]>
git.ipfire.org Git - people/ms/u-boot.git/blob - lib/hashtable.c
2 * This implementation is based on code from uClibc-0.9.30.3 but was
3 * modified and extended for use within U-Boot.
5 * Copyright (C) 2010 Wolfgang Denk <wd@denx.de>
7 * Original license header:
9 * Copyright (C) 1993, 1995, 1996, 1997, 2002 Free Software Foundation, Inc.
10 * This file is part of the GNU C Library.
11 * Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1993.
13 * The GNU C Library is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
18 * The GNU C Library is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with the GNU C Library; if not, write to the Free
25 * Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
32 #ifdef USE_HOSTCC /* HOST build */
38 # define debug(fmt,args...) printf(fmt ,##args)
40 # define debug(fmt,args...)
43 #else /* U-Boot build */
45 # include <linux/string.h>
51 * [Aho,Sethi,Ullman] Compilers: Principles, Techniques and Tools, 1986
52 * [Knuth] The Art of Computer Programming, part 3 (6.4)
56 * The non-reentrant version use a global space for storing the hash table.
58 static struct hsearch_data htab
;
61 * The reentrant version has no static variables to maintain the state.
62 * Instead the interface of all functions is extended to take an argument
63 * which describes the current status.
65 typedef struct _ENTRY
{
76 * For the used double hash method the table size has to be a prime. To
77 * correct the user given table size we need a prime test. This trivial
78 * algorithm is adequate because
79 * a) the code is (most probably) called a few times per program run and
80 * b) the number is small because the table must fit in the core
82 static int isprime(unsigned int number
)
84 /* no even number will be passed */
87 while (div
* div
< number
&& number
% div
!= 0)
90 return number
% div
!= 0;
93 int hcreate(size_t nel
)
95 return hcreate_r(nel
, &htab
);
99 * Before using the hash table we must allocate memory for it.
100 * Test for an existing table are done. We allocate one element
101 * more as the found prime number says. This is done for more effective
102 * indexing as explained in the comment for the hsearch function.
103 * The contents of the table is zeroed, especially the field used
106 int hcreate_r(size_t nel
, struct hsearch_data
*htab
)
108 /* Test for correct arguments. */
114 /* There is still another table active. Return with error. */
115 if (htab
->table
!= NULL
)
118 /* Change nel to the first prime number not smaller as nel. */
119 nel
|= 1; /* make odd */
120 while (!isprime(nel
))
126 /* allocate memory and zero out */
127 htab
->table
= (_ENTRY
*) calloc(htab
->size
+ 1, sizeof(_ENTRY
));
128 if (htab
->table
== NULL
)
131 /* everything went alright */
145 * After using the hash table it has to be destroyed. The used memory can
146 * be freed and the local static variable can be marked as not used.
148 void hdestroy_r(struct hsearch_data
*htab
)
152 /* Test for correct arguments. */
158 /* free used memory */
159 for (i
= 1; i
<= htab
->size
; ++i
) {
160 if (htab
->table
[i
].used
) {
161 ENTRY
*ep
= &htab
->table
[i
].entry
;
169 /* the sign for an existing table is an value != NULL in htable */
178 * This is the search function. It uses double hashing with open addressing.
179 * The argument item.key has to be a pointer to an zero terminated, most
180 * probably strings of chars. The function for generating a number of the
181 * strings is simple but fast. It can be replaced by a more complex function
182 * like ajw (see [Aho,Sethi,Ullman]) if the needs are shown.
184 * We use an trick to speed up the lookup. The table is created by hcreate
185 * with one more element available. This enables us to use the index zero
186 * special. This index will never be used because we store the first hash
187 * index in the field used where zero means not used. Every other value
188 * means used. The used field can be used as a first fast comparison for
189 * equality of the stored and the parameter value. This helps to prevent
190 * unnecessary expensive calls of strcmp.
