/* Hash tables.
Copyright (C) 2000, 2001 Free Software Foundation, Inc.
-This file is part of Wget.
+This file is part of GNU Wget.
-This program is free software; you can redistribute it and/or modify
+GNU Wget is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or (at
your option) any later version.
-This program is distributed in the hope that it will be useful,
+GNU Wget is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
-along with this program; if not, write to the Free Software
-Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
+along with Wget; if not, write to the Free Software
+Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+
+In addition, as a special exception, the Free Software Foundation
+gives permission to link the code of its release of Wget with the
+OpenSSL project's "OpenSSL" library (or with modified versions of it
+that use the same license as the "OpenSSL" library), and distribute
+the linked executables. You must obey the GNU General Public License
+in all respects for all of the code used other than "OpenSSL". If you
+modify this file, you may extend this exception to your version of the
+file, but you are not obligated to do so. If you do not wish to do
+so, delete this exception statement from your version. */
#ifdef HAVE_CONFIG_H
# include <config.h>
# define xmalloc malloc
# define xrealloc realloc
# define xfree free
+
+# undef TOLOWER
+# define TOLOWER(x) ('A' <= (x) && (x) <= 'Z' ? (x) - 32 : (x))
#endif
/* INTERFACE:
Hash tables are an implementation technique used to implement
- mapping between objects. Provided a good hashing function is used,
- they guarantee constant-time access and storing of information.
+ mapping between objects. Assuming a good hashing function is used,
+ they provide near-constant-time access and storing of information.
Duplicate keys are not allowed.
- The basics are all covered. hash_table_new creates a hash table,
- and hash_table_destroy deletes it. hash_table_put establishes a
- mapping between a key and a value. hash_table_get retrieves the
- value that corresponds to a key. hash_table_exists queries whether
- a key is stored in a table at all. hash_table_remove removes a
- mapping that corresponds to a key. hash_table_map allows you to
- map through all the entries in a hash table. hash_table_clear
- clears all the entries from the hash table.
+ This file defines the following entry points: hash_table_new
+ creates a hash table, and hash_table_destroy