1 | /* Copyright (C) 2000 Free Software Foundation, Inc. |
---|
2 | This file is part of the GNU C Library. |
---|
3 | Contributed by Bruno Haible <haible@clisp.cons.org>, 2000. |
---|
4 | |
---|
5 | |
---|
6 | NOTE: The canonical source of this file is maintained with the GNU C Library. |
---|
7 | Bugs can be reported to bug-glibc@gnu.org. |
---|
8 | |
---|
9 | This program is free software; you can redistribute it and/or modify it |
---|
10 | under the terms of the GNU General Public License as published by the |
---|
11 | Free Software Foundation; either version 2, or (at your option) any |
---|
12 | later version. |
---|
13 | |
---|
14 | This program is distributed in the hope that it will be useful, |
---|
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
17 | GNU General Public License for more details. |
---|
18 | |
---|
19 | You should have received a copy of the GNU General Public License |
---|
20 | along with this program; if not, write to the Free Software |
---|
21 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, |
---|
22 | USA. */ |
---|
23 | |
---|
24 | /* Construction of sparse 3-level tables. |
---|
25 | See wchar-lookup.h or coll-lookup.h for their structure and the |
---|
26 | meaning of p and q. |
---|
27 | |
---|
28 | Before including this file, set |
---|
29 | TABLE to the name of the structure to be defined |
---|
30 | ELEMENT to the type of every entry |
---|
31 | DEFAULT to the default value for empty entries |
---|
32 | ITERATE if you want the TABLE_iterate function to be defined |
---|
33 | NO_FINALIZE if you don't want the TABLE_finalize function to be defined |
---|
34 | |
---|
35 | This will define |
---|
36 | |
---|
37 | struct TABLE; |
---|
38 | void TABLE_init (struct TABLE *t); |
---|
39 | ELEMENT TABLE_get (struct TABLE *t, uint32_t wc); |
---|
40 | void TABLE_add (struct TABLE *t, uint32_t wc, ELEMENT value); |
---|
41 | void TABLE_iterate (struct TABLE *t, |
---|
42 | void (*fn) (uint32_t wc, ELEMENT value)); |
---|
43 | void TABLE_finalize (struct TABLE *t); |
---|
44 | */ |
---|
45 | |
---|
46 | #define CONCAT(a,b) CONCAT1(a,b) |
---|
47 | #define CONCAT1(a,b) a##b |
---|
48 | |
---|
49 | struct TABLE |
---|
50 | { |
---|
51 | /* Parameters. */ |
---|
52 | unsigned int p; |
---|
53 | unsigned int q; |
---|
54 | /* Working representation. */ |
---|
55 | size_t level1_alloc; |
---|
56 | size_t level1_size; |
---|
57 | uint32_t *level1; |
---|
58 | size_t level2_alloc; |
---|
59 | size_t level2_size; |
---|
60 | uint32_t *level2; |
---|
61 | size_t level3_alloc; |
---|
62 | size_t level3_size; |
---|
63 | ELEMENT *level3; |
---|
64 | /* Compressed representation. */ |
---|
65 | size_t result_size; |
---|
66 | char *result; |
---|
67 | }; |
---|
68 | |
---|
69 | /* Initialize. Assumes t->p and t->q have already been set. */ |
---|
70 | static inline void |
---|
71 | CONCAT(TABLE,_init) (struct TABLE *t) |
---|
72 | { |
---|
73 | t->level1 = NULL; |
---|
74 | t->level1_alloc = t->level1_size = 0; |
---|
