3 * $Id: sym.c,v 1.5 1999/05/13 22:48:37 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 1.5 1999/05/13 22:48:37 mdw
34 * Twiddle the extension threshold. Change `-ise' to `-ize' throughout.
36 * Revision 1.4 1999/05/06 19:51:35 mdw
37 * Reformatted the LGPL notice a little bit.
39 * Revision 1.3 1999/05/05 18:50:31 mdw
40 * Change licensing conditions to LGPL.
42 * Revision 1.2 1998/11/26 19:27:33 mdw
43 * Move SYM_NAME into the header file. Fix bugs.
45 * Revision 1.1.1.1 1998/06/17 23:44:42 mdw
46 * Initial version of mLib
50 /*----- Header files ------------------------------------------------------*/
52 /* --- ANSI headers --- */
58 /* --- Local headers --- */
67 /*----- Tuning parameters -------------------------------------------------*/
69 /* --- Initial hash table size --- *
71 * This is the initial @mask@ value. It must be of the form %$2^n - 1$%,
72 * so that it can be used to mask of the bottom bits of a hash value.
75 #define SYM_INITSZ 255 /* Size of a new hash table */
77 /* --- Maximum load factor --- *
79 * This parameter controls how much the table has to be loaded before the
80 * table is extended. The number of elements %$n$%, the number of bins %$b$%
81 * and the limit %$l$% satisfy the relation %$n < bl$%; if a new item is
82 * added to the table and this relation is found to be false, the table is
86 #define SYM_LIMIT(n) ((n) * 4) /* Load factor for growing table */
88 /*----- Main code ---------------------------------------------------------*/
90 /* --- @sym_createTable@ --- *
92 * Arguments: @sym_table *t@ = symbol table to initialize
96 * Use: Initializes the given symbol table. Raises @EXC_NOMEM@ if
97 * there isn't enough memory.
100 void sym_createTable(sym_table
*t
)
104 TRACK_CTX("symbol table creation");
107 t
->mask
= SYM_INITSZ
;
108 t
->c
= SYM_LIMIT(SYM_INITSZ
);
109 t
->a
= xmalloc((t
->mask
+ 1) * sizeof(sym_base
*));
111 for (i
= 0; i
< SYM_INITSZ
+ 1; i
++)
117 /* --- @sym_destroyTable@ --- *
119 * Arguments: @sym_table *t@ = pointer to symbol table in question
123 * Use: Destroys a symbol table, freeing all the memory it used to
127 void sym_destroyTable(sym_table
*t
)
132 TRACK_CTX("symbol table deletion");
135 for (i
= 0; i
<= t
->mask
; i
++) {
139 if (p
->len
> SYM_BUFSZ
)
140 sub_free(p
->name
.p
, p
->len
);
150 /* --- @sym_find@ --- *
152 * Arguments: @sym_table *t@ = pointer to symbol table in question
153 * @const char *n@ = pointer to symbol table to look up
154 * @long l@ = length of the name string or negative to measure
155 * @size_t sz@ = size of desired symbol object, or zero
156 * @unsigned *f@ = pointer to a flag, or null.
158 * Returns: The address of a @sym_base@ structure, or null if not found
161 * Use: Looks up a symbol in a given symbol table. The name is
162 * passed by the address of its first character. The length
163 * may be given, in which case the name may contain arbitrary
164 * binary data, or it may be given as a negative number, in
165 * which case the length of the name is calculated as
168 * The return value is the address of a pointer to a @sym_base@
169 * block (which may have other things on the end, as above). If
170 * the symbol could be found, the return value points to the
171 * symbol block. If the symbol wasn't there, then if @sz@ is
172 * nonzero, a new symbol is created and its address is returned;
173 * otherwise a null pointer is returned. The exception
174 * @EXC_NOMEM@ is raised if the block can't be allocated.
176 * The value of @*f@ indicates whether a new symbol entry was
177 * created: a nonzero value indicates that an old value was
181 void *sym_find(sym_table
*t
, const char *n
, long l
, size_t sz
, unsigned *f
)
183 unsigned long hash
; /* Hash value for user's name */
184 size_t len
= l
< 0 ?
strlen(n
) + 1 : l
; /* Find length of user's name */
185 sym_base
*bin
; /* Bin containing our item */
186 sym_base
*p
, *q
; /* Pointer wandering through list */
188 /* --- Find the correct bin --- */
190 CRC32(hash
, 0, n
, len
); /* Find hash value for this name */
191 bin
= p
= (sym_base
*)(t
->a
+ (hash
& t
->mask
));
193 /* --- Search the bin list --- */
196 if (hash
== p
->next
->hash
&& /* Check full hash values first */
197 len
== p
->next
->len
&& /* Then compare string lengths */
198 !memcmp(n
, SYM_NAME(p
->next
), len
)) /* And finally compare names */
200 /* --- Found a match --- *
202 * As a minor, and probably pointless, tweak, move the item to the
203 * front of its bin list.
