |
1188 | embeddedcode* e = new embeddedcode; | |
1189 | const token* t = next (); | |
1190 | if (t->type != tok_embedded) | |
24cb178f FCE |
1191 | throw parse_error ("expected '%{'"); |
1192 | ||
1193 | if (! privileged) | |
cd7116b8 FCE |
1194 | throw parse_error ("embedded code in unprivileged script", |
1195 | false /* don't skip tokens for parse resumption */); | |
54dfabe9 FCE |
1196 | |
1197 | e->tok = t; | |
1198 | e->code = t->content; | |
1199 | return e; | |
2f1a1aea FCE |
1200 | } |
1201 | ||
1202 | ||
1203 | block* | |
56099f08 | 1204 | parser::parse_stmt_block () |
2f1a1aea FCE |
1205 | { |
1206 | block* pb = new block; | |
1207 | ||
56099f08 FCE |
1208 | const token* t = next (); |
1209 | if (! (t->type == tok_operator && t->content == "{")) | |
1210 | throw parse_error ("expected '{'"); | |
1211 | ||
1212 | pb->tok = t; | |
2b066ec1 | 1213 | |
2f1a1aea FCE |
1214 | while (1) |
1215 | { | |
1216 | try | |
1217 | { | |
2b066ec1 FCE |
1218 | t = peek (); |
1219 | if (t && t->type == tok_operator && t->content == "}") | |
1220 | { | |
1221 | next (); | |
1222 | break; | |
1223 | } | |
1224 | ||
2f1a1aea | 1225 | pb->statements.push_back (parse_statement ()); |
2f1a1aea FCE |
1226 | } |
1227 | catch (parse_error& pe) | |
1228 | { | |
1229 | print_error (pe); | |
54dfabe9 | 1230 | |
2f1a1aea FCE |
1231 | // Quietly swallow all tokens until the next ';' or '}'. |
1232 | while (1) | |
1233 | { | |
1234 | const token* t = peek (); | |
54dfabe9 | 1235 | if (! t) return 0; |
2f1a1aea | 1236 | next (); |
54dfabe9 FCE |
1237 | if (t->type == tok_operator |
1238 | && (t->content == "}" || t->content == ";")) | |
2f1a1aea FCE |
1239 | break; |
1240 | } | |
1241 | } | |
1242 | } | |
1243 | ||
1244 | return pb; | |
1245 | } | |
1246 | ||
1247 | ||
1248 | statement* | |
1249 | parser::parse_statement () | |
1250 | { | |
40b71c47 | 1251 | statement *ret; |
2f1a1aea FCE |
1252 | const token* t = peek (); |
1253 | if (t && t->type == tok_operator && t->content == ";") | |
1254 | { | |
69c68955 FCE |
1255 | null_statement* n = new null_statement (); |
1256 | n->tok = next (); | |
1257 | return n; | |
2f1a1aea | 1258 | } |
dff50e09 | 1259 | else if (t && t->type == tok_operator && t->content == "{") |
40b71c47 | 1260 | return parse_stmt_block (); // Don't squash semicolons. |
6e213f58 | 1261 | else if (t && t->type == tok_keyword && t->content == "if") |
40b71c47 | 1262 | return parse_if_statement (); // Don't squash semicolons. |
6e213f58 | 1263 | else if (t && t->type == tok_keyword && t->content == "for") |
40b71c47 | 1264 | return parse_for_loop (); // Don't squash semicolons. |
6e213f58 | 1265 | else if (t && t->type == tok_keyword && t->content == "foreach") |
40b71c47 MW |
1266 | return parse_foreach_loop (); // Don't squash semicolons. |
1267 | else if (t && t->type == tok_keyword && t->content == "while") | |
1268 | return parse_while_loop (); // Don't squash semicolons. | |
6e213f58 | 1269 | else if (t && t->type == tok_keyword && t->content == "return") |
40b71c47 | 1270 | ret = parse_return_statement (); |
6e213f58 | 1271 | else if (t && t->type == tok_keyword && t->content == "delete") |
40b71c47 | 1272 | ret = parse_delete_statement (); |
6e213f58 | 1273 | else if (t && t->type == tok_keyword && t->content == "break") |
40b71c47 | 1274 | ret = parse_break_statement (); |
6e213f58 | 1275 | else if (t && t->type == tok_keyword && t->content == "continue") |
40b71c47 | 1276 | ret = parse_continue_statement (); |
6e213f58 | 1277 | else if (t && t->type == tok_keyword && t->content == "next") |
40b71c47 | 1278 | ret = parse_next_statement (); |
2f1a1aea FCE |
1279 | else if (t && (t->type == tok_operator || // expressions are flexible |
1280 | t->type == tok_identifier || | |
1281 | t->type == tok_number || | |
1282 | t->type == tok_string)) | |
40b71c47 | 1283 | ret = parse_expr_statement (); |
54dfabe9 | 1284 | // XXX: consider generally accepting tok_embedded here too |
2f1a1aea FCE |
1285 | else |
1286 | throw parse_error ("expected statement"); | |
40b71c47 MW |
1287 | |
1288 | // Squash "empty" trailing colons after any "non-block-like" statement. | |
1289 | t = peek (); | |
1290 | if (t && t->type == tok_operator && t->content == ";") | |
1291 | { | |
1292 | next (); // Silently eat trailing ; after statement | |
1293 | } | |
1294 | ||
1295 | return ret; | |
2f1a1aea FCE |
1296 | } |
1297 | ||
1298 | ||
56099f08 | 1299 | void |
78f6bba6 | 1300 | parser::parse_global (vector <vardecl*>& globals, vector<probe*>&) |
2f1a1aea | 1301 | { |
82919855 | 1302 | const token* t0 = next (); |
6e213f58 | 1303 | if (! (t0->type == tok_keyword && t0->content == "global")) |
82919855 FCE |
1304 | throw parse_error ("expected 'global'"); |
1305 | ||
56099f08 FCE |
1306 | while (1) |
1307 | { | |
1308 | const token* t = next (); | |
1309 | if (! (t->type == tok_identifier)) | |
1310 | throw parse_error ("expected identifier"); | |
1311 | ||
2b066ec1 FCE |
1312 | for (unsigned i=0; i<globals.size(); i++) |
1313 | if (globals[i]->name == t->content) | |
57b73400 | 1314 | throw parse_error ("duplicate global name"); |
dff50e09 | 1315 | |
24cb178f FCE |
1316 | vardecl* d = new vardecl; |
1317 | d->name = t->content; | |
1318 | d->tok = t; | |
1319 | globals.push_back (d); | |
56099f08 | 1320 | |
82919855 | 1321 | t = peek (); |
ef474d24 JS |
1322 | |
1323 | if (t && t->type == tok_operator && t->content == "[") // array size | |
1324 | { | |
1325 | int64_t size; | |
1326 | next (); | |
1327 | expect_number(size); | |
1328 | if (size <= 0 || size > 1000000) // arbitrary max | |
1329 | throw parse_error("array size out of range"); | |
1330 | d->maxsize = (int)size; | |
1331 | expect_known(tok_operator, "]"); | |
1332 | t = peek (); | |
1333 | } | |
1334 | ||
4b5f3e45 | 1335 | if (t && t->type == tok_operator && t->content == "=") // initialization |
ef474d24 JS |
1336 | { |
1337 | if (!d->compatible_arity(0)) | |
1338 | throw parse_error("only scalar globals can be initialized"); | |
1339 | d->set_arity(0); | |
1340 | next (); | |
1341 | d->init = parse_literal (); | |
1342 | d->type = d->init->type; | |
1343 | t = peek (); | |
1344 | } | |
4b5f3e45 | 1345 | |
c3799d72 AM |