75 | t->level2 = NULL; |
---|
76 | t->level2_alloc = t->level2_size = 0; |
---|
77 | t->level3 = NULL; |
---|
78 | t->level3_alloc = t->level3_size = 0; |
---|
79 | } |
---|
80 | |
---|
81 | /* Retrieve an entry. */ |
---|
82 | static inline ELEMENT |
---|
83 | CONCAT(TABLE,_get) (struct TABLE *t, uint32_t wc) |
---|
84 | { |
---|
85 | uint32_t index1 = wc >> (t->q + t->p); |
---|
86 | if (index1 < t->level1_size) |
---|
87 | { |
---|
88 | uint32_t lookup1 = t->level1[index1]; |
---|
89 | if (lookup1 != ~((uint32_t) 0)) |
---|
90 | { |
---|
91 | uint32_t index2 = ((wc >> t->p) & ((1 << t->q) - 1)) |
---|
92 | + (lookup1 << t->q); |
---|
93 | uint32_t lookup2 = t->level2[index2]; |
---|
94 | if (lookup2 != ~((uint32_t) 0)) |
---|
95 | { |
---|
96 | uint32_t index3 = (wc & ((1 << t->p) - 1)) |
---|
97 | + (lookup2 << t->p); |
---|
98 | ELEMENT lookup3 = t->level3[index3]; |
---|
99 | |
---|
100 | return lookup3; |
---|
101 | } |
---|
102 | } |
---|
103 | } |
---|
104 | return DEFAULT; |
---|
105 | } |
---|
106 | |
---|
107 | /* Add one entry. */ |
---|
108 | static void |
---|
109 | CONCAT(TABLE,_add) (struct TABLE *t, uint32_t wc, ELEMENT value) |
---|
110 | { |
---|
111 | uint32_t index1 = wc >> (t->q + t->p); |
---|
112 | uint32_t index2 = (wc >> t->p) & ((1 << t->q) - 1); |
---|
113 | uint32_t index3 = wc & ((1 << t->p) - 1); |
---|
114 | size_t i, i1, i2; |
---|
115 | |
---|
116 | if (value == CONCAT(TABLE,_get) (t, wc)) |
---|
117 | return; |
---|
118 | |
---|
119 | if (index1 >= t->level1_size) |
---|
120 | { |
---|
121 | if (index1 >= t->level1_alloc) |
---|
122 | { |
---|
123 | size_t alloc = 2 * t->level1_alloc; |
---|
124 | if (alloc <= index1) |
---|
125 | alloc = index1 + 1; |
---|
126 | t->level1 = (uint32_t *) xrealloc ((char *) t->level1, |
---|
127 | alloc * sizeof (uint32_t)); |
---|
128 | t->level1_alloc = alloc; |
---|
129 | } |
---|
130 | while (index1 >= t->level1_size) |
---|
131 | t->level1[t->level1_size++] = ~((uint32_t) 0); |
---|
132 | } |
---|
133 | |
---|
134 | if (t->level1[index1] == ~((uint32_t) 0)) |
---|
135 | { |
---|
136 | if (t->level2_size == t->level2_alloc) |
---|
137 | { |
---|
138 | size_t alloc = 2 * t->level2_alloc + 1; |
---|
139 | t->level2 = (uint32_t *) xrealloc ((char *) t->level2, |
---|
140 | (alloc << t->q) * sizeof (uint32_t)); |
---|
141 | t->level2_alloc = alloc; |
---|
142 | } |
---|
143 | i1 = t->level2_size << t->q; |
---|
144 | i2 = (t->level2_size + 1) << t->q; |
---|
145 | for (i = i1; i < i2; i++) |
---|
146 | t->level2[i] = ~((uint32_t) 0); |
---|
147 | t->level1[index1] = t->level2_size++; |
---|
148 | } |
---|
149 | |
---|
150 | index2 += t->level1[index1] << t->q; |
---|
151 | |
---|
152 | if (t->level2[index2] == ~((uint32_t) 0)) |
---|
153 | { |
---|
154 | if (t->level3_size == t->level3_alloc) |
---|
155 | { |
---|
156 | size_t alloc = 2 * t->level3_alloc + 1; |
---|
157 | t->level3 = (ELEMENT *) xrealloc ((char *) t->level3, |
---|
158 | (alloc << t->p) * sizeof (ELEMENT)); |
---|
