]>
Commit | Line | Data |
---|---|---|
1 | /* Generate the LR(0) parser states for Bison. | |
2 | ||
3 | Copyright (C) 1984, 1986, 1989, 2000-2002, 2004-2012 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 | ||
30 | #include "LR0.h" | |
31 | #include "closure.h" | |
32 | #include "complain.h" | |
33 | #include "getargs.h" | |
34 | #include "gram.h" | |
35 | #include "lalr.h" | |
36 | #include "reader.h" | |
37 | #include "reduce.h" | |
38 | #include "state.h" | |
39 | #include "symtab.h" | |
40 | ||
41 | typedef struct state_list | |
42 | { | |
43 | struct state_list *next; | |
44 | state *state; | |
45 | } state_list; | |
46 | ||
47 | static state_list *first_state = NULL; | |
48 | static state_list *last_state = NULL; | |
49 | ||
50 | ||
51 | /*------------------------------------------------------------------. | |
52 | | A state was just discovered from another state. Queue it for | | |
53 | | later examination, in order to find its transitions. Return it. | | |
54 | `------------------------------------------------------------------*/ | |
55 | ||
56 | static state * | |
57 | state_list_append (symbol_number sym, size_t core_size, item_number *core) | |
58 | { | |
59 | state_list *node = xmalloc (sizeof *node); | |
60 | state *s = state_new (sym, core_size, core); | |
61 | ||
62 | if (trace_flag & trace_automaton) | |
63 | fprintf (stderr, "state_list_append (state = %d, symbol = %d (%s))\n", | |
64 | nstates, sym, symbols[sym]->tag); | |
65 | ||
66 | node->next = NULL; | |
67 | node->state = s; | |
68 | ||
69 | if (!first_state) | |
70 | first_state = node; | |
71 | if (last_state) | |
72 | last_state->next = node; | |
73 | last_state = node; | |
74 | ||
75 | return s; | |
76 | } | |
77 | ||
78 | static int nshifts; | |
79 | static symbol_number *shift_symbol; | |
80 | ||
81 | static rule **redset; | |
82 | static state **shiftset; | |
83 | ||
84 | static item_number **kernel_base; | |
85 | static int *kernel_size; | |
86 | static item_number *kernel_items; | |
87 | ||
88 | \f | |
89 | static void | |
90 | allocate_itemsets (void) | |
91 | { | |
92 | symbol_number i; | |
93 | rule_number r; | |
94 | item_number *rhsp; | |
95 | ||
96 | /* Count the number of occurrences of all the symbols in RITEMS. | |
97 | Note that useless productions (hence useless nonterminals) are | |
98 | browsed too, hence we need to allocate room for _all_ the | |
99 | symbols. */ | |
100 | size_t count = 0; | |
101 | size_t *symbol_count = xcalloc (nsyms + nuseless_nonterminals, | |
102 | sizeof *symbol_count); | |
103 | ||
104 | for (r = 0; r < nrules; ++r) | |
105 | for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) | |
106 | { | |
107 | count++; | |
108 | symbol_count[*rhsp]++; | |
109 | } | |
110 | ||
111 | /* See comments before new_itemsets. All the vectors of items | |
112 | live inside KERNEL_ITEMS. The number of active items after | |
113 | some symbol S cannot be more than the number of times that S | |
114 | appears as an item, which is SYMBOL_COUNT[S]. | |
115 | We allocate that much space for each symbol. */ | |
116 | ||
117 | kernel_base = xnmalloc (nsyms, sizeof *kernel_base); | |
118 | kernel_items = xnmalloc (count, sizeof *kernel_items); | |
119 | ||
120 | count = 0; | |
121 | for (i = 0; i < nsyms; i++) | |
122 | { | |
123 | kernel_base[i] = kernel_items + count; | |
124 | count += symbol_count[i]; | |
125 | } | |
126 | ||
127 | free (symbol_count); | |
128 | kernel_size = xnmalloc (nsyms, sizeof *kernel_size); | |
129 | } | |
130 | ||
131 | ||
132 | static void | |
133 | allocate_storage (void) | |
134 | { | |
135 | allocate_itemsets (); | |
136 | ||
137 | shiftset = xnmalloc (nsyms, sizeof *shiftset); | |
138 | redset = xnmalloc (nrules, sizeof *redset); | |
139 | state_hash_new (); | |
140 | shift_symbol = xnmalloc (nsyms, sizeof *shift_symbol); | |
141 | } | |
142 | ||
143 | ||
144 | static void | |
145 | free_storage (void) | |
146 | { | |
147 | free (shift_symbol); | |
148 | free (redset); | |
