3 * $Id: sym.c,v 1.1 1998/06/17 23:44:42 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 General Public License as published by
16 * the Free Software Foundation; either version 2 of the License, or
17 * (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 General Public License for more details.
24 * You should have received a copy of the GNU General Public License
25 * along with mLib; if not, write to the Free Software Foundation,
26 * Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
29 /*----- Revision history --------------------------------------------------*
32 * Revision 1.1 1998/06/17 23:44:42 mdw
37 /*----- Header files ------------------------------------------------------*/
39 /* --- ANSI headers --- */
45 /* --- Local headers --- */
54 /*----- Tuning parameters -------------------------------------------------*/
56 /* --- Initial hash table size --- *
58 * This is the initial @mask@ value. It must be of the form %$2^n - 1$%,
59 * so that it can be used to mask of the bottom bits of a hash value.
62 #define SYM_INITSZ 255 /* Size of a new hash table */
64 /* --- Maximum load factor --- *
66 * This parameter controls how much the table has to be loaded before the
67 * table is extended. The number of elements %$n$%, the number of bins %$b$%
68 * and the limit %$l$% satisfy the relation %$n < bl$%; if a new item is
69 * added to the table and this relation is found to be false, the table is
72 * The current function gives %$l = {3n \over 4}$%, which appears to be
73 * reasonable on the face of things.
76 #define SYM_LIMIT(n) (((n) * 3) >> 2) /* Load factor for growing table */
78 /*----- Useful macros -----------------------------------------------------*/
80 #define SYM_NAME(sy) \
81 ((sy)->len > SYM_BUFSZ ? (sy)->name.p : (sy)->name.b)
83 /*----- Main code ---------------------------------------------------------*/
85 /* --- @sym_createTable@ --- *
87 * Arguments: @sym_table *t@ = symbol table to initialise
91 * Use: Initialises the given symbol table. Raises @EXC_NOMEM@ if
92 * there isn't enough memory.
95 void sym_createTable(sym_table
*t
)
99 TRACK_CTX("symbol table creation");
102 t
->mask
= SYM_INITSZ
;
103 t
->c
= SYM_LIMIT(SYM_INITSZ
);
104 t
->a
= xmalloc((t
->mask
+ 1) * sizeof(sym_base
*));
106 for (i
= 0; i
< SYM_INITSZ
+ 1; i
++)
112 /* --- @sym_destroyTable@ --- *
114 * Arguments: @sym_table *t@ = pointer to symbol table in question
118 * Use: Destroys a symbol table, freeing all the memory it used to
122 void sym_destroyTable(sym_table
*t
)
127 TRACK_CTX("symbol table deletion");
130 for (i
= 0; i
<= t
->mask
; i
++) {
134 if (p
->len
> SYM_BUFSZ
)
135 sub_free(p
->name
.p
, p
->len
);
145 /* --- @sym_find@ --- *
147 * Arguments: @sym_table *t@ = pointer to symbol table in question
148 * @const char *n@ = pointer to symbol table to look up
149 * @long l@ = length of the name string or negative to measure
150 * @size_t sz@ = size of desired symbol object, or zero
151 * @unsigned *f@ = pointer to a flag, or null.
153 * Returns: The address of a @sym_base@ structure, or null if not found
156 * Use: Looks up a symbol in a given symbol table. The name is
157 * passed by the address of its first character. The length
158 * may be given, in which case the name may contain arbitrary
159 * binary data, or it may be given as a negative number, in
160 * which case the length of the name is calculated as
163 * The return value is the address of a pointer to a @sym_base@
164 * block (which may have other things on the end, as above). If
165 * the symbol could be found, the return value points to the
166 * symbol block. If the symbol wasn't there, then if @sz@ is
167 * nonzero, a new symbol is created and its address is returned;
168 * otherwise a null pointer is returned. The exception
169 * @EXC_NOMEM@ is raised if the block can't be allocated.
171 * The value of @*f@ indicates whether a new symbol entry was
172 * created: a nonzero value indicates that an old value was
176 void *sym_find(sym_table
*t
, const char *n
, long l
, size_t sz
, unsigned *f
)
178 unsigned long hash
; /* Hash value for user's name */
179 size_t len
= l
< 0 ?
strlen(n
) + 1 : l
; /* Find length of user's name */
180 sym_base
*bin
; /* Bin containing our item */
181 sym_base
*p
, *q
; /* Pointer wandering through list */
183 /* --- Find the correct bin --- */
185 CRC32(hash
, 0, n
, len
); /* Find hash value for this name */
186 bin
= p
= (sym_base
*)(t
->a
+ (hash
& t
->mask
));
188 /* --- Search the bin list --- */
191 if (hash
== p
->next
->hash
&& /* Check full hash values first */
192 len
== p
->next
->len
&& /* Then compare string lengths */
193 !memcmp(n
, SYM_NAME(p
->next
), len
)) /* And finally compare names */
195 /* --- Found a match --- *
197 * As a minor, and probably pointless, tweak, move the item to the
198 * front of its bin list.