159 | t->level3_alloc = alloc; |
---|
160 | } |
---|
161 | i1 = t->level3_size << t->p; |
---|
162 | i2 = (t->level3_size + 1) << t->p; |
---|
163 | for (i = i1; i < i2; i++) |
---|
164 | t->level3[i] = DEFAULT; |
---|
165 | t->level2[index2] = t->level3_size++; |
---|
166 | } |
---|
167 | |
---|
168 | index3 += t->level2[index2] << t->p; |
---|
169 | |
---|
170 | t->level3[index3] = value; |
---|
171 | } |
---|
172 | |
---|
173 | #ifdef ITERATE |
---|
174 | /* Apply a function to all entries in the table. */ |
---|
175 | static void |
---|
176 | CONCAT(TABLE,_iterate) (struct TABLE *t, |
---|
177 | void (*fn) (uint32_t wc, ELEMENT value)) |
---|
178 | { |
---|
179 | uint32_t index1; |
---|
180 | for (index1 = 0; index1 < t->level1_size; index1++) |
---|
181 | { |
---|
182 | uint32_t lookup1 = t->level1[index1]; |
---|
183 | if (lookup1 != ~((uint32_t) 0)) |
---|
184 | { |
---|
185 | uint32_t lookup1_shifted = lookup1 << t->q; |
---|
186 | uint32_t index2; |
---|
187 | for (index2 = 0; index2 < (1 << t->q); index2++) |
---|
188 | { |
---|
189 | uint32_t lookup2 = t->level2[index2 + lookup1_shifted]; |
---|
190 | if (lookup2 != ~((uint32_t) 0)) |
---|
191 | { |
---|
192 | uint32_t lookup2_shifted = lookup2 << t->p; |
---|
193 | uint32_t index3; |
---|
194 | for (index3 = 0; index3 < (1 << t->p); index3++) |
---|
195 | { |
---|
196 | ELEMENT lookup3 = t->level3[index3 + lookup2_shifted]; |
---|
197 | if (lookup3 != DEFAULT) |
---|
198 | fn ((((index1 << t->q) + index2) << t->p) + index3, |
---|
199 | lookup3); |
---|
200 | } |
---|
201 | } |
---|
202 | } |
---|
203 | } |
---|
204 | } |
---|
205 | } |
---|
206 | #endif |
---|
207 | |
---|
208 | #ifndef NO_FINALIZE |
---|
209 | /* Finalize and shrink. */ |
---|
210 | static void |
---|
211 | CONCAT(TABLE,_finalize) (struct TABLE *t) |
---|
212 | { |
---|
213 | size_t i, j, k; |
---|
214 | uint32_t reorder3[t->level3_size]; |
---|
215 | uint32_t reorder2[t->level2_size]; |
---|
216 | uint32_t level1_offset, level2_offset, level3_offset, last_offset; |
---|
217 | |
---|
218 | /* Uniquify level3 blocks. */ |
---|
219 | k = 0; |
---|
220 | for (j = 0; j < t->level3_size; j++) |
---|
221 | { |
---|
222 | for (i = 0; i < k; i++) |
---|
223 | if (memcmp (&t->level3[i << t->p], &t->level3[j << t->p], |
---|
224 | (1 << t->p) * sizeof (ELEMENT)) == 0) |
---|
225 | break; |
---|
226 | /* Relocate block j to block i. */ |
---|
227 | reorder3[j] = i; |
---|
228 | if (i == k) |
---|
229 | { |
---|
230 | if (i != j) |
---|
231 | memcpy (&t->level3[i << t->p], &t->level3[j << t->p], |
---|
232 | (1 << t->p) * sizeof (ELEMENT)); |
---|
233 | k++; |
---|
234 | } |
---|
235 | } |
---|
236 | t->level3_size = k; |
---|
237 | |
---|
238 | for (i = 0; i < (t->level2_size << t->q); i++) |
---|
239 | if (t->level2[i] != ~((uint32_t) 0)) |
---|
240 | t->level2[i] = reorder3[t->level2[i]]; |
---|
241 | |
---|
242 | /* Uniquify level2 blocks. */ |
---|
243 | k = 0; |
---|
244 | for (j = 0; j < t->level2_size; j++) |
---|
245 | { |
---|