192 * This implementation differs from the standard library version of
193 * this function in a number of ways:
195 * - While the standard version does not make any assumptions about
196 * the type of the stored data objects at all, this implementation
197 * works with NUL terminated strings only.
198 * - Instead of storing just pointers to the original objects, we
199 * create local copies so the caller does not need to care about the
201 * - The standard implementation does not provide a way to update an
202 * existing entry. This version will create a new entry or update an
203 * existing one when both "action == ENTER" and "item.data != NULL".
204 * - Instead of returning 1 on success, we return the index into the
205 * internal hash table, which is also guaranteed to be positive.
206 * This allows us direct access to the found hash table slot for
207 * example for functions like hdelete().
210 ENTRY
*hsearch(ENTRY item
, ACTION action
)
214 (void) hsearch_r(item
, action
, &result
, &htab
);
219 int hsearch_r(ENTRY item
, ACTION action
, ENTRY
** retval
,
220 struct hsearch_data
*htab
)
224 unsigned int len
= strlen(item
.key
);
227 /* Compute an value for the given string. Perhaps use a better method. */
230 while (count
-- > 0) {
232 hval
+= item
.key
[count
];
236 * First hash function:
237 * simply take the modul but prevent zero.
243 /* The first index tried. */
246 if (htab
->table
[idx
].used
) {
248 * Further action might be required according to the
253 if (htab
->table
[idx
].used
== hval
254 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
255 /* Overwrite existing value? */
256 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
257 free(htab
->table
[idx
].entry
.data
);
258 htab
->table
[idx
].entry
.data
=
260 if (!htab
->table
[idx
].entry
.data
) {
266 /* return found entry */
267 *retval
= &htab
->table
[idx
].entry
;
272 * Second hash function:
273 * as suggested in [Knuth]
275 hval2
= 1 + hval
% (htab
->size
- 2);
279 * Because SIZE is prime this guarantees to
280 * step through all available indices.
283 idx
= htab
->size
+ idx
- hval2
;
288 * If we visited all entries leave the loop
294 /* If entry is found use it. */
295 if ((htab
->table
[idx
].used
== hval
)
296 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
297 /* Overwrite existing value? */
298 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
299 free(htab
->table
[idx
].entry
.data
);
300 htab
->table
[idx
].entry
.data
=
302 if (!htab
->table
[idx
].entry
.data
) {
308 /* return found entry */
309 *retval
= &htab
->table
[idx
].entry
;
313 while (htab
->table
[idx
].used
);
316 /* An empty bucket has been found. */
317 if (action
== ENTER
) {
319 * If table is full and another entry should be
320 * entered return with error.
322 if (htab
->filled
== htab
->size
) {
330 * create copies of item.key and item.data
332 htab
->table
[idx
].used
= hval
;
333 htab
->table
[idx
].entry
.key
= strdup(item
.key
);
334 htab
->table
[idx
].entry
.data
= strdup(item
.data
);
335 if (!htab
->table
[idx
].entry
.key
||
336 !htab
->table
[idx
].entry
.data
) {
344 /* return new entry */
345 *retval
= &htab
->table
[idx
].entry
;
360 * The standard implementation of hsearch(3) does not provide any way
361 * to delete any entries from the hash table. We extend the code to
365 int hdelete(const char *key
)
367 return hdelete_r(key
, &htab
);
370 int hdelete_r(const char *key
, struct hsearch_data
*htab
)
375 debug("hdelete: DELETE key \"%s\"\n", key
);
379 if ((idx
= hsearch_r(e
, FIND
, &ep
, htab
)) == 0) {
381 return 0; /* not found */
384 /* free used ENTRY */
385 debug("hdelete: DELETING key \"%s\"\n", key
);
389 htab
->table
[idx
].used
= 0;
401 * Export the data stored in the hash table in linearized form.