1346 | if (t && t->type == tok_operator && t->content == ";") // termination |
1347 | next(); | |
1348 | ||
4b5f3e45 | 1349 | if (t && t->type == tok_operator && t->content == ",") // next global |
82919855 FCE |
1350 | { |
1351 | next (); | |
1352 | continue; | |
1353 | } | |
56099f08 | 1354 | else |
82919855 | 1355 | break; |
56099f08 FCE |
1356 | } |
1357 | } | |
1358 | ||
1359 | ||
24cb178f FCE |
1360 | void |
1361 | parser::parse_functiondecl (std::vector<functiondecl*>& functions) | |
56099f08 | 1362 | { |
82919855 | 1363 | const token* t = next (); |
6e213f58 | 1364 | if (! (t->type == tok_keyword && t->content == "function")) |
82919855 FCE |
1365 | throw parse_error ("expected 'function'"); |
1366 | ||
56099f08 | 1367 | |
82919855 | 1368 | t = next (); |
6e213f58 DS |
1369 | if (! (t->type == tok_identifier) |
1370 | && ! (t->type == tok_keyword | |
1371 | && (t->content == "string" || t->content == "long"))) | |
56099f08 | 1372 | throw parse_error ("expected identifier"); |
24cb178f FCE |
1373 | |
1374 | for (unsigned i=0; i<functions.size(); i++) | |
1375 | if (functions[i]->name == t->content) | |
1376 | throw parse_error ("duplicate function name"); | |
1377 | ||
1378 | functiondecl *fd = new functiondecl (); | |
56099f08 FCE |
1379 | fd->name = t->content; |
1380 | fd->tok = t; | |
1381 | ||
1382 | t = next (); | |
6a505121 FCE |
1383 | if (t->type == tok_operator && t->content == ":") |
1384 | { | |
1385 | t = next (); | |
6e213f58 | 1386 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 1387 | fd->type = pe_string; |
6e213f58 | 1388 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 FCE |
1389 | fd->type = pe_long; |
1390 | else throw parse_error ("expected 'string' or 'long'"); | |
1391 | ||
1392 | t = next (); | |
1393 | } | |
1394 | ||
56099f08 FCE |
1395 | if (! (t->type == tok_operator && t->content == "(")) |
1396 | throw parse_error ("expected '('"); | |
1397 | ||
1398 | while (1) | |
1399 | { | |
1400 | t = next (); | |
1401 | ||
1402 | // permit zero-argument fuctions | |
1403 | if (t->type == tok_operator && t->content == ")") | |
1404 | break; | |
1405 | else if (! (t->type == tok_identifier)) | |
1406 | throw parse_error ("expected identifier"); | |
1407 | vardecl* vd = new vardecl; | |
1408 | vd->name = t->content; | |
1409 | vd->tok = t; | |
1410 | fd->formal_args.push_back (vd); | |
1411 | ||
1412 | t = next (); | |
6a505121 FCE |
1413 | if (t->type == tok_operator && t->content == ":") |
1414 | { | |
1415 | t = next (); | |
6e213f58 | 1416 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 1417 | vd->type = pe_string; |
6e213f58 | 1418 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 FCE |
1419 | vd->type = pe_long; |
1420 | else throw parse_error ("expected 'string' or 'long'"); | |
dff50e09 | 1421 | |
6a505121 FCE |
1422 | t = next (); |
1423 | } | |
56099f08 FCE |
1424 | if (t->type == tok_operator && t->content == ")") |
1425 | break; | |
1426 | if (t->type == tok_operator && t->content == ",") | |
1427 | continue; | |
1428 | else | |
1429 | throw parse_error ("expected ',' or ')'"); | |
1430 | } | |
1431 | ||
54dfabe9 FCE |
1432 | t = peek (); |
1433 | if (t && t->type == tok_embedded) | |
1434 | fd->body = parse_embeddedcode (); | |
1435 | else | |
1436 | fd->body = parse_stmt_block (); | |
24cb178f FCE |
1437 | |
1438 | functions.push_back (fd); | |
2f1a1aea FCE |
1439 | } |
1440 | ||
1441 | ||
9c0c0e46 FCE |
1442 | probe_point* |
1443 | parser::parse_probe_point () | |
2f1a1aea | 1444 | { |
9c0c0e46 | 1445 | probe_point* pl = new probe_point; |
2f1a1aea | 1446 | |
9c0c0e46 | 1447 | while (1) |
2f1a1aea | 1448 | { |
0c218afb | 1449 | const token* t = next (true); // wildcard scanning here |
6e213f58 DS |
1450 | if (! (t->type == tok_identifier |
1451 | // we must allow ".return" and ".function", which are keywords | |
0c218afb | 1452 | || t->type == tok_keyword)) |
b4ceace2 | 1453 | throw parse_error ("expected identifier or '*'"); |
9c0c0e46 | 1454 | |
9c0c0e46 FCE |
1455 | |
1456 | probe_point::component* c = new probe_point::component; | |
1457 | c->functor = t->content; | |
f1a0157a | 1458 | c->tok = t; |
9c0c0e46 | 1459 | pl->components.push_back (c); |
6e3347a9 | 1460 | // NB we may add c->arg soon |
9c0c0e46 FCE |
1461 | |
1462 | t = peek (); | |
a477f3f1 | 1463 | |
6e3347a9 | 1464 | // consume optional parameter |
9c0c0e46 FCE |
1465 | if (t && t->type == tok_operator && t->content == "(") |
1466 | { | |
1467 | next (); // consume "(" | |
1468 | c->arg = parse_literal (); | |
1469 | ||
1470 | t = next (); | |
1471 | if (! (t->type == tok_operator && t->content == ")")) | |
1472 | throw parse_error ("expected ')'"); | |
1473 | ||
1474 | t = peek (); | |
9c0c0e46 | 1475 | } |
9c0c0e46 FCE |
1476 | |
1477 | if (t && t->type == tok_operator && t->content == ".") | |
6e3347a9 FCE |
1478 | { |
1479 | next (); | |
1480 | continue; | |
1481 | } | |
1482 | ||
f1a0157a | 1483 | // We only fall through here at the end of a probe point (past |
6e3347a9 FCE |
1484 | // all the dotted/parametrized components). |
1485 | ||
d898100a FCE |
1486 | if (t && t->type == tok_operator && |
1487 | (t->content == "?" || t->content == "!")) | |
6e3347a9 FCE |
1488 | { |
1489 | pl->optional = true; | |
d898100a FCE |
1490 | if (t->content == "!") pl->sufficient = true; |
1491 | // NB: sufficient implies optional | |
6e3347a9 FCE |
1492 | next (); |
1493 | t = peek (); | |
1494 | // fall through | |
cbbe8080 MH |
1495 | } |
1496 | ||
1497 | if (t && t->type == tok_keyword && t->content == "if") | |
1498 | { | |
1499 | next (); | |
1500 | t = peek (); | |
75686668 | 1501 | if (t && ! (t->type == tok_operator && t->content == "(")) |
cbbe8080 MH |
1502 | throw parse_error ("expected '('"); |
1503 | next (); | |
1504 | ||
1505 | pl->condition = parse_expression (); | |
1506 | ||
1507 | t = peek (); | |
75686668 | 1508 | if (t && ! (t->type == tok_operator && t->content == ")")) |
cbbe8080 MH |
1509 | throw parse_error ("expected ')'"); |
1510 | next (); | |
1511 | t = peek (); | |
1512 | // fall through | |
6e3347a9 FCE |
1513 | } |
1514 | ||
dff50e09 | 1515 | if (t && t->type == tok_operator |
6e3347a9 FCE |
1516 | && (t->content == "{" || t->content == "," || |
1517 | t->content == "=" || t->content == "+=" )) | |