149 | free (shiftset); | |
150 | free (kernel_base); | |
151 | free (kernel_size); | |
152 | free (kernel_items); | |
153 | state_hash_free (); | |
154 | } | |
155 | ||
156 | ||
157 | ||
158 | ||
159 | /*---------------------------------------------------------------. | |
160 | | Find which symbols can be shifted in S, and for each one | | |
161 | | record which items would be active after that shift. Uses the | | |
162 | | contents of itemset. | | |
163 | | | | |
164 | | shift_symbol is set to a vector of the symbols that can be | | |
165 | | shifted. For each symbol in the grammar, kernel_base[symbol] | | |
166 | | points to a vector of item numbers activated if that symbol is | | |
167 | | shifted, and kernel_size[symbol] is their numbers. | | |
168 | | | | |
169 | | itemset is sorted on item index in ritem, which is sorted on | | |
170 | | rule number. Compute each kernel_base[symbol] with the same | | |
171 | | sort. | | |
172 | `---------------------------------------------------------------*/ | |
173 | ||
174 | static void | |
175 | new_itemsets (state *s) | |
176 | { | |
177 | size_t i; | |
178 | ||
179 | if (trace_flag & trace_automaton) | |
180 | fprintf (stderr, "Entering new_itemsets, state = %d\n", s->number); | |
181 | ||
182 | memset (kernel_size, 0, nsyms * sizeof *kernel_size); | |
183 | ||
184 | nshifts = 0; | |
185 | ||
186 | for (i = 0; i < nitemset; ++i) | |
187 | if (item_number_is_symbol_number (ritem[itemset[i]])) | |
188 | { | |
189 | symbol_number sym = item_number_as_symbol_number (ritem[itemset[i]]); | |
190 | if (!kernel_size[sym]) | |
191 | { | |
192 | shift_symbol[nshifts] = sym; | |
193 | nshifts++; | |
194 | } | |
195 | ||
196 | kernel_base[sym][kernel_size[sym]] = itemset[i] + 1; | |
197 | kernel_size[sym]++; | |
198 | } | |
199 | } | |
200 | ||
201 | ||
202 | ||
203 | /*--------------------------------------------------------------. | |
204 | | Find the state we would get to (from the current state) by | | |
205 | | shifting SYM. Create a new state if no equivalent one exists | | |
206 | | already. Used by append_states. | | |
207 | `--------------------------------------------------------------*/ | |
208 | ||
209 | static state * | |
210 | get_state (symbol_number sym, size_t core_size, item_number *core) | |
211 | { | |
212 | state *s; | |
213 | ||
214 | if (trace_flag & trace_automaton) | |
215 | fprintf (stderr, "Entering get_state, symbol = %d (%s)\n", | |
216 | sym, symbols[sym]->tag); | |
217 | ||
218 | s = state_hash_lookup (core_size, core); | |
219 | if (!s) | |
220 | s = state_list_append (sym, core_size, core); | |
221 | ||
222 | if (trace_flag & trace_automaton) | |
223 | fprintf (stderr, "Exiting get_state => %d\n", s->number); | |
224 | ||
225 | return s; | |
226 | } | |
227 | ||
228 | /*---------------------------------------------------------------. | |
229 | | Use the information computed by new_itemsets to find the state | | |
230 | | numbers reached by each shift transition from S. | | |
231 | | | | |
232 | | SHIFTSET is set up as a vector of those states. | | |
233 | `---------------------------------------------------------------*/ | |
234 | ||
235 | static void | |
236 | append_states (state *s) | |
237 | { | |
238 | int i; | |
239 | ||
240 | if (trace_flag & trace_automaton) | |
241 | fprintf (stderr, "Entering append_states, state = %d\n", s->number); | |
242 | ||
243 | /* First sort shift_symbol into increasing order. */ | |
244 | ||
245 | for (i = 1; i < nshifts; i++) | |
246 | { | |
247 | symbol_number sym = shift_symbol[i]; | |
248 | int j; | |
249 | for (j = i; 0 < j && sym < shift_symbol[j - 1]; j--) | |
250 | shift_symbol[j] = shift_symbol[j - 1]; | |
251 | shift_symbol[j] = sym; | |
252 | } | |
253 | ||
254 | for (i = 0; i < nshifts; i++) | |
255 | { | |
256 | symbol_number sym = shift_symbol[i]; | |
257 | shiftset[i] = get_state (sym, kernel_size[sym], kernel_base[sym]); | |
258 | } | |
259 | } | |
260 | ||
261 | ||