206 q
= p
->next
; /* Find the actual symbol block */
207 p
->next
= q
->next
; /* Extract block from bin list */
208 q
->next
= bin
->next
; /* Set up symbol's next pointer */
209 bin
->next
= q
; /* And reinsert the block */
211 /* --- Return the block --- */
213 if (f
) *f
= 1; /* Didn't fail to find the item */
214 return (q
); /* And return the block */
217 p
= p
->next
; /* Move onto the next item */
220 /* --- Couldn't find the item there --- */
222 if (f
) *f
= 0; /* Failed to find the block */
223 if (!sz
) return (0); /* Return zero if not creating */
225 /* --- Create a new symbol block and initialize it --- */
228 TRACK_CTX("new symbol creation");
231 p
= xmalloc(sz
); /* Create a new symbol block */
232 p
->next
= bin
->next
; /* Set up the next pointer */
233 p
->hash
= hash
; /* Set up the hash value too */
234 p
->len
= len
; /* And set up the string length */
235 if (len
<= SYM_BUFSZ
)
236 memcpy(p
->name
.b
, n
, len
); /* And copy the string over */
239 p
->name
.p
= sub_alloc(len
); /* Allocate a block for the name */
240 memcpy(p
->name
.p
, n
, len
); /* And copy the string over */
251 bin
->next
= p
; /* Put the pointer into the bin */
253 /* --- Consider growing the array --- */
258 unsigned long m
= t
->mask
+ 1; /* Next set bit in has word */
259 sym_base
*p
, *q
, *r
; /* More useful pointers */
260 size_t i
, lim
; /* Loop counter and limit */
262 TRACK_CTX("symbol table extension");
265 /* --- Update values in the anchor block --- */
268 t
->a
= xrealloc(t
->a
, (t
->mask
+ 1) * 2 * sizeof(sym_base
*));
269 } CATCH
switch (exc_type
) {
278 t
->c
= SYM_LIMIT(t
->mask
+ 1); /* Set load value */
279 t
->mask
= (t
->mask
+ 1) * 2 - 1; /* Set the new mask value */
281 /* --- Now wander through the table rehashing things --- *
283 * This loop is very careful to avoid problems with aliasing. The items
284 * are dealt with from the end backwards to avoid overwriting bins before
285 * they've been processed.
288 lim
= (t
->mask
+ 1) >> 1;
289 for (i
= 0; i
< lim
; i
++) {
291 /* --- Some initialisation --- */
293 r
= t
->a
[i
]; /* Find the list we're dissecting */
294 p
= (sym_base
*)(t
->a
+ i
); /* Find bit-clear list */
295 q
= (sym_base
*)(t
->a
+ i
+ lim
); /* And the bit-set lsit */
297 /* --- Now go through the @r@ list --- */
300 if (r
->hash
& m
) /* Is the next bit set? */
301 q
= q
->next
= r
; /* Yes, so fit up previous link */
303 p
= p
->next
= r
; /* No, so fit up previous link */
304 r
= r
->next
; /* Move onto the next item */
306 p
->next
= q
->next
= 0; /* Null terminate the new lists */
312 /* --- Finished that, so return the new symbol block --- */
317 /* --- @sym_remove@ --- *
319 * Arguments: @sym_table *i@ = pointer to a symbol table object
320 * @void *b@ = pointer to symbol table entry
324 * Use: Removes the object from the symbol table. The space occupied
325 * by the object and its name is freed; anything else attached
326 * to the entry should already be gone by this point.
329 void sym_remove(sym_table
*t
, void *b
)
331 /* --- A quick comment --- *
333 * Since the @sym_base@ block contains the hash, finding the element in the
334 * bin list is really quick -- it's not worth bothering with things like
335 * doubly linked lists.
339 sym_base
*bin
= (sym_base
*)(t
->a
+ (p
->hash
& t
->mask
));
341 /* --- Find the item in the bin list --- */
351 /* --- Now just remove the item from the list and free it --- *
353 * Oh, and bump the load counter.
357 if (p
->len
> SYM_BUFSZ
)
358 sub_free(p
->name
.p
, p
->len
);
363 /* --- @sym_createIter@ --- *
365 * Arguments: @sym_iter *i@ = pointer to an iterator object
366 * @sym_table *t@ = pointer to a symbol table object
370 * Use: Creates a new symbol table iterator which may be used to
371 * iterate through a symbol table.