1518 | break; | |
dff50e09 | 1519 | |
d898100a | 1520 | throw parse_error ("expected one of '. , ( ? ! { = +='"); |
2f1a1aea FCE |
1521 | } |
1522 | ||
1523 | return pl; | |
1524 | } | |
1525 | ||
1526 | ||
1527 | literal* | |
1528 | parser::parse_literal () | |
1529 | { | |
1530 | const token* t = next (); | |
56099f08 | 1531 | literal* l; |
2f1a1aea | 1532 | if (t->type == tok_string) |
56099f08 | 1533 | l = new literal_string (t->content); |
16e8f21f | 1534 | else |
9c0c0e46 | 1535 | { |
16e8f21f JS |
1536 | bool neg = false; |
1537 | if (t->type == tok_operator && t->content == "-") | |
1538 | { | |
1539 | neg = true; | |
1540 | t = next (); | |
1541 | } | |
1542 | ||
1543 | if (t->type == tok_number) | |
1544 | { | |
1545 | const char* startp = t->content.c_str (); | |
1546 | char* endp = (char*) startp; | |
1547 | ||
1548 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
1549 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
79e6d33f JS |
1550 | // since the lexer only gives us positive digit strings, but we'll |
1551 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
16e8f21f JS |
1552 | errno = 0; |
1553 | long long value = (long long) strtoull (startp, & endp, 0); | |
16e8f21f | 1554 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' |
79e6d33f | 1555 | || (neg && (unsigned long long) value > 9223372036854775808ULL) |
16e8f21f JS |
1556 | || (unsigned long long) value > 18446744073709551615ULL |
1557 | || value < -9223372036854775807LL-1) | |
dff50e09 | 1558 | throw parse_error ("number invalid or out of range"); |
16e8f21f | 1559 | |
79e6d33f JS |
1560 | if (neg) |
1561 | value = -value; | |
1562 | ||
16e8f21f JS |
1563 | l = new literal_number (value); |
1564 | } | |
1565 | else | |
1566 | throw parse_error ("expected literal string or number"); | |
9c0c0e46 | 1567 | } |
56099f08 FCE |
1568 | |
1569 | l->tok = t; | |
1570 | return l; | |
2f1a1aea FCE |
1571 | } |
1572 | ||
1573 | ||
1574 | if_statement* | |
1575 | parser::parse_if_statement () | |
1576 | { | |
1577 | const token* t = next (); | |
6e213f58 | 1578 | if (! (t->type == tok_keyword && t->content == "if")) |
56099f08 FCE |
1579 | throw parse_error ("expected 'if'"); |
1580 | if_statement* s = new if_statement; | |
1581 | s->tok = t; | |
1582 | ||
1583 | t = next (); | |
2f1a1aea FCE |
1584 | if (! (t->type == tok_operator && t->content == "(")) |
1585 | throw parse_error ("expected '('"); | |
1586 | ||
2f1a1aea FCE |
1587 | s->condition = parse_expression (); |
1588 | ||
1589 | t = next (); | |
1590 | if (! (t->type == tok_operator && t->content == ")")) | |
1591 | throw parse_error ("expected ')'"); | |
1592 | ||
1593 | s->thenblock = parse_statement (); | |
1594 | ||
1595 | t = peek (); | |
6e213f58 | 1596 | if (t && t->type == tok_keyword && t->content == "else") |
2f1a1aea FCE |
1597 | { |
1598 | next (); | |
1599 | s->elseblock = parse_statement (); | |
1600 | } | |
ed10c639 FCE |
1601 | else |
1602 | s->elseblock = 0; // in case not otherwise initialized | |
2f1a1aea FCE |
1603 | |
1604 | return s; | |
1605 | } | |
1606 | ||
1607 | ||
69c68955 FCE |
1608 | expr_statement* |
1609 | parser::parse_expr_statement () | |
1610 | { | |
1611 | expr_statement *es = new expr_statement; | |
1612 | const token* t = peek (); | |
1613 | es->tok = t; | |
1614 | es->value = parse_expression (); | |
1615 | return es; | |
1616 | } | |
1617 | ||
1618 | ||
56099f08 FCE |
1619 | return_statement* |
1620 | parser::parse_return_statement () | |
1621 | { | |
1622 | const token* t = next (); | |
6e213f58 | 1623 | if (! (t->type == tok_keyword && t->content == "return")) |
56099f08 | 1624 | throw parse_error ("expected 'return'"); |
6e213f58 DS |
1625 | if (context != con_function) |
1626 | throw parse_error ("found 'return' not in function context"); | |
56099f08 FCE |
1627 | return_statement* s = new return_statement; |
1628 | s->tok = t; | |
1629 | s->value = parse_expression (); | |
1630 | return s; | |
1631 | } | |
1632 | ||
1633 | ||
1634 | delete_statement* | |
1635 | parser::parse_delete_statement () | |
1636 | { | |
1637 | const token* t = next (); | |
6e213f58 | 1638 | if (! (t->type == tok_keyword && t->content == "delete")) |
56099f08 FCE |
1639 | throw parse_error ("expected 'delete'"); |
1640 | delete_statement* s = new delete_statement; | |
1641 | s->tok = t; | |
1642 | s->value = parse_expression (); | |
1643 | return s; | |
1644 | } | |
1645 | ||
1646 | ||
f3c26ea5 FCE |
1647 | next_statement* |
1648 | parser::parse_next_statement () | |
1649 | { | |
1650 | const token* t = next (); | |
6e213f58 | 1651 | if (! (t->type == tok_keyword && t->content == "next")) |
f3c26ea5 | 1652 | throw parse_error ("expected 'next'"); |
6e213f58 DS |
1653 | if (context != con_probe) |
1654 | throw parse_error ("found 'next' not in probe context"); | |
f3c26ea5 FCE |
1655 | next_statement* s = new next_statement; |
1656 | s->tok = t; | |
1657 | return s; | |
1658 | } | |
1659 | ||
1660 | ||
1661 | break_statement* | |
1662 | parser::parse_break_statement () | |
1663 | { | |
1664 | const token* t = next (); | |
6e213f58 | 1665 | if (! (t->type == tok_keyword && t->content == "break")) |
f3c26ea5 FCE |
1666 | throw parse_error ("expected 'break'"); |
1667 | break_statement* s = new break_statement; | |
1668 | s->tok = t; | |
1669 | return s; | |
1670 | } | |
1671 | ||
1672 | ||
1673 | continue_statement* | |
1674 | parser::parse_continue_statement () | |
1675 | { | |
1676 | const token* t = next (); | |
6e213f58 | 1677 | if (! (t->type == tok_keyword && t->content == "continue")) |
f3c26ea5 FCE |
1678 | throw parse_error ("expected 'continue'"); |
1679 | continue_statement* s = new continue_statement; | |
1680 | s->tok = t; | |
1681 | return s; | |
1682 | } | |
1683 | ||
1684 | ||
69c68955 FCE |
1685 | for_loop* |
1686 | parser::parse_for_loop () | |
1687 | { | |
f3c26ea5 | 1688 | const token* t = next (); |
6e213f58 | 1689 | if (! (t->type == tok_keyword && t->content == "for")) |
f3c26ea5 FCE |
1690 | throw parse_error ("expected 'for'"); |
1691 | for_loop* s = new for_loop; | |
1692 | s->tok = t; | |
1693 | ||
1694 | t = next (); | |
1695 | if (! (t->type == tok_operator && t->content == "(")) | |