262 | /*----------------------------------------------------------------. | |
263 | | Find which rules can be used for reduction transitions from the | | |
264 | | current state and make a reductions structure for the state to | | |
265 | | record their rule numbers. | | |
266 | `----------------------------------------------------------------*/ | |
267 | ||
268 | static void | |
269 | save_reductions (state *s) | |
270 | { | |
271 | int count = 0; | |
272 | size_t i; | |
273 | ||
274 | /* Find and count the active items that represent ends of rules. */ | |
275 | for (i = 0; i < nitemset; ++i) | |
276 | { | |
277 | item_number item = ritem[itemset[i]]; | |
278 | if (item_number_is_rule_number (item)) | |
279 | { | |
280 | rule_number r = item_number_as_rule_number (item); | |
281 | redset[count++] = &rules[r]; | |
282 | if (r == 0) | |
283 | { | |
284 | /* This is "reduce 0", i.e., accept. */ | |
285 | aver (!final_state); | |
286 | final_state = s; | |
287 | } | |
288 | } | |
289 | } | |
290 | ||
291 | /* Make a reductions structure and copy the data into it. */ | |
292 | state_reductions_set (s, count, redset); | |
293 | } | |
294 | ||
295 | \f | |
296 | /*---------------. | |
297 | | Build STATES. | | |
298 | `---------------*/ | |
299 | ||
300 | static void | |
301 | set_states (void) | |
302 | { | |
303 | states = xcalloc (nstates, sizeof *states); | |
304 | ||
305 | while (first_state) | |
306 | { | |
307 | state_list *this = first_state; | |
308 | ||
309 | /* Pessimization, but simplification of the code: make sure all | |
310 | the states have valid transitions and reductions members, | |
311 | even if reduced to 0. It is too soon for errs, which are | |
312 | computed later, but set_conflicts. */ | |
313 | state *s = this->state; | |
314 | if (!s->transitions) | |
315 | state_transitions_set (s, 0, 0); | |
316 | if (!s->reductions) | |
317 | state_reductions_set (s, 0, 0); | |
318 | ||
319 | states[s->number] = s; | |
320 | ||
321 | first_state = this->next; | |
322 | free (this); | |
323 | } | |
324 | first_state = NULL; | |
325 | last_state = NULL; | |
326 | } | |
327 | ||
328 | ||
329 | /*-------------------------------------------------------------------. | |
330 | | Compute the LR(0) parser states (see state.h for details) from the | | |
331 | | grammar. | | |
332 | `-------------------------------------------------------------------*/ | |
333 | ||
334 | void | |
335 | generate_states (void) | |
336 | { | |
337 | item_number initial_core = 0; | |
338 | state_list *list = NULL; | |
339 | allocate_storage (); | |
340 | new_closure (nritems); | |
341 | ||
342 | /* Create the initial state. The 0 at the lhs is the index of the | |
343 | item of this initial rule. */ | |
344 | state_list_append (0, 1, &initial_core); | |
345 | ||
346 | /* States are queued when they are created; process them all. */ | |
347 | for (list = first_state; list; list = list->next) | |
348 | { | |
349 | state *s = list->state; | |
350 | if (trace_flag & trace_automaton) | |
351 | fprintf (stderr, "Processing state %d (reached by %s)\n", | |
352 | s->number, | |
353 | symbols[s->accessing_symbol]->tag); | |
354 | /* Set up itemset for the transitions out of this state. itemset gets a | |
355 | vector of all the items that could be accepted next. */ | |
356 | closure (s->items, s->nitems); | |
357 | /* Record the reductions allowed out of this state. */ | |
358 | save_reductions (s); | |
359 | /* Find the itemsets of the states that shifts can reach. */ | |
360 | new_itemsets (s); | |
361 | /* Find or create the core structures for those states. */ | |
362 | append_states (s); | |
363 | ||
364 | /* Create the shifts structures for the shifts to those states, | |
365 | now that the state numbers transitioning to are known. */ | |
366 | state_transitions_set (s, nshifts, shiftset); | |
367 | } | |
368 | ||
369 | /* discard various storage */ | |
370 | free_closure (); | |
371 | free_storage (); | |
372 | ||
373 | /* Set up STATES. */ | |
374 | set_states (); | |
375 | } |