246 | for (i = 0; i < k; i++) |
---|
247 | if (memcmp (&t->level2[i << t->q], &t->level2[j << t->q], |
---|
248 | (1 << t->q) * sizeof (uint32_t)) == 0) |
---|
249 | break; |
---|
250 | /* Relocate block j to block i. */ |
---|
251 | reorder2[j] = i; |
---|
252 | if (i == k) |
---|
253 | { |
---|
254 | if (i != j) |
---|
255 | memcpy (&t->level2[i << t->q], &t->level2[j << t->q], |
---|
256 | (1 << t->q) * sizeof (uint32_t)); |
---|
257 | k++; |
---|
258 | } |
---|
259 | } |
---|
260 | t->level2_size = k; |
---|
261 | |
---|
262 | for (i = 0; i < t->level1_size; i++) |
---|
263 | if (t->level1[i] != ~((uint32_t) 0)) |
---|
264 | t->level1[i] = reorder2[t->level1[i]]; |
---|
265 | |
---|
266 | /* Create and fill the resulting compressed representation. */ |
---|
267 | last_offset = |
---|
268 | 5 * sizeof (uint32_t) |
---|
269 | + t->level1_size * sizeof (uint32_t) |
---|
270 | + (t->level2_size << t->q) * sizeof (uint32_t) |
---|
271 | + (t->level3_size << t->p) * sizeof (ELEMENT); |
---|
272 | t->result_size = (last_offset + 3) & ~3ul; |
---|
273 | t->result = (char *) xmalloc (t->result_size); |
---|
274 | |
---|
275 | level1_offset = |
---|
276 | 5 * sizeof (uint32_t); |
---|
277 | level2_offset = |
---|
278 | 5 * sizeof (uint32_t) |
---|
279 | + t->level1_size * sizeof (uint32_t); |
---|
280 | level3_offset = |
---|
281 | 5 * sizeof (uint32_t) |
---|
282 | + t->level1_size * sizeof (uint32_t) |
---|
283 | + (t->level2_size << t->q) * sizeof (uint32_t); |
---|
284 | |
---|
285 | ((uint32_t *) t->result)[0] = t->q + t->p; |
---|
286 | ((uint32_t *) t->result)[1] = t->level1_size; |
---|
287 | ((uint32_t *) t->result)[2] = t->p; |
---|
288 | ((uint32_t *) t->result)[3] = (1 << t->q) - 1; |
---|
289 | ((uint32_t *) t->result)[4] = (1 << t->p) - 1; |
---|
290 | |
---|
291 | for (i = 0; i < t->level1_size; i++) |
---|
292 | ((uint32_t *) (t->result + level1_offset))[i] = |
---|
293 | (t->level1[i] == ~((uint32_t) 0) |
---|
294 | ? 0 |
---|
295 | : (t->level1[i] << t->q) * sizeof (uint32_t) + level2_offset); |
---|
296 | |
---|
297 | for (i = 0; i < (t->level2_size << t->q); i++) |
---|
298 | ((uint32_t *) (t->result + level2_offset))[i] = |
---|
299 | (t->level2[i] == ~((uint32_t) 0) |
---|
300 | ? 0 |
---|
301 | : (t->level2[i] << t->p) * sizeof (ELEMENT) + level3_offset); |
---|
302 | |
---|
303 | for (i = 0; i < (t->level3_size << t->p); i++) |
---|
304 | ((ELEMENT *) (t->result + level3_offset))[i] = t->level3[i]; |
---|
305 | |
---|
306 | if (last_offset < t->result_size) |
---|
307 | memset (t->result + last_offset, 0, t->result_size - last_offset); |
---|
308 | |
---|
309 | if (t->level1_alloc > 0) |
---|
310 | free (t->level1); |
---|
311 | if (t->level2_alloc > 0) |
---|
312 | free (t->level2); |
---|
313 | if (t->level3_alloc > 0) |
---|
314 | free (t->level3); |
---|
315 | } |
---|
316 | #endif |
---|
317 | |
---|
318 | #undef TABLE |
---|
319 | #undef ELEMENT |
---|
320 | #undef DEFAULT |
---|
321 | #undef ITERATE |
---|
322 | #undef NO_FINALIZE |
---|