374 void sym_createIter(sym_iter
*i
, sym_table
*t
)
381 /* --- @sym_next@ --- *
383 * Arguments: @sym_iter *i@ = pointer to iterator object
385 * Returns: Pointer to the next symbol found, or null when finished.
387 * Use: Returns the next symbol from the table. Symbols are not
388 * returned in any particular order.
391 void *sym_next(sym_iter
*i
)
395 /* --- Find the next item --- */
398 if (i
->i
> i
->t
->mask
)
400 i
->n
= i
->t
->a
[i
->i
++];
403 /* --- Update the iterator block --- */
413 /*----- Symbol table test code --------------------------------------------*/
420 typedef struct sym_word
{
426 /* --- What it does --- *
428 * Reads the file /usr/dict/words (change to some other file full of
429 * interesting and appropriate bits of text to taste) into a big buffer and
430 * picks apart into lines. Then picks lines at random and enters them into
436 char *buff
, *p
, *lim
;
445 /* --- Initialize for reading the file --- */
448 buff
= xmalloc(sz
+ 1);
452 if ((fp
= fopen("/usr/dict/words", "r")) == 0)
453 fprintf(stderr
, "buggered ;-( (%s)\n", strerror(errno
));
455 /* --- Read buffers of text --- *
457 * Read a buffer. If more to come, double the buffer size and try again.
458 * This is the method I recommended to comp.lang.c, so I may as well try
463 i
= fread(buff
+ done
, 1, sz
- done
, fp
);
468 buff
= xrealloc(buff
, sz
+ 1);
471 /* --- Count the lines --- */
476 for (p
= buff
; p
< lim
; p
++)
477 if (*p
== '\n') sz
++;
479 /* --- Build a table of line starts --- */
481 line
= xmalloc(sz
* sizeof(char *));
484 for (p
= buff
; p
< lim
; p
++)
485 if (*p
== '\n') *p
= 0, line
[i
++] = p
+ 1;
488 /* --- Build a table of lines --- *
490 * This reverses the mapping which the symbol table performs, so that its
491 * accuracy can be tested.
494 flag
= xmalloc(sz
* sizeof(sym_word
*));
495 for (i
= 0; i
< sz
; i
++)
499 sym_createTable(&tbl
);
502 i
= (unsigned)rand() % sz
;
509 printf("find `%s'\n", line
[i
]);
510 if ((rand() & 1023) == 0) {
511 putchar('.'); fflush(stdout
);
514 w
= sym_find(&tbl
, line
[i
], -1, 0, 0);
516 printf("*** error: find `%s' gave %p not %p\n",
517 line
[i
], (void *)w
, (void *)flag
[i
]);
518 else if (w
&& w
->i
!= i
)
519 printf("*** error: find sym for `%s' gives index %i not %i\n",
527 printf("create `%s'\n", line
[i
]);
528 if ((rand() & 1023) == 0) {
529 putchar('+'); fflush(stdout
);
532 w
= sym_find(&tbl
, line
[i
], -1, sizeof(sym_word
), &f
);
536 printf("*** error: create `%s' gave %p not %p\n",
537 line
[i
], (void *)w
, (void *)flag
[i
]);
538 else if (w
&& w
->i
!= i
)
539 printf("*** error: create sym for `%s' gives index %i not %i\n",
545 printf("*** error: create `%s' gave new block, should be %p\n",
546 line
[i
], (void *)flag
[i
]);
562 v
= (rand() % entries
) == 0;
565 printf("\niterated %i entries\n", entries
);
570 ntbl
= xmalloc(sz
* sizeof(sym_word
*));
571 memcpy(ntbl
, flag
, sz
* sizeof(sym_word
*));
572 sym_createIter(&it
, &tbl
);
574 while ((w
= sym_next(&it
)) != 0) {
576 printf("*** error: iterate returned duff item %i\n", w
->i
);
581 for (i
= 0; i
< sz
; i
++)
582 if (ntbl
[i
]) printf("*** error: iterate didn't return item %i\n",
589 int v
= rand() & 255 ?
0 : 1;
594 for (i
= 0; i
<= tbl
.mask
; i
++) {
595 if (!tbl
.a
[i
]) continue;
596 if (v
) printf(" %i: ", i
);
599 if ((b
->hash
& tbl
.mask
) != i
)
600 printf("*** error: bad hash value found");
601 if (v
) printf("`%s'(%08lx:%lu) ",
602 line
[((sym_word
*)b
)->i
],
607 if (v
) putchar('\n');
613 printf("remove `%s'\n", SYM_NAME(&flag
[i
]->base
));
614 if ((rand() & 1023) == 0) {
615 putchar('-'); fflush(stdout
);
617 sym_remove(&tbl
, flag
[i
]);
631 /*----- That's all, folks -------------------------------------------------*/