1696 | throw parse_error ("expected '('"); | |
1697 | ||
1698 | // initializer + ";" | |
1699 | t = peek (); | |
1700 | if (t && t->type == tok_operator && t->content == ";") | |
1701 | { | |
cbfbbf69 FCE |
1702 | s->init = 0; |
1703 | next (); | |
f3c26ea5 FCE |
1704 | } |
1705 | else | |
1706 | { | |
1707 | s->init = parse_expr_statement (); | |
1708 | t = next (); | |
1709 | if (! (t->type == tok_operator && t->content == ";")) | |
1710 | throw parse_error ("expected ';'"); | |
1711 | } | |
1712 | ||
1713 | // condition + ";" | |
1714 | t = peek (); | |
1715 | if (t && t->type == tok_operator && t->content == ";") | |
1716 | { | |
1717 | literal_number* l = new literal_number(1); | |
1718 | s->cond = l; | |
1719 | s->cond->tok = next (); | |
1720 | } | |
1721 | else | |
1722 | { | |
1723 | s->cond = parse_expression (); | |
1724 | t = next (); | |
1725 | if (! (t->type == tok_operator && t->content == ";")) | |
1726 | throw parse_error ("expected ';'"); | |
1727 | } | |
dff50e09 | 1728 | |
f3c26ea5 FCE |
1729 | // increment + ")" |
1730 | t = peek (); | |
1731 | if (t && t->type == tok_operator && t->content == ")") | |
1732 | { | |
cbfbbf69 FCE |
1733 | s->incr = 0; |
1734 | next (); | |
f3c26ea5 FCE |
1735 | } |
1736 | else | |
1737 | { | |
1738 | s->incr = parse_expr_statement (); | |
1739 | t = next (); | |
1740 | if (! (t->type == tok_operator && t->content == ")")) | |
c958a431 | 1741 | throw parse_error ("expected ')'"); |
f3c26ea5 FCE |
1742 | } |
1743 | ||
1744 | // block | |
1745 | s->block = parse_statement (); | |
1746 | ||
1747 | return s; | |
1748 | } | |
1749 | ||
1750 | ||
1751 | for_loop* | |
1752 | parser::parse_while_loop () | |
1753 | { | |
1754 | const token* t = next (); | |
6e213f58 | 1755 | if (! (t->type == tok_keyword && t->content == "while")) |
f3c26ea5 FCE |
1756 | throw parse_error ("expected 'while'"); |
1757 | for_loop* s = new for_loop; | |
1758 | s->tok = t; | |
1759 | ||
1760 | t = next (); | |
1761 | if (! (t->type == tok_operator && t->content == "(")) | |
1762 | throw parse_error ("expected '('"); | |
1763 | ||
1764 | // dummy init and incr fields | |
cbfbbf69 FCE |
1765 | s->init = 0; |
1766 | s->incr = 0; | |
f3c26ea5 FCE |
1767 | |
1768 | // condition | |
1769 | s->cond = parse_expression (); | |
1770 | ||
f3c26ea5 FCE |
1771 | t = next (); |
1772 | if (! (t->type == tok_operator && t->content == ")")) | |
1773 | throw parse_error ("expected ')'"); | |
dff50e09 | 1774 | |
f3c26ea5 FCE |
1775 | // block |
1776 | s->block = parse_statement (); | |
1777 | ||
1778 | return s; | |
69c68955 FCE |
1779 | } |
1780 | ||
1781 | ||
1782 | foreach_loop* | |
1783 | parser::parse_foreach_loop () | |
1784 | { | |
1785 | const token* t = next (); | |
6e213f58 | 1786 | if (! (t->type == tok_keyword && t->content == "foreach")) |
69c68955 FCE |
1787 | throw parse_error ("expected 'foreach'"); |
1788 | foreach_loop* s = new foreach_loop; | |
1789 | s->tok = t; | |
93484556 | 1790 | s->sort_direction = 0; |
27f21e8c | 1791 | s->limit = NULL; |
69c68955 FCE |
1792 | |
1793 | t = next (); | |
1794 | if (! (t->type == tok_operator && t->content == "(")) | |
1795 | throw parse_error ("expected '('"); | |
1796 | ||
1797 | // see also parse_array_in | |
1798 | ||
1799 | bool parenthesized = false; | |
1800 | t = peek (); | |
1801 | if (t && t->type == tok_operator && t->content == "[") | |
1802 | { | |
1803 | next (); | |
1804 | parenthesized = true; | |
1805 | } | |
1806 | ||
1807 | while (1) | |
1808 | { | |
1809 | t = next (); | |
1810 | if (! (t->type == tok_identifier)) | |
1811 | throw parse_error ("expected identifier"); | |
1812 | symbol* sym = new symbol; | |
1813 | sym->tok = t; | |
1814 | sym->name = t->content; | |
1815 | s->indexes.push_back (sym); | |
1816 | ||
93484556 FCE |
1817 | t = peek (); |
1818 | if (t && t->type == tok_operator && | |
1819 | (t->content == "+" || t->content == "-")) | |
1820 | { | |
1821 | if (s->sort_direction) | |
1822 | throw parse_error ("multiple sort directives"); | |
1823 | s->sort_direction = (t->content == "+") ? 1 : -1; | |
1824 | s->sort_column = s->indexes.size(); | |
1825 | next(); | |
1826 | } | |
1827 | ||
69c68955 FCE |
1828 | if (parenthesized) |
1829 | { | |
93484556 | 1830 | t = peek (); |
69c68955 FCE |
1831 | if (t && t->type == tok_operator && t->content == ",") |
1832 | { | |
1833 | next (); | |
1834 | continue; | |
1835 | } | |
1836 | else if (t && t->type == tok_operator && t->content == "]") | |
1837 | { | |
1838 | next (); | |
1839 | break; | |
1840 | } | |
dff50e09 | 1841 | else |
69c68955 FCE |
1842 | throw parse_error ("expected ',' or ']'"); |
1843 | } | |
1844 | else | |
1845 | break; // expecting only one expression | |
1846 | } | |
1847 | ||
1848 | t = next (); | |
6e213f58 | 1849 | if (! (t->type == tok_keyword && t->content == "in")) |
69c68955 | 1850 | throw parse_error ("expected 'in'"); |
dff50e09 | 1851 | |
d02548c0 | 1852 | s->base = parse_indexable(); |
69c68955 | 1853 | |
93484556 FCE |
1854 | t = peek (); |
1855 | if (t && t->type == tok_operator && | |
1856 | (t->content == "+" || t->content == "-")) | |
1857 | { | |
1858 | if (s->sort_direction) | |
1859 | throw parse_error ("multiple sort directives"); | |
1860 | s->sort_direction = (t->content == "+") ? 1 : -1; | |
1861 | s->sort_column = 0; | |
1862 | next(); | |
1863 | } | |
1864 | ||
27f21e8c DS |
1865 | t = peek (); |
1866 | if (tok_is(t, tok_keyword, "limit")) | |
1867 | { | |
1868 | next (); // get past the "limit" | |
1869 | s->limit = parse_expression (); | |
1870 | } | |
1871 | ||
69c68955 FCE |
1872 | t = next (); |
1873 | if (! (t->type == tok_operator && t->content == ")")) | |
1874 | throw parse_error ("expected ')'"); | |
1875 | ||
1876 | s->block = parse_statement (); | |
1877 | return s; | |
1878 | } | |
1879 | ||
1880 | ||
2f1a1aea FCE |
1881 | expression* |
1882 | parser::parse_expression () | |
1883 | { | |
1884 | return parse_assignment (); | |
1885 | } | |
1886 | ||
2f1a1aea FCE |
1887 | |
1888 | expression* | |
1889 | parser::parse_assignment () | |
1890 | { | |
1891 | expression* op1 = parse_ternary (); | |
1892 | ||
1893 | const token* t = peek (); | |
82919855 | 1894 | // right-associative operators |
dff50e09 | 1895 | if (t && t->type == tok_operator |
