3 * $Id: sym.c,v 1.8 1999/08/02 14:45:48 mdw Exp $
5 * Symbol table management
7 * (c) 1998 Straylight/Edgeware
10 /*----- Licensing notice --------------------------------------------------*
12 * This file is part of the mLib utilities library.
14 * mLib is free software; you can redistribute it and/or modify
15 * it under the terms of the GNU Library General Public License as
16 * published by the Free Software Foundation; either version 2 of the
17 * License, or (at your option) any later version.
19 * mLib is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22 * GNU Library General Public License for more details.
24 * You should have received a copy of the GNU Library General Public
25 * License along with mLib; if not, write to the Free
26 * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
30 /*----- Revision history --------------------------------------------------*
33 * Revision 1.8 1999/08/02 14:45:48 mdw
34 * Break low-level hashtable code out from sym.
36 * Revision 1.7 1999/06/01 09:49:08 mdw
37 * Allow things to be looked up by just their caller-supplied hashes. This
38 * actually needs to be thought through better.
40 * Revision 1.6 1999/05/26 21:08:31 mdw
41 * Rename symbols in line with newer conventions.
43 * Revision 1.5 1999/05/13 22:48:37 mdw
44 * Twiddle the extension threshold. Change `-ise' to `-ize' throughout.
46 * Revision 1.4 1999/05/06 19:51:35 mdw
47 * Reformatted the LGPL notice a little bit.
49 * Revision 1.3 1999/05/05 18:50:31 mdw
50 * Change licensing conditions to LGPL.
52 * Revision 1.2 1998/11/26 19:27:33 mdw
53 * Move SYM_NAME into the header file. Fix bugs.
55 * Revision 1.1.1.1 1998/06/17 23:44:42 mdw
56 * Initial version of mLib
60 /*----- Header files ------------------------------------------------------*/
62 /* --- ANSI headers --- */
68 /* --- Local headers --- */
79 /*----- Tuning parameters -------------------------------------------------*/
81 /* --- Initial hash table size --- *
83 * This is the initial @mask@ value. It must be of the form %$2^n - 1$%,
84 * so that it can be used to mask of the bottom bits of a hash value.
87 #define SYM_INITSZ 64 /* Size of a new hash table */
89 /* --- Maximum load factor --- *
91 * This parameter controls how much the table has to be loaded before the
92 * table is extended. The number of elements %$n$%, the number of bins %$b$%
93 * and the limit %$l$% satisfy the relation %$n < bl$%; if a new item is
94 * added to the table and this relation is found to be false, the table is
98 #define SYM_LIMIT(n) ((n) * 2) /* Load factor for growing table */
100 /*----- Main code ---------------------------------------------------------*/
102 /* --- @sym_create@ --- *
104 * Arguments: @sym_table *t@ = symbol table to initialize
108 * Use: Initializes the given symbol table. Raises @EXC_NOMEM@ if
109 * there isn't enough memory.
112 void sym_create(sym_table
*t
)
114 TRACK_CTX("symbol table creation");
116 hash_create(&t
->t
, SYM_INITSZ
);
117 t
->load
= SYM_LIMIT(SYM_INITSZ
);
121 /* --- @sym_destroy@ --- *
123 * Arguments: @sym_table *t@ = pointer to symbol table in question
127 * Use: Destroys a symbol table, freeing all the memory it used to
131 void sym_destroy(sym_table
*t
)
135 TRACK_CTX("symbol table destruction");
144 if (p
->len
> SYM_BUFSZ
)
145 sub_free(p
->name
.p
, p
->len
);
153 /* --- @sym_find@ --- *
155 * Arguments: @sym_table *t@ = pointer to symbol table in question
156 * @const char *n@ = pointer to symbol table to look up
157 * @long l@ = length of the name string or negative to measure
158 * @size_t sz@ = size of desired symbol object, or zero
159 * @unsigned *f@ = pointer to a flag, or null.
