]>
Commit | Line | Data |
---|---|---|
1 | /* Generate the LR(0) parser states for Bison. | |
2 | ||
3 | Copyright (C) 1984, 1986, 1989, 2000-2002, 2004-2010 Free Software | |
4 | Foundation, Inc. | |
5 | ||
6 | This file is part of Bison, the GNU Compiler Compiler. | |
7 | ||
8 | This program is free software: you can redistribute it and/or modify | |
9 | it under the terms of the GNU General Public License as published by | |
10 | the Free Software Foundation, either version 3 of the License, or | |
11 | (at your option) any later version. | |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ | |
20 | ||
21 | ||
22 | /* See comments in state.h for the data structures that represent it. | |
23 | The entry point is generate_states. */ | |
24 | ||
25 | #include <config.h> | |
26 | #include "system.h" | |
27 | ||
28 | #include <bitset.h> | |
29 | #include <quotearg.h> | |
30 | ||
31 | #include "LR0.h" | |
32 | #include "closure.h" | |
33 | #include "complain.h" | |
34 | #include "getargs.h" | |
35 | #include "gram.h" | |
36 | #include "lalr.h" | |
37 | #include "reader.h" | |
38 | #include "reduce.h" | |
39 | #include "state.h" | |
40 | #include "symtab.h" | |
41 | ||
42 | typedef struct state_list | |
43 | { | |
44 | struct state_list *next; | |
45 | state *state; | |
46 | } state_list; | |
47 | ||
48 | static state_list *first_state = NULL; | |
49 | static state_list *last_state = NULL; | |
50 | ||
51 | ||
52 | /*------------------------------------------------------------------. | |
53 | | A state was just discovered from another state. Queue it for | | |
54 | | later examination, in order to find its transitions. Return it. | | |
55 | `------------------------------------------------------------------*/ | |
56 | ||
57 | static state * | |
58 | state_list_append (symbol_number sym, size_t core_size, item_number *core) | |
59 | { | |
60 | state_list *node = xmalloc (sizeof *node); | |
61 | state *s = state_new (sym, core_size, core); | |
62 | ||
63 | if (trace_flag & trace_automaton) | |
64 | fprintf (stderr, "state_list_append (state = %d, symbol = %d (%s))\n", | |
65 | nstates, sym, symbols[sym]->tag); | |
66 | ||
67 | node->next = NULL; | |
68 | node->state = s; | |
69 | ||
70 | if (!first_state) | |
71 | first_state = node; | |
72 | if (last_state) | |
73 | last_state->next = node; | |
74 | last_state = node; | |
75 | ||
76 | return s; | |
77 | } | |
78 | ||
79 | static int nshifts; | |
80 | static symbol_number *shift_symbol; | |
81 | ||
82 | static rule **redset; | |
83 | static state **shiftset; | |
84 | ||
85 | static item_number **kernel_base; | |
86 | static int *kernel_size; | |
87 | static item_number *kernel_items; | |
88 | ||
89 | \f | |
90 | static void | |
91 | allocate_itemsets (void) | |
92 | { | |
93 | symbol_number i; | |
94 | rule_number r; | |
95 | item_number *rhsp; | |
96 | ||
97 | /* Count the number of occurrences of all the symbols in RITEMS. | |
98 | Note that useless productions (hence useless nonterminals) are | |
99 | browsed too, hence we need to allocate room for _all_ the | |
100 | symbols. */ | |
101 | size_t count = 0; | |
102 | size_t *symbol_count = xcalloc (nsyms + nuseless_nonterminals, | |
103 | sizeof *symbol_count); | |
104 | ||
105 | for (r = 0; r < nrules; ++r) | |
106 | for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) | |
107 | { | |
108 | count++; | |
109 | symbol_count[*rhsp]++; | |
110 | } | |
111 | ||
112 | /* See comments before new_itemsets. All the vectors of items | |
113 | live inside KERNEL_ITEMS. The number of active items after | |
114 | some symbol S cannot be more than the number of times that S | |
115 | appears as an item, which is SYMBOL_COUNT[S]. | |
116 | We allocate that much space for each symbol. */ | |
117 | ||
118 | kernel_base = xnmalloc (nsyms, sizeof *kernel_base); | |
119 | kernel_items = xnmalloc (count, sizeof *kernel_items); | |
120 | ||
121 | count = 0; | |
122 | for (i = 0; i < nsyms; i++) | |
123 | { | |
124 | kernel_base[i] = kernel_items + count; | |
125 | count += symbol_count[i]; | |
126 | } | |
127 | ||
128 | free (symbol_count); | |
129 | kernel_size = xnmalloc (nsyms, sizeof *kernel_size); | |
130 | } | |
131 | ||
132 | ||
133 | static void | |