2f1a1aea | 1896 | && (t->content == "=" || |
82919855 | 1897 | t->content == "<<<" || |
2f1a1aea | 1898 | t->content == "+=" || |
bb2e3076 FCE |
1899 | t->content == "-=" || |
1900 | t->content == "*=" || | |
1901 | t->content == "/=" || | |
1902 | t->content == "%=" || | |
1903 | t->content == "<<=" || | |
1904 | t->content == ">>=" || | |
1905 | t->content == "&=" || | |
1906 | t->content == "^=" || | |
1907 | t->content == "|=" || | |
d5d7c2cc | 1908 | t->content == ".=" || |
dff50e09 | 1909 | false)) |
2f1a1aea | 1910 | { |
bb2e3076 | 1911 | // NB: lvalueness is checked during elaboration / translation |
2f1a1aea | 1912 | assignment* e = new assignment; |
56099f08 | 1913 | e->left = op1; |
2f1a1aea | 1914 | e->op = t->content; |
56099f08 | 1915 | e->tok = t; |
2f1a1aea | 1916 | next (); |
82919855 | 1917 | e->right = parse_expression (); |
56099f08 | 1918 | op1 = e; |
2f1a1aea | 1919 | } |
56099f08 FCE |
1920 | |
1921 | return op1; | |
2f1a1aea FCE |
1922 | } |
1923 | ||
1924 | ||
1925 | expression* | |
1926 | parser::parse_ternary () | |
1927 | { | |
1928 | expression* op1 = parse_logical_or (); | |
1929 | ||
1930 | const token* t = peek (); | |
1931 | if (t && t->type == tok_operator && t->content == "?") | |
1932 | { | |
2f1a1aea | 1933 | ternary_expression* e = new ternary_expression; |
56099f08 | 1934 | e->tok = t; |
2f1a1aea | 1935 | e->cond = op1; |
56099f08 FCE |
1936 | next (); |
1937 | e->truevalue = parse_expression (); // XXX | |
2f1a1aea FCE |
1938 | |
1939 | t = next (); | |
1940 | if (! (t->type == tok_operator && t->content == ":")) | |
1941 | throw parse_error ("expected ':'"); | |
1942 | ||
56099f08 | 1943 | e->falsevalue = parse_expression (); // XXX |
2f1a1aea FCE |
1944 | return e; |
1945 | } | |
1946 | else | |
1947 | return op1; | |
1948 | } | |
1949 | ||
1950 | ||
1951 | expression* | |
1952 | parser::parse_logical_or () | |
1953 | { | |
1954 | expression* op1 = parse_logical_and (); | |
dff50e09 | 1955 | |
2f1a1aea | 1956 | const token* t = peek (); |
56099f08 | 1957 | while (t && t->type == tok_operator && t->content == "||") |
2f1a1aea | 1958 | { |
2f1a1aea | 1959 | logical_or_expr* e = new logical_or_expr; |
56099f08 FCE |
1960 | e->tok = t; |
1961 | e->op = t->content; | |
2f1a1aea | 1962 | e->left = op1; |
56099f08 FCE |
1963 | next (); |
1964 | e->right = parse_logical_and (); | |
1965 | op1 = e; | |
1966 | t = peek (); | |
2f1a1aea | 1967 | } |
56099f08 FCE |
1968 | |
1969 | return op1; | |
2f1a1aea FCE |
1970 | } |
1971 | ||
1972 | ||
1973 | expression* | |
1974 | parser::parse_logical_and () | |
1975 | { | |
bb2e3076 | 1976 | expression* op1 = parse_boolean_or (); |
2f1a1aea FCE |
1977 | |
1978 | const token* t = peek (); | |
56099f08 | 1979 | while (t && t->type == tok_operator && t->content == "&&") |
2f1a1aea | 1980 | { |
2f1a1aea FCE |
1981 | logical_and_expr *e = new logical_and_expr; |
1982 | e->left = op1; | |
56099f08 FCE |
1983 | e->op = t->content; |
1984 | e->tok = t; | |
1985 | next (); | |
bb2e3076 FCE |
1986 | e->right = parse_boolean_or (); |
1987 | op1 = e; | |
1988 | t = peek (); | |
1989 | } | |
1990 | ||
1991 | return op1; | |
1992 | } | |
1993 | ||
1994 | ||
1995 | expression* | |
1996 | parser::parse_boolean_or () | |
1997 | { | |
1998 | expression* op1 = parse_boolean_xor (); | |
1999 | ||
2000 | const token* t = peek (); | |
2001 | while (t && t->type == tok_operator && t->content == "|") | |
2002 | { | |
2003 | binary_expression* e = new binary_expression; | |
2004 | e->left = op1; | |
2005 | e->op = t->content; | |
2006 | e->tok = t; | |
2007 | next (); | |
2008 | e->right = parse_boolean_xor (); | |
2009 | op1 = e; | |
2010 | t = peek (); | |
2011 | } | |
2012 | ||
2013 | return op1; | |
2014 | } | |
2015 | ||
2016 | ||
2017 | expression* | |
2018 | parser::parse_boolean_xor () | |
2019 | { | |
2020 | expression* op1 = parse_boolean_and (); | |
2021 | ||
2022 | const token* t = peek (); | |
2023 | while (t && t->type == tok_operator && t->content == "^") | |
2024 | { | |
2025 | binary_expression* e = new binary_expression; | |
2026 | e->left = op1; | |
2027 | e->op = t->content; | |
2028 | e->tok = t; | |
2029 | next (); | |
2030 | e->right = parse_boolean_and (); | |
2031 | op1 = e; | |
2032 | t = peek (); | |
2033 | } | |
2034 | ||
2035 | return op1; | |
2036 | } | |
2037 | ||
2038 | ||
2039 | expression* | |
2040 | parser::parse_boolean_and () | |
2041 | { | |
2042 | expression* op1 = parse_array_in (); | |
2043 | ||
2044 | const token* t = peek (); | |
2045 | while (t && t->type == tok_operator && t->content == "&") | |
2046 | { | |
2047 | binary_expression* e = new binary_expression; | |
2048 | e->left = op1; | |
2049 | e->op = t->content; | |
2050 | e->tok = t; | |
2051 | next (); | |
56099f08 FCE |
2052 | e->right = parse_array_in (); |
2053 | op1 = e; | |
2054 | t = peek (); | |
2f1a1aea | 2055 | } |
56099f08 FCE |
2056 | |
2057 | return op1; | |
2f1a1aea FCE |
2058 | } |
2059 | ||
2060 | ||
2061 | expression* | |
2062 | parser::parse_array_in () | |
2063 | { | |
ce10591c | 2064 | // This is a very tricky case. All these are legit expressions: |
69c68955 | 2065 | // "a in b" "a+0 in b" "[a,b] in c" "[c,(d+0)] in b" |
ce10591c FCE |
2066 | vector<expression*> indexes; |
2067 | bool parenthesized = false; | |
2f1a1aea FCE |
2068 | |
2069 | const token* t = peek (); | |
69c68955 | 2070 | if (t && t->type == tok_operator && t->content == "[") |
ce10591c FCE |
2071 | { |
2072 | next (); | |
2073 | parenthesized = true; | |
2074 | } | |
2075 | ||
2076 | while (1) | |
2077 | { | |
2078 | expression* op1 = parse_comparison (); | |
2079 | indexes.push_back (op1); | |
2080 | ||
2081 | if (parenthesized) | |
2082 | { | |
2083 | const token* t = peek (); | |
2084 | if (t && t->type == tok_operator && t->content == ",") | |
2085 | { | |
2086 | next (); | |
2087 | continue; | |
2088 | } | |
69c68955 | 2089 | else if (t && t->type == tok_operator && t->content == "]") |
ce10591c FCE |
2090 | { |
2091 | next (); | |
2092 | break; | |
2093 | } | |
dff50e09 | 2094 | else |
69c68955 | 2095 | throw parse_error ("expected ',' or ']'"); |
ce10591c FCE |
2096 | } |
2097 | else | |
2098 | break; // expecting only one expression | |