403 * Entries are exported as "name=value" strings, separated by an
404 * arbitrary (non-NUL, of course) separator character. This allows to
405 * use this function both when formatting the U-Boot environment for
406 * external storage (using '\0' as separator), but also when using it
407 * for the "printenv" command to print all variables, simply by using
408 * as '\n" as separator. This can also be used for new features like
409 * exporting the environment data as text file, including the option
410 * for later re-import.
412 * The entries in the result list will be sorted by ascending key
415 * If the separator character is different from NUL, then any
416 * separator characters and backslash characters in the values will
417 * be escaped by a preceeding backslash in output. This is needed for
418 * example to enable multi-line values, especially when the output
419 * shall later be parsed (for example, for re-import).
421 * There are several options how the result buffer is handled:
425 * NULL 0 A string of sufficient length will be allocated.
426 * NULL >0 A string of the size given will be
427 * allocated. An error will be returned if the size is
428 * not sufficient. Any unused bytes in the string will
430 * !NULL 0 The user-supplied buffer will be used. No length
431 * checking will be performed, i. e. it is assumed that
432 * the buffer size will always be big enough. DANGEROUS.
433 * !NULL >0 The user-supplied buffer will be used. An error will
434 * be returned if the size is not sufficient. Any unused
435 * bytes in the string will be '\0'-padded.
438 ssize_t
hexport(const char sep
, char **resp
, size_t size
)
440 return hexport_r(&htab
, sep
, resp
, size
);
443 static int cmpkey(const void *p1
, const void *p2
)
445 ENTRY
*e1
= *(ENTRY
**) p1
;
446 ENTRY
*e2
= *(ENTRY
**) p2
;
448 return (strcmp(e1
->key
, e2
->key
));
451 ssize_t
hexport_r(struct hsearch_data
*htab
, const char sep
,
452 char **resp
, size_t size
)
454 ENTRY
*list
[htab
->size
];
459 /* Test for correct arguments. */
460 if ((resp
== NULL
) || (htab
== NULL
)) {
465 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
466 htab
, htab
->size
, htab
->filled
, size
);
469 * search used entries,
470 * save addresses and compute total length
472 for (i
= 1, n
= 0, totlen
= 0; i
<= htab
->size
; ++i
) {
474 if (htab
->table
[i
].used
) {
475 ENTRY
*ep
= &htab
->table
[i
].entry
;
479 totlen
+= strlen(ep
->key
) + 2;
482 totlen
+= strlen(ep
->data
);
483 } else { /* check if escapes are needed */
488 /* add room for needed escape chars */
489 if ((*s
== sep
) || (*s
== '\\'))
494 totlen
+= 2; /* for '=' and 'sep' char */
499 /* Pass 1a: print unsorted list */
500 printf("Unsorted: n=%d\n", n
);
501 for (i
= 0; i
< n
; ++i
) {
502 printf("\t%3d: %p ==> %-10s => %s\n",
503 i
, list
[i
], list
[i
]->key
, list
[i
]->data
);
507 /* Sort list by keys */
508 qsort(list
, n
, sizeof(ENTRY
*), cmpkey
);
510 /* Check if the user supplied buffer size is sufficient */
512 if (size
< totlen
+ 1) { /* provided buffer too small */
513 debug("### buffer too small: %d, but need %d\n",
522 /* Check if the user provided a buffer */
526 memset(res
, '\0', size
);
528 /* no, allocate and clear one */
529 *resp
= res
= calloc(1, size
);
537 * export sorted list of result data
539 for (i
= 0, p
= res
; i
< n
; ++i
) {
550 if ((*s
== sep
) || (*s
== '\\'))
551 *p
++ = '\\'; /* escape */
556 *p
= '\0'; /* terminate result */
567 * Import linearized data into hash table.
569 * This is the inverse function to hexport(): it takes a linear list
570 * of "name=value" pairs and creates hash table entries from it.
572 * Entries without "value", i. e. consisting of only "name" or
573 * "name=", will cause this entry to be deleted from the hash table.
575 * The "flag" argument can be used to control the behaviour: when the
576 * H_NOCLEAR bit is set, then an existing hash table will kept, i. e.