161 * Returns: The address of a @sym_base@ structure, or null if not found
164 * Use: Looks up a symbol in a given symbol table. The name is
165 * passed by the address of its first character. The length
166 * may be given, in which case the name may contain arbitrary
167 * binary data, or it may be given as a negative number, in
168 * which case the length of the name is calculated as
171 * The return value is the address of a pointer to a @sym_base@
172 * block (which may have other things on the end, as above). If
173 * the symbol could be found, the return value points to the
174 * symbol block. If the symbol wasn't there, then if @sz@ is
175 * nonzero, a new symbol is created and its address is returned;
176 * otherwise a null pointer is returned. The exception
177 * @EXC_NOMEM@ is raised if the block can't be allocated.
179 * The value of @*f@ indicates whether a new symbol entry was
180 * created: a nonzero value indicates that an old value was
184 void *sym_find(sym_table
*t
, const char *n
, long l
, size_t sz
, unsigned *f
)
188 hash_base
**bin
, **p
;
191 /* --- Find the correct bin --- */
193 len
= l
< 0 ?
strlen(n
) + 1 : l
;
194 CRC32(hash
, 0, n
, len
);
195 bin
= HASH_BIN(&t
->t
, hash
);
197 /* --- Search the bin list --- */
199 for (p
= bin
; *p
; p
= &(*p
)->next
) {
201 if (hash
== q
->b
.hash
&& len
== q
->len
&& !memcmp(n
, SYM_NAME(q
), len
)) {
203 /* --- Found a match --- *
205 * As a minor, and probably pointless, tweak, move the item to the
206 * front of its bin list.
213 /* --- Return the block --- */
220 /* --- Couldn't find the item there --- */
225 /* --- Create a new symbol block and initialize it --- */
228 TRACK_CTX("new symbol creation");
236 if (len
<= SYM_BUFSZ
)
237 memcpy(q
->name
.b
, n
, len
);
240 q
->name
.p
= sub_alloc(len
);
241 memcpy(q
->name
.p
, n
, len
);
255 /* --- Consider growing the array --- */
259 if (!t
->load
&& hash_extend(&t
->t
))
260 t
->load
= SYM_LIMIT(t
->t
.mask
/ 2 + 1);
262 /* --- Finished that, so return the new symbol block --- */
267 /* --- @sym_remove@ --- *
269 * Arguments: @sym_table *i@ = pointer to a symbol table object
270 * @void *p@ = pointer to symbol table entry
274 * Use: Removes the object from the symbol table. The space occupied
275 * by the object and its name is freed; anything else attached
276 * to the entry should already be gone by this point.
279 void sym_remove(sym_table
*t
, void *p
)
282 hash_remove(&t
->t
, &q
->b
);
283 if (q
->len
> SYM_BUFSZ
)
284 sub_free(q
->name
.p
, q
->len
);
289 /* --- @sym_mkiter@ --- *
291 * Arguments: @sym_iter *i@ = pointer to an iterator object
292 * @sym_table *t@ = pointer to a symbol table object
296 * Use: Creates a new symbol table iterator which may be used to
297 * iterate through a symbol table.
300 void sym_mkiter(sym_iter
*i
, sym_table
*t
) { SYM_MKITER(i
, t
); }
302 /* --- @sym_next@ --- *
304 * Arguments: @sym_iter *i@ = pointer to iterator object
306 * Returns: Pointer to the next symbol found, or null when finished.
308 * Use: Returns the next symbol from the table. Symbols are not
309 * returned in any particular order.