2099 | } | |
2100 | ||
2101 | t = peek (); | |
6e213f58 | 2102 | if (t && t->type == tok_keyword && t->content == "in") |
2f1a1aea | 2103 | { |
2f1a1aea | 2104 | array_in *e = new array_in; |
56099f08 | 2105 | e->tok = t; |
ce10591c FCE |
2106 | next (); // swallow "in" |
2107 | ||
2108 | arrayindex* a = new arrayindex; | |
2109 | a->indexes = indexes; | |
d02548c0 GH |
2110 | a->base = parse_indexable(); |
2111 | a->tok = a->base->get_tok(); | |
ce10591c | 2112 | e->operand = a; |
2f1a1aea FCE |
2113 | return e; |
2114 | } | |
ce10591c FCE |
2115 | else if (indexes.size() == 1) // no "in" - need one expression only |
2116 | return indexes[0]; | |
2f1a1aea | 2117 | else |
ce10591c | 2118 | throw parse_error ("unexpected comma-separated expression list"); |
2f1a1aea FCE |
2119 | } |
2120 | ||
2121 | ||
2122 | expression* | |
2123 | parser::parse_comparison () | |
2124 | { | |
bb2e3076 | 2125 | expression* op1 = parse_shift (); |
2f1a1aea FCE |
2126 | |
2127 | const token* t = peek (); | |
dff50e09 | 2128 | while (t && t->type == tok_operator |
553d27a5 FCE |
2129 | && (t->content == ">" || |
2130 | t->content == "<" || | |
2131 | t->content == "==" || | |
2132 | t->content == "!=" || | |
2133 | t->content == "<=" || | |
bb2e3076 | 2134 | t->content == ">=")) |
2f1a1aea FCE |
2135 | { |
2136 | comparison* e = new comparison; | |
2137 | e->left = op1; | |
2138 | e->op = t->content; | |
56099f08 | 2139 | e->tok = t; |
2f1a1aea | 2140 | next (); |
bb2e3076 FCE |
2141 | e->right = parse_shift (); |
2142 | op1 = e; | |
2143 | t = peek (); | |
2144 | } | |
2145 | ||
2146 | return op1; | |
2147 | } | |
2148 | ||
2149 | ||
2150 | expression* | |
2151 | parser::parse_shift () | |
2152 | { | |
2153 | expression* op1 = parse_concatenation (); | |
2154 | ||
2155 | const token* t = peek (); | |
dff50e09 | 2156 | while (t && t->type == tok_operator && |
bb2e3076 FCE |
2157 | (t->content == "<<" || t->content == ">>")) |
2158 | { | |
2159 | binary_expression* e = new binary_expression; | |
2160 | e->left = op1; | |
2161 | e->op = t->content; | |
2162 | e->tok = t; | |
2163 | next (); | |
56099f08 FCE |
2164 | e->right = parse_concatenation (); |
2165 | op1 = e; | |
2166 | t = peek (); | |
2f1a1aea | 2167 | } |
56099f08 FCE |
2168 | |
2169 | return op1; | |
2f1a1aea FCE |
2170 | } |
2171 | ||
2172 | ||
2173 | expression* | |
2174 | parser::parse_concatenation () | |
2175 | { | |
2176 | expression* op1 = parse_additive (); | |
2177 | ||
2178 | const token* t = peek (); | |
2179 | // XXX: the actual awk string-concatenation operator is *whitespace*. | |
2180 | // I don't know how to easily to model that here. | |
56099f08 | 2181 | while (t && t->type == tok_operator && t->content == ".") |
2f1a1aea FCE |
2182 | { |
2183 | concatenation* e = new concatenation; | |
2184 | e->left = op1; | |
2185 | e->op = t->content; | |
56099f08 | 2186 | e->tok = t; |
2f1a1aea | 2187 | next (); |
56099f08 FCE |
2188 | e->right = parse_additive (); |
2189 | op1 = e; | |
2190 | t = peek (); | |
2f1a1aea | 2191 | } |
56099f08 FCE |
2192 | |
2193 | return op1; | |
2f1a1aea FCE |
2194 | } |
2195 | ||
2196 | ||
2197 | expression* | |
2198 | parser::parse_additive () | |
2199 | { | |
2200 | expression* op1 = parse_multiplicative (); | |
2201 | ||
2202 | const token* t = peek (); | |
dff50e09 | 2203 | while (t && t->type == tok_operator |
2f1a1aea FCE |
2204 | && (t->content == "+" || t->content == "-")) |
2205 | { | |
2206 | binary_expression* e = new binary_expression; | |
2207 | e->op = t->content; | |
2208 | e->left = op1; | |
56099f08 | 2209 | e->tok = t; |
2f1a1aea | 2210 | next (); |
56099f08 FCE |
2211 | e->right = parse_multiplicative (); |
2212 | op1 = e; | |
2213 | t = peek (); | |
2f1a1aea | 2214 | } |
56099f08 FCE |
2215 | |
2216 | return op1; | |
2f1a1aea FCE |
2217 | } |
2218 | ||
2219 | ||
2220 | expression* | |
2221 | parser::parse_multiplicative () | |
2222 | { | |
2223 | expression* op1 = parse_unary (); | |
2224 | ||
2225 | const token* t = peek (); | |
dff50e09 | 2226 | while (t && t->type == tok_operator |
2f1a1aea FCE |
2227 | && (t->content == "*" || t->content == "/" || t->content == "%")) |
2228 | { | |
2229 | binary_expression* e = new binary_expression; | |
2230 | e->op = t->content; | |
2231 | e->left = op1; | |
56099f08 | 2232 | e->tok = t; |
2f1a1aea | 2233 | next (); |
56099f08 FCE |
2234 | e->right = parse_unary (); |
2235 | op1 = e; | |
2236 | t = peek (); | |
2f1a1aea | 2237 | } |
56099f08 FCE |
2238 | |
2239 | return op1; | |
2f1a1aea FCE |
2240 | } |
2241 | ||
2242 | ||
2243 | expression* | |
2244 | parser::parse_unary () | |
2245 | { | |
2246 | const token* t = peek (); | |
dff50e09 FCE |
2247 | if (t && t->type == tok_operator |
2248 | && (t->content == "+" || | |
2249 | t->content == "-" || | |
bb2e3076 FCE |
2250 | t->content == "!" || |
2251 | t->content == "~" || | |
2252 | false)) | |
2f1a1aea FCE |
2253 | { |
2254 | unary_expression* e = new unary_expression; | |
2255 | e->op = t->content; | |
56099f08 | 2256 | e->tok = t; |
2f1a1aea | 2257 | next (); |
3a20432b | 2258 | e->operand = parse_crement (); |
2f1a1aea FCE |
2259 | return e; |
2260 | } | |
2261 | else | |
bb2e3076 | 2262 | return parse_crement (); |
2f1a1aea FCE |
2263 | } |
2264 | ||
2265 | ||
2266 | expression* | |
2267 | parser::parse_crement () // as in "increment" / "decrement" | |
2268 | { | |
cbfbbf69 FCE |
2269 | // NB: Ideally, we'd parse only a symbol as an operand to the |
2270 | // *crement operators, instead of a general expression value. We'd | |
2271 | // need more complex lookahead code to tell apart the postfix cases. | |
2272 | // So we just punt, and leave it to pass-3 to signal errors on | |
2273 | // cases like "4++". | |
2274 | ||
2f1a1aea | 2275 | const token* t = peek (); |
dff50e09 | 2276 | if (t && t->type == tok_operator |
2f1a1aea FCE |
2277 | && (t->content == "++" || t->content == "--")) |
2278 | { | |
2279 | pre_crement* e = new pre_crement; | |
2280 | e->op = t->content; | |
56099f08 | 2281 | e->tok = t; |
2f1a1aea FCE |
2282 | next (); |
2283 | e->operand = parse_value (); | |
2284 | return e; | |
2285 | } | |
2286 | ||
2287 | // post-crement or non-crement | |
2288 | expression *op1 = parse_value (); | |