134 | allocate_storage (void) | |
135 | { | |
136 | allocate_itemsets (); | |
137 | ||
138 | shiftset = xnmalloc (nsyms, sizeof *shiftset); | |
139 | redset = xnmalloc (nrules, sizeof *redset); | |
140 | state_hash_new (); | |
141 | shift_symbol = xnmalloc (nsyms, sizeof *shift_symbol); | |
142 | } | |
143 | ||
144 | ||
145 | static void | |
146 | free_storage (void) | |
147 | { | |
148 | free (shift_symbol); | |
149 | free (redset); | |
150 | free (shiftset); | |
151 | free (kernel_base); | |
152 | free (kernel_size); | |
153 | free (kernel_items); | |
154 | state_hash_free (); | |
155 | } | |
156 | ||
157 | ||
158 | ||
159 | ||
160 | /*---------------------------------------------------------------. | |
161 | | Find which symbols can be shifted in S, and for each one | | |
162 | | record which items would be active after that shift. Uses the | | |
163 | | contents of itemset. | | |
164 | | | | |
165 | | shift_symbol is set to a vector of the symbols that can be | | |
166 | | shifted. For each symbol in the grammar, kernel_base[symbol] | | |
167 | | points to a vector of item numbers activated if that symbol is | | |
168 | | shifted, and kernel_size[symbol] is their numbers. | | |
169 | | | | |
170 | | itemset is sorted on item index in ritem, which is sorted on | | |
171 | | rule number. Compute each kernel_base[symbol] with the same | | |
172 | | sort. | | |
173 | `---------------------------------------------------------------*/ | |
174 | ||
175 | static void | |
176 | new_itemsets (state *s) | |
177 | { | |
178 | size_t i; | |
179 | ||
180 | if (trace_flag & trace_automaton) | |
181 | fprintf (stderr, "Entering new_itemsets, state = %d\n", s->number); | |
182 | ||
183 | memset (kernel_size, 0, nsyms * sizeof *kernel_size); | |
184 | ||
185 | nshifts = 0; | |
186 | ||
187 | for (i = 0; i < nitemset; ++i) | |
188 | if (item_number_is_symbol_number (ritem[itemset[i]])) | |
189 | { | |
190 | symbol_number sym = item_number_as_symbol_number (ritem[itemset[i]]); | |
191 | if (!kernel_size[sym]) | |
192 | { | |
193 | shift_symbol[nshifts] = sym; | |
194 | nshifts++; | |
195 | } | |
196 | ||
197 | kernel_base[sym][kernel_size[sym]] = itemset[i] + 1; | |
198 | kernel_size[sym]++; | |
199 | } | |
200 | } | |
201 | ||
202 | ||
203 | ||
204 | /*--------------------------------------------------------------. | |
205 | | Find the state we would get to (from the current state) by | | |
206 | | shifting SYM. Create a new state if no equivalent one exists | | |
207 | | already. Used by append_states. | | |
208 | `--------------------------------------------------------------*/ | |
209 | ||
210 | static state * | |
211 | get_state (symbol_number sym, size_t core_size, item_number *core) | |
212 | { | |
213 | state *s; | |
214 | ||
215 | if (trace_flag & trace_automaton) | |
216 | fprintf (stderr, "Entering get_state, symbol = %d (%s)\n", | |
217 | sym, symbols[sym]->tag); | |
218 | ||
219 | s = state_hash_lookup (core_size, core); | |
220 | if (!s) | |
221 | s = state_list_append (sym, core_size, core); | |
222 | ||
223 | if (trace_flag & trace_automaton) | |
224 | fprintf (stderr, "Exiting get_state => %d\n", s->number); | |
225 | ||
226 | return s; | |
227 | } | |
228 | ||
229 | /*---------------------------------------------------------------. | |
230 | | Use the information computed by new_itemsets to find the state | | |
231 | | numbers reached by each shift transition from S. | | |
232 | | | | |
233 | | SHIFTSET is set up as a vector of those states. | | |
234 | `---------------------------------------------------------------*/ | |
235 | ||
236 | static void | |
237 | append_states (state *s) | |
238 | { | |
239 | int i; | |
240 | ||
241 | if (trace_flag & trace_automaton) | |
242 | fprintf (stderr, "Entering append_states, state = %d\n", s->number); | |
243 | ||
244 | /* First sort shift_symbol into increasing order. */ | |
245 | ||
246 | for (i = 1; i < nshifts; i++) | |
247 | { | |
248 | symbol_number sym = shift_symbol[i]; | |
249 | int j; | |
250 | for (j = i; 0 < j && sym < shift_symbol[j - 1]; j--) | |
251 | shift_symbol[j] = shift_symbol[j - 1]; | |
252 | shift_symbol[j] = sym; | |
253 | } | |
254 | ||
255 | for (i = 0; i < nshifts; i++) | |
256 | { | |
257 | symbol_number sym = shift_symbol[i]; | |
258 | shiftset[i] = get_state (sym, kernel_size[sym], kernel_base[sym]); | |
259 | } | |
260 | } | |
261 | ||
262 | ||
263 | /*----------------------------------------------------------------. | |
264 | | Find which rules can be used for reduction transitions from the | | |
265 | | current state and make a reductions structure for the state to | | |
266 | | record their rule numbers. | | |
267 | `----------------------------------------------------------------*/ | |
268 | ||
269 | static void | |
270 | save_reductions (state *s) | |
271 | { | |
272 | int count = 0; | |
273 | size_t i; | |
274 | ||
275 | /* Find and count the active items that represent ends of rules. */ | |
276 | for (i = 0; i < nitemset; ++i) | |
277 | { | |
278 | item_number item = ritem[itemset[i]]; | |
279 | if (item_number_is_rule_number (item)) | |
280 | { | |
281 | rule_number r = item_number_as_rule_number (item); | |
282 | redset[count++] = &rules[r]; | |
283 | if (r == 0) | |
284 | { | |
285 | /* This is "reduce 0", i.e., accept. */ | |
286 | aver (!final_state); | |
287 | final_state = s; | |
288 | } | |
289 | } | |
290 | } | |
291 | ||
292 | /* Make a reductions structure and copy the data into it. */ | |
293 | state_reductions_set (s, count, redset); | |
294 | } | |
295 | ||
296 | \f | |
297 | /*---------------. | |
298 | | Build STATES. | | |
299 | `---------------*/ | |
300 | ||
301 | static void | |
302 | set_states (void) | |
303 | { | |
304 | states = xcalloc (nstates, sizeof *states); | |
305 | ||
306 | while (first_state) | |
307 | { | |
308 | state_list *this = first_state; | |
309 | ||
310 | /* Pessimization, but simplification of the code: make sure all | |
311 | the states have valid transitions and reductions members, | |
312 | even if reduced to 0. It is too soon for errs, which are | |
313 | computed later, but set_conflicts. */ | |
314 | state *s = this->state; | |
315 | if (!s->transitions) | |
316 | state_transitions_set (s, 0, 0); | |
317 | if (!s->reductions) | |
318 | state_reductions_set (s, 0, 0); | |
319 | ||
320 | states[s->number] = s; | |
321 | ||
322 | first_state = this->next; | |
323 | free (this); | |
324 | } | |
325 | first_state = NULL; | |
326 | last_state = NULL; | |
327 | } | |
328 | ||
329 | ||
330 | /*-------------------------------------------------------------------. | |
331 | | Compute the LR(0) parser states (see state.h for details) from the | | |
332 | | grammar. | | |
333 | `-------------------------------------------------------------------*/ | |
334 | ||
335 | void | |
336 | generate_states (void) | |
337 | { | |
338 | item_number initial_core = 0; | |
339 | state_list *list = NULL; | |
340 | allocate_storage (); | |
341 | new_closure (nritems); | |
342 | ||
343 | /* Create the initial state. The 0 at the lhs is the index of the | |
344 | item of this initial rule. */ | |
345 | state_list_append (0, 1, &initial_core); | |
346 | ||
347 | /* States are queued when they are created; process them all. */ | |
348 | for (list = first_state; list; list = list->next) | |
349 | { | |
350 | state *s = list->state; | |
351 | if (trace_flag & trace_automaton) | |
352 | fprintf (stderr, "Processing state %d (reached by %s)\n", | |
353 | s->number, | |
354 | symbols[s->accessing_symbol]->tag); | |
355 | /* Set up itemset for the transitions out of this state. itemset gets a | |
356 | vector of all the items that could be accepted next. */ | |
357 | closure (s->items, s->nitems); | |
358 | /* Record the reductions allowed out of this state. */ | |
359 | save_reductions (s); | |
360 | /* Find the itemsets of the states that shifts can reach. */ | |
361 | new_itemsets (s); | |
362 | /* Find or create the core structures for those states. */ | |
363 | append_states (s); | |
364 | ||
365 | /* Create the shifts structures for the shifts to those states, | |
366 | now that the state numbers transitioning to are known. */ | |
367 | state_transitions_set (s, nshifts, shiftset); | |
368 | } | |
369 | ||
370 | /* discard various storage */ | |
371 | free_closure (); | |
372 | free_storage (); | |
373 | ||
374 | /* Set up STATES. */ | |
375 | set_states (); | |
376 | } |