312 void *sym_next(sym_iter
*i
)
319 /*----- Symbol table test code --------------------------------------------*/
326 typedef struct sym_word
{
332 /* --- What it does --- *
334 * Reads the file /usr/dict/words (change to some other file full of
335 * interesting and appropriate bits of text to taste) into a big buffer and
336 * picks apart into lines. Then picks lines at random and enters them into
342 char *buff
, *p
, *lim
;
351 /* --- Initialize for reading the file --- */
354 buff
= xmalloc(sz
+ 1);
358 if ((fp
= fopen("/usr/dict/words", "r")) == 0)
359 fprintf(stderr
, "buggered ;-( (%s)\n", strerror(errno
));
361 /* --- Read buffers of text --- *
363 * Read a buffer. If more to come, double the buffer size and try again.
364 * This is the method I recommended to comp.lang.c, so I may as well try
369 i
= fread(buff
+ done
, 1, sz
- done
, fp
);
374 buff
= xrealloc(buff
, sz
+ 1);
377 /* --- Count the lines --- */
382 for (p
= buff
; p
< lim
; p
++)
383 if (*p
== '\n') sz
++;
385 /* --- Build a table of line starts --- */
387 line
= xmalloc(sz
* sizeof(char *));
390 for (p
= buff
; p
< lim
; p
++)
391 if (*p
== '\n') *p
= 0, line
[i
++] = p
+ 1;
394 /* --- Build a table of lines --- *
396 * This reverses the mapping which the symbol table performs, so that its
397 * accuracy can be tested.
400 flag
= xmalloc(sz
* sizeof(sym_word
*));
401 for (i
= 0; i
< sz
; i
++)
408 i
= (unsigned)rand() % sz
;
415 printf("? %s\n", line
[i
]);
417 w
= sym_find(&tbl
, line
[i
], -1, 0, 0);
419 printf("*** error: find `%s' gave %p not %p\n",
420 line
[i
], (void *)w
, (void *)flag
[i
]);
421 else if (w
&& w
->i
!= i
)
422 printf("*** error: find sym for `%s' gives index %i not %i\n",
430 printf("+ %s\n", line
[i
]);
432 w
= sym_find(&tbl
, line
[i
], -1, sizeof(sym_word
), &f
);
436 printf("*** error: create `%s' gave %p not %p\n",
437 line
[i
], (void *)w
, (void *)flag
[i
]);
438 else if (w
&& w
->i
!= i
)
439 printf("*** error: create sym for `%s' gives index %i not %i\n",
445 printf("*** error: create `%s' gave new block, should be %p\n",
446 line
[i
], (void *)flag
[i
]);
462 v
= (rand() % entries
) == 0;
468 ntbl
= xmalloc(sz
* sizeof(sym_word
*));
469 memcpy(ntbl
, flag
, sz
* sizeof(sym_word
*));
470 sym_mkiter(&it
, &tbl
);
472 while ((w
= sym_next(&it
)) != 0) {
474 printf("*** error: iterate returned duff item %s\n", SYM_NAME(w
));
476 printf(": %s\n", SYM_NAME(w
));
481 for (i
= 0; i
< sz
; i
++)
482 if (ntbl
[i
]) printf("*** error: iterate didn't return item %s\n",
489 int v
= rand() & 255 ?
0 : 1;
494 for (i
= 0; i
<= tbl
.b
.mask
; i
++) {
495 if (!tbl
.b
.v
[i
]) continue;
496 if (v
) printf(" %i: ", i
);
497 b
= (sym_base
*)tbl
.b
.v
[i
];
499 if ((b
->b
.hash
& tbl
.b
.mask
) != i
)
500 printf("*** error: bad hash value found");
501 if (v
) printf("`%s'(%08lx:%lu) ",
502 line
[((sym_word
*)b
)->i
],
504 b
->b
.hash
& tbl
.b
.mask
);
505 b
= (sym_base
*)b
->b
.next
;
507 if (v
) putchar('\n');
513 printf("- %s\n", SYM_NAME(&flag
[i
]->base
));
514 if ((rand() & 1023) == 0) {
515 putchar('-'); fflush(stdout
);
517 sym_remove(&tbl
, flag
[i
]);
531 /*----- That's all, folks -------------------------------------------------*/