201 q
= p
->next
; /* Find the actual symbol block */
202 p
->next
= q
->next
; /* Extract block from bin list */
203 q
->next
= bin
->next
; /* Set up symbol's next pointer */
204 bin
->next
= q
; /* And reinsert the block */
206 /* --- Return the block --- */
208 if (f
) *f
= 1; /* Didn't fail to find the item */
209 return (q
); /* And return the block */
212 p
= p
->next
; /* Move onto the next item */
215 /* --- Couldn't find the item there --- */
217 if (f
) *f
= 0; /* Failed to find the block */
218 if (!sz
) return (0); /* Return zero if not creating */
220 /* --- Create a new symbol block and initialise it --- */
223 TRACK_CTX("new symbol creation");
226 p
= xmalloc(sz
); /* Create a new symbol block */
227 p
->next
= bin
->next
; /* Set up the next pointer */
228 p
->hash
= hash
; /* Set up the hash value too */
229 p
->len
= len
; /* And set up the string length */
230 if (len
<= SYM_BUFSZ
)
231 memcpy(p
->name
.b
, n
, len
); /* And copy the string over */
234 p
->name
.p
= sub_alloc(len
); /* Allocate a block for the name */
235 memcpy(p
->name
.p
, n
, len
); /* And copy the string over */
246 bin
->next
= p
; /* Put the pointer into the bin */
248 /* --- Consider growing the array --- */
251 unsigned long m
= t
->mask
+ 1; /* Next set bit in has word */
252 sym_base
*p
, *q
, *r
; /* More useful pointers */
253 size_t i
, lim
; /* Loop counter and limit */
255 TRACK_CTX("symbol table extension");
258 /* --- Update values in the anchor block --- */
261 t
->a
= xrealloc(t
->a
, (t
->mask
+ 1) * 2 * sizeof(sym_base
*));
262 } CATCH
switch (exc_type
) {
271 t
->c
= SYM_LIMIT(t
->mask
+ 1); /* Set load value */
272 t
->mask
= (t
->mask
+ 1) * 2 - 1; /* Set the new mask value */
274 /* --- Now wander through the table rehashing things --- *
276 * This loop is very careful to avoid problems with aliasing. The items
277 * are dealt with from the end backwards to avoid overwriting bins before
278 * they've been processed.
281 lim
= (t
->mask
+ 1) >> 1;
282 for (i
= 0; i
< lim
; i
++) {
284 /* --- Some initialisation --- */
286 r
= t
->a
[i
]; /* Find the list we're dissecting */
287 p
= (sym_base
*)(t
->a
+ i
); /* Find bit-clear list */
288 q
= (sym_base
*)(t
->a
+ i
+ lim
); /* And the bit-set lsit */
290 /* --- Now go through the @r@ list --- */
293 if (r
->hash
& m
) /* Is the next bit set? */
294 q
= q
->next
= r
; /* Yes, so fit up previous link */
296 p
= p
->next
= r
; /* No, so fit up previous link */
297 r
= r
->next
; /* Move onto the next item */
299 p
->next
= q
->next
= 0; /* Null terminate the new lists */
305 /* --- Finished that, so return the new symbol block --- */
310 /* --- @sym_remove@ --- *
312 * Arguments: @sym_table *i@ = pointer to a symbol table object
313 * @void *b@ = pointer to symbol table entry
317 * Use: Removes the object from the symbol table. The space occupied
318 * by the object and its name is freed; anything else attached
319 * to the entry should already be gone by this point.
322 void sym_remove(sym_table
*t
, void *b
)
324 /* --- A quick comment --- *
326 * Since the @sym_base@ block contains the hash, finding the element in the
327 * bin list is really quick -- it's not worth bothering with things like
328 * doubly linked lists.
332 sym_base
*bin
= (sym_base
*)(t
->a
+ (p
->hash
& t
->mask
));
334 /* --- Find the item in the bin list --- */
344 /* --- Now just remove the item from the list and free it --- *
346 * Oh, and bump the load counter.
350 if (p
->len
> SYM_BUFSZ
)
351 sub_free(p
->name
.p
, p
->len
);
356 /* --- @sym_createIter@ --- *
358 * Arguments: @sym_iter *i@ = pointer to an iterator object
359 * @sym_table *t@ = pointer to a symbol table object
363 * Use: Creates a new symbol table iterator which may be used to
364 * iterate through a symbol table.