dff50e09 | 2289 | |
2f1a1aea | 2290 | t = peek (); |
dff50e09 | 2291 | if (t && t->type == tok_operator |
2f1a1aea FCE |
2292 | && (t->content == "++" || t->content == "--")) |
2293 | { | |
2294 | post_crement* e = new post_crement; | |
2295 | e->op = t->content; | |
56099f08 | 2296 | e->tok = t; |
2f1a1aea FCE |
2297 | next (); |
2298 | e->operand = op1; | |
2299 | return e; | |
2300 | } | |
2301 | else | |
2302 | return op1; | |
2303 | } | |
2304 | ||
2305 | ||
2306 | expression* | |
2307 | parser::parse_value () | |
2308 | { | |
2309 | const token* t = peek (); | |
2310 | if (! t) | |
2311 | throw parse_error ("expected value"); | |
2312 | ||
2313 | if (t->type == tok_operator && t->content == "(") | |
2314 | { | |
2315 | next (); | |
2316 | expression* e = parse_expression (); | |
2317 | t = next (); | |
2318 | if (! (t->type == tok_operator && t->content == ")")) | |
2319 | throw parse_error ("expected ')'"); | |
2320 | return e; | |
2321 | } | |
03c75a4a JS |
2322 | else if (t->type == tok_operator && t->content == "&") |
2323 | { | |
2324 | next (); | |
2325 | t = peek (); | |
2326 | if (t->type != tok_identifier || | |
2327 | (t->content != "@cast" && t->content[0] != '$')) | |
2328 | throw parse_error ("expected @cast or $var"); | |
2329 | ||
2330 | target_symbol *ts = static_cast<target_symbol*>(parse_symbol()); | |
2331 | ts->addressof = true; | |
2332 | return ts; | |
2333 | } | |
2f1a1aea FCE |
2334 | else if (t->type == tok_identifier) |
2335 | return parse_symbol (); | |
2336 | else | |
2337 | return parse_literal (); | |
2338 | } | |
2339 | ||
2340 | ||
d02548c0 GH |
2341 | const token * |
2342 | parser::parse_hist_op_or_bare_name (hist_op *&hop, string &name) | |
2343 | { | |
2344 | hop = NULL; | |
2345 | const token* t = expect_ident (name); | |
2346 | if (name == "@hist_linear" || name == "@hist_log") | |
2347 | { | |
2348 | hop = new hist_op; | |
2349 | if (name == "@hist_linear") | |
2350 | hop->htype = hist_linear; | |
2351 | else if (name == "@hist_log") | |
2352 | hop->htype = hist_log; | |
2353 | hop->tok = t; | |
2354 | expect_op("("); | |
2355 | hop->stat = parse_expression (); | |
2356 | int64_t tnum; | |
2357 | if (hop->htype == hist_linear) | |
2358 | { | |
2359 | for (size_t i = 0; i < 3; ++i) | |
2360 | { | |
2361 | expect_op (","); | |
2362 | expect_number (tnum); | |
2363 | hop->params.push_back (tnum); | |
2364 | } | |
2365 | } | |
d02548c0 GH |
2366 | expect_op(")"); |
2367 | } | |
2368 | return t; | |
2369 | } | |
2370 | ||
2371 | ||
2372 | indexable* | |
2373 | parser::parse_indexable () | |
2374 | { | |
2375 | hist_op *hop = NULL; | |
2376 | string name; | |
2377 | const token *tok = parse_hist_op_or_bare_name(hop, name); | |
2378 | if (hop) | |
2379 | return hop; | |
2380 | else | |
2381 | { | |
2382 | symbol* sym = new symbol; | |
2383 | sym->name = name; | |
2384 | sym->tok = tok; | |
2385 | return sym; | |
2386 | } | |
2387 | } | |
2388 | ||
2389 | ||
2390 | // var, indexable[index], func(parms), printf("...", ...), $var, $var->member, @stat_op(stat) | |
2f1a1aea | 2391 | expression* |
dff50e09 | 2392 | parser::parse_symbol () |
2f1a1aea | 2393 | { |
d02548c0 GH |
2394 | hist_op *hop = NULL; |
2395 | symbol *sym = NULL; | |
d7f3e0c5 | 2396 | string name; |
d02548c0 GH |
2397 | const token *t = parse_hist_op_or_bare_name(hop, name); |
2398 | ||
2399 | if (!hop) | |
0fefb486 | 2400 | { |
dff50e09 | 2401 | // If we didn't get a hist_op, then we did get an identifier. We can |
d02548c0 GH |
2402 | // now scrutinize this identifier for the various magic forms of identifier |
2403 | // (printf, @stat_op, and $var...) | |
2404 | ||
9b5af295 JS |
2405 | if (name == "@cast") |
2406 | { | |
2407 | // type-punning time | |
2408 | cast_op *cop = new cast_op; | |
2409 | cop->tok = t; | |
2410 | cop->base_name = name; | |
2411 | expect_op("("); | |
2412 | cop->operand = parse_expression (); | |
2413 | expect_op(","); | |
2414 | expect_unknown(tok_string, cop->type); | |
c15b6083 MW |
2415 | // types never start with "struct<space>" or "union<space>", |
2416 | // so gobble it up. | |
2417 | if (cop->type.compare(0, 7, "struct ") == 0) | |
2418 | cop->type = cop->type.substr(7); | |
2419 | if (cop->type.compare(0, 6, "union ") == 0) | |
2420 | cop->type = cop->type.substr(6); | |
9b5af295 JS |
2421 | if (peek_op (",")) |
2422 | { | |
2423 | next(); | |
2424 | expect_unknown(tok_string, cop->module); | |
2425 | } | |
2426 | expect_op(")"); | |
81931eab JS |
2427 | parse_target_symbol_components(cop); |
2428 | ||
9b5af295 JS |
2429 | // if there aren't any dereferences, then the cast is pointless |
2430 | if (cop->components.empty()) | |
2431 | { | |
2432 | expression *op = cop->operand; | |
2433 | delete cop; | |
2434 | return op; | |
2435 | } | |
2436 | return cop; | |
2437 | } | |
2438 | ||
2439 | else if (name.size() > 0 && name[0] == '@') | |
d7f3e0c5 | 2440 | { |
d02548c0 GH |
2441 | stat_op *sop = new stat_op; |
2442 | if (name == "@avg") | |
2443 | sop->ctype = sc_average; | |
2444 | else if (name == "@count") | |
2445 | sop->ctype = sc_count; | |
2446 | else if (name == "@sum") | |
2447 | sop->ctype = sc_sum; | |
2448 | else if (name == "@min") | |
2449 | sop->ctype = sc_min; | |
2450 | else if (name == "@max") | |
2451 | sop->ctype = sc_max; | |
2452 | else | |
2453 | throw parse_error("unknown statistic operator " + name); | |
2454 | expect_op("("); | |
2455 | sop->tok = t; | |
2456 | sop->stat = parse_expression (); | |
2457 | expect_op(")"); | |
2458 | return sop; | |
2459 | } | |
dff50e09 | 2460 | |
d5e178c1 | 2461 | else if (print_format *fmt = print_format::create(t)) |
d02548c0 | 2462 | { |
d02548c0 | 2463 | expect_op("("); |
b15c465c PP |
2464 | if ((name == "print" || name == "println" || |
2465 | name == "sprint" || name == "sprintln") && | |
3cb17058 | 2466 | (peek_kw("@hist_linear") || peek_kw("@hist_log"))) |
a4636912 GH |
2467 | { |
2468 | // We have a special case where we recognize | |
2469 | // print(@hist_foo(bar)) as a magic print-the-histogram | |
2470 | // construct. This is sort of gross but it avoids | |
2471 | // promoting histogram references to typeful | |
2472 | // expressions. | |
dff50e09 | 2473 | |
1bbeef03 GH |
2474 | hop = NULL; |