577 * new data will be added to an existing hash table; otherwise, old
578 * data will be discarded and a new hash table will be created.
580 * The separator character for the "name=value" pairs can be selected,
581 * so we both support importing from externally stored environment
582 * data (separated by NUL characters) and from plain text files
583 * (entries separated by newline characters).
585 * To allow for nicely formatted text input, leading white space
586 * (sequences of SPACE and TAB chars) is ignored, and entries starting
587 * (after removal of any leading white space) with a '#' character are
588 * considered comments and ignored.
590 * [NOTE: this means that a variable name cannot start with a '#'
593 * When using a non-NUL separator character, backslash is used as
594 * escape character in the value part, allowing for example for
597 * In theory, arbitrary separator characters can be used, but only
598 * '\0' and '\n' have really been tested.
601 int himport(const char *env
, size_t size
, const char sep
, int flag
)
603 return himport_r(&htab
, env
, size
, sep
, flag
);
606 int himport_r(struct hsearch_data
*htab
,
607 const char *env
, size_t size
, const char sep
, int flag
)
609 char *data
, *sp
, *dp
, *name
, *value
;
611 /* Test for correct arguments. */
617 /* we allocate new space to make sure we can write to the array */
618 if ((data
= malloc(size
)) == NULL
) {
619 debug("himport_r: can't malloc %d bytes\n", size
);
623 memcpy(data
, env
, size
);
626 if ((flag
& H_NOCLEAR
) == 0) {
627 /* Destroy old hash table if one exists */
628 debug("Destroy Hash Table: %p table = %p\n", htab
,
635 * Create new hash table (if needed). The computation of the hash
636 * table size is based on heuristics: in a sample of some 70+
637 * existing systems we found an average size of 39+ bytes per entry
638 * in the environment (for the whole key=value pair). Assuming a
639 * size of 7 per entry (= safety factor of >5) should provide enough
640 * safety margin for any existing environment definitons and still
641 * allow for more than enough dynamic additions. Note that the
642 * "size" argument is supposed to give the maximum enviroment size
649 debug("Create Hash Table: N=%d\n", nent
);
651 if (hcreate_r(nent
, htab
) == 0) {
657 /* Parse environment; allow for '\0' and 'sep' as separators */
661 /* skip leading white space */
662 while ((*dp
== ' ') || (*dp
== '\t'))
665 /* skip comment lines */
667 while (*dp
&& (*dp
!= sep
))
674 for (name
= dp
; *dp
!= '=' && *dp
&& *dp
!= sep
; ++dp
)
677 /* deal with "name" and "name=" entries (delete var) */
678 if (*dp
== '\0' || *(dp
+ 1) == '\0' ||
679 *dp
== sep
|| *(dp
+ 1) == sep
) {
682 *dp
++ = '\0'; /* terminate name */
684 debug("DELETE CANDIDATE: \"%s\"\n", name
);
686 if (hdelete_r(name
, htab
) == 0)
687 debug("DELETE ERROR ##############################\n");
691 *dp
++ = '\0'; /* terminate name */
693 /* parse value; deal with escapes */
694 for (value
= sp
= dp
; *dp
&& (*dp
!= sep
); ++dp
) {
695 if ((*dp
== '\\') && *(dp
+ 1))
699 *sp
++ = '\0'; /* terminate value */
702 /* enter into hash table */
706 hsearch_r(e
, ENTER
, &rv
, htab
);
708 printf("himport_r: can't insert \"%s=%s\" into hash table\n", name
, value
);
712 debug("INSERT: %p ==> name=\"%s\" value=\"%s\"\n", rv
, name
,
714 debug(" table = %p, size = %d, filled = %d\n", htab
,
715 htab
->size
, htab
->filled
);
716 } while ((dp
< data
+ size
) && *dp
); /* size check needed for text */
717 /* without '\0' termination */
720 return 1; /* everything OK */