367 void sym_createIter(sym_iter
*i
, sym_table
*t
)
374 /* --- @sym_next@ --- *
376 * Arguments: @sym_iter *i@ = pointer to iterator object
378 * Returns: Pointer to the next symbol found, or null when finished.
380 * Use: Returns the next symbol from the table. Symbols are not
381 * returned in any particular order.
384 void *sym_next(sym_iter
*i
)
388 /* --- Find the next item --- */
391 if (i
->i
> i
->t
->mask
)
393 i
->n
= i
->t
->a
[i
->i
++];
396 /* --- Update the iterator block --- */
406 /*----- Symbol table test code --------------------------------------------*/
413 typedef struct sym_word
{
419 /* --- What it does --- *
421 * Reads the file /usr/dict/words (change to some other file full of
422 * interesting and appropriate bits of text to taste) into a big buffer and
423 * picks apart into lines. Then picks lines at random and enters them into
429 char *buff
, *p
, *lim
;
438 /* --- Initialise for reading the file --- */
441 buff
= xmalloc(sz
+ 1);
445 if ((fp
= fopen("/usr/dict/words", "r")) == 0)
446 fprintf(stderr
, "buggered ;-( (%s)\n", strerror(errno
));
448 /* --- Read buffers of text --- *
450 * Read a buffer. If more to come, double the buffer size and try again.
451 * This is the method I recommended to comp.lang.c, so I may as well try
456 i
= fread(buff
+ done
, 1, sz
- done
, fp
);
461 buff
= xrealloc(buff
, sz
+ 1);
464 /* --- Count the lines --- */
469 for (p
= buff
; p
< lim
; p
++)
470 if (*p
== '\n') sz
++;
472 /* --- Build a table of line starts --- */
474 line
= xmalloc(sz
* sizeof(char *));
477 for (p
= buff
; p
< lim
; p
++)
478 if (*p
== '\n') *p
= 0, line
[i
++] = p
+ 1;
481 /* --- Build a table of lines --- *
483 * This reverses the mapping which the symbol table performs, so that its
484 * accuracy can be tested.
487 flag
= xmalloc(sz
* sizeof(sym_word
*));
488 for (i
= 0; i
< sz
; i
++)
492 sym_createTable(&tbl
);
495 i
= (unsigned)rand() % sz
;
502 printf("find `%s'\n", line
[i
]);
503 if ((rand() & 1023) == 0) {
504 putchar('.'); fflush(stdout
);
507 w
= sym_find(&tbl
, line
[i
], -1, 0, 0);
509 printf("*** error: find `%s' gave %p not %p\n",
510 line
[i
], (void *)w
, (void *)flag
[i
]);
511 else if (w
&& w
->i
!= i
)
512 printf("*** error: find sym for `%s' gives index %i not %i\n",
520 printf("create `%s'\n", line
[i
]);
521 if ((rand() & 1023) == 0) {
522 putchar('+'); fflush(stdout
);
525 w
= sym_find(&tbl
, line
[i
], -1, sizeof(sym_word
), &f
);
529 printf("*** error: create `%s' gave %p not %p\n",
530 line
[i
], (void *)w
, (void *)flag
[i
]);
531 else if (w
&& w
->i
!= i
)
532 printf("*** error: create sym for `%s' gives index %i not %i\n",
538 printf("*** error: create `%s' gave new block, should be %p\n",
539 line
[i
], (void *)flag
[i
]);
555 v
= (rand() % entries
) == 0;
558 printf("\niterated %i entries\n", entries
);
563 ntbl
= xmalloc(sz
* sizeof(sym_word
*));
564 memcpy(ntbl
, flag
, sz
* sizeof(sym_word
*));
565 sym_createIter(&it
, &tbl
);
567 while ((w
= sym_next(&it
)) != 0) {
569 printf("*** error: iterate returned duff item %i\n", w
->i
);
574 for (i
= 0; i
< sz
; i
++)
575 if (ntbl
[i
]) printf("*** error: iterate didn't return item %i\n",
582 int v
= rand() & 255 ?
0 : 1;
587 for (i
= 0; i
<= tbl
.mask
; i
++) {
588 if (!tbl
.a
[i
]) continue;
589 if (v
) printf(" %i: ", i
);
592 if ((b
->hash
& tbl
.mask
) != i
)
593 printf("*** error: bad hash value found");
594 if (v
) printf("`%s'(%08lx:%lu) ",
595 line
[((sym_word
*)b
)->i
],
600 if (v
) putchar('\n');
606 printf("remove `%s'\n", SYM_NAME(&flag
[i
]->base
));
607 if ((rand() & 1023) == 0) {
608 putchar('-'); fflush(stdout
);
610 sym_remove(&tbl
, flag
[i
]);
624 /*----- That's all, folks -------------------------------------------------*/