2475 | t = parse_hist_op_or_bare_name(hop, name); | |
2476 | assert(hop); | |
dff50e09 | 2477 | |
1bbeef03 GH |
2478 | // It is, sadly, possible that even while parsing a |
2479 | // hist_op, we *mis-guessed* and the user wishes to | |
2480 | // print(@hist_op(foo)[bucket]), a scalar. In that case | |
2481 | // we must parse the arrayindex and print an expression. | |
dff50e09 | 2482 | |
1bbeef03 GH |
2483 | if (!peek_op ("[")) |
2484 | fmt->hist = hop; | |
2485 | else | |
2486 | { | |
2487 | // This is simplified version of the | |
2488 | // multi-array-index parser below, because we can | |
2489 | // only ever have one index on a histogram anyways. | |
2490 | expect_op("["); | |
2491 | struct arrayindex* ai = new arrayindex; | |
2492 | ai->tok = t; | |
2493 | ai->base = hop; | |
2494 | ai->indexes.push_back (parse_expression ()); | |
2495 | expect_op("]"); | |
2496 | fmt->args.push_back(ai); | |
2497 | } | |
a4636912 | 2498 | } |
d7f3e0c5 | 2499 | else |
d02548c0 | 2500 | { |
3cb17058 JS |
2501 | int min_args = 0; |
2502 | if (fmt->print_with_format) | |
2503 | { | |
2504 | // Consume and convert a format string. Agreement between the | |
2505 | // format string and the arguments is postponed to the | |
2506 | // typechecking phase. | |
2507 | string tmp; | |
2508 | expect_unknown (tok_string, tmp); | |
2509 | fmt->raw_components = tmp; | |
2510 | fmt->components = print_format::string_to_components (tmp); | |
2511 | } | |
2512 | else if (fmt->print_with_delim) | |
2513 | { | |
2514 | // Consume a delimiter to separate arguments. | |
2515 | fmt->delimiter.clear(); | |
2516 | fmt->delimiter.type = print_format::conv_literal; | |
2517 | expect_unknown (tok_string, fmt->delimiter.literal_string); | |
2518 | min_args = 2; | |
2519 | } | |
2520 | else | |
2521 | { | |
2522 | // If we are not printing with a format string, we must have | |
2523 | // at least one argument (of any type). | |
2524 | expression *e = parse_expression (); | |
2525 | fmt->args.push_back(e); | |
2526 | } | |
2527 | ||
2528 | // Consume any subsequent arguments. | |
2529 | while (min_args || !peek_op (")")) | |
2530 | { | |
2531 | expect_op(","); | |
2532 | expression *e = parse_expression (); | |
2533 | fmt->args.push_back(e); | |
2534 | if (min_args) | |
2535 | --min_args; | |
2536 | } | |
d02548c0 GH |
2537 | } |
2538 | expect_op(")"); | |
2539 | return fmt; | |
2540 | } | |
dff50e09 | 2541 | |
d02548c0 GH |
2542 | else if (name.size() > 0 && name[0] == '$') |
2543 | { | |
2544 | // target_symbol time | |
2545 | target_symbol *tsym = new target_symbol; | |
2546 | tsym->tok = t; | |
2547 | tsym->base_name = name; | |
81931eab | 2548 | parse_target_symbol_components(tsym); |
d02548c0 GH |
2549 | return tsym; |
2550 | } | |
2551 | ||
2552 | else if (peek_op ("(")) // function call | |
2553 | { | |
2554 | next (); | |
2555 | struct functioncall* f = new functioncall; | |
2556 | f->tok = t; | |
2557 | f->function = name; | |
2558 | // Allow empty actual parameter list | |
2559 | if (peek_op (")")) | |
2560 | { | |
2561 | next (); | |
2562 | return f; | |
2563 | } | |
2564 | while (1) | |
2565 | { | |
2566 | f->args.push_back (parse_expression ()); | |
2567 | if (peek_op (")")) | |
2568 | { | |
2569 | next(); | |
2570 | break; | |
2571 | } | |
2572 | else if (peek_op (",")) | |
2573 | { | |
2574 | next(); | |
2575 | continue; | |
2576 | } | |
2577 | else | |
2578 | throw parse_error ("expected ',' or ')'"); | |
2579 | } | |
2580 | return f; | |
2581 | } | |
2582 | ||
2583 | else | |
2584 | { | |
2585 | sym = new symbol; | |
2586 | sym->name = name; | |
2587 | sym->tok = t; | |
d7f3e0c5 | 2588 | } |
0fefb486 | 2589 | } |
dff50e09 FCE |
2590 | |
2591 | // By now, either we had a hist_op in the first place, or else | |
d02548c0 GH |
2592 | // we had a plain word and it was converted to a symbol. |
2593 | ||
70c743d8 | 2594 | assert (!hop != !sym); // logical XOR |
d02548c0 GH |
2595 | |
2596 | // All that remains is to check for array indexing | |
2597 | ||
d7f3e0c5 | 2598 | if (peek_op ("[")) // array |
2f1a1aea FCE |
2599 | { |
2600 | next (); | |
2601 | struct arrayindex* ai = new arrayindex; | |
d02548c0 GH |
2602 | ai->tok = t; |
2603 | ||
2604 | if (hop) | |
2605 | ai->base = hop; | |
2606 | else | |
2607 | ai->base = sym; | |
2608 | ||
2f1a1aea FCE |
2609 | while (1) |
2610 | { | |
2611 | ai->indexes.push_back (parse_expression ()); | |
d7f3e0c5 | 2612 | if (peek_op ("]")) |
dff50e09 FCE |
2613 | { |
2614 | next(); | |
2615 | break; | |
d7f3e0c5 GH |
2616 | } |
2617 | else if (peek_op (",")) | |
2618 | { | |
2619 | next(); | |
2620 | continue; | |
2621 | } | |
2f1a1aea FCE |
2622 | else |
2623 | throw parse_error ("expected ',' or ']'"); | |
2624 | } | |
2625 | return ai; | |
2626 | } | |
d02548c0 GH |
2627 | |
2628 | // If we got to here, we *should* have a symbol; if we have | |
2629 | // a hist_op on its own, it doesn't count as an expression, | |
2630 | // so we throw a parse error. | |
2631 | ||
2632 | if (hop) | |
2633 | throw parse_error("base histogram operator where expression expected", t); | |
dff50e09 FCE |
2634 | |
2635 | return sym; | |
2f1a1aea | 2636 | } |
56099f08 | 2637 | |
81931eab JS |
2638 | |
2639 | void | |
2640 | parser::parse_target_symbol_components (target_symbol* e) | |
2641 | { | |
2642 | while (true) | |
2643 | { | |
81931eab JS |
2644 | if (peek_op ("->")) |
2645 | { | |
c67847a0 JS |
2646 | const token* t = next(); |
2647 | string member; | |
2648 | expect_ident_or_keyword (member); | |
2649 | e->components.push_back (target_symbol::component(t, member)); | |
81931eab JS |
2650 | } |
2651 | else if (peek_op ("[")) | |
2652 | { | |
c67847a0 | 2653 | const token* t = next(); |
6fda2dff JS |
2654 | expression* index = parse_expression(); |
2655 | literal_number* ln = dynamic_cast<literal_number*>(index); | |
2656 | if (ln) | |
2657 | e->components.push_back (target_symbol::component(t, ln->value)); | |
2658 | else | |
2659 | e->components.push_back (target_symbol::component(t, index)); | |
81931eab | 2660 | expect_op ("]"); |
81931eab JS |
2661 | } |
2662 | else | |
2663 | break; | |
2664 | } | |
2665 | } | |
2666 | ||
73267b89 | 2667 | /* vim: set sw=2 ts=8 cino=>4,n-2,{2,^-2,t0,(0,u0,w1,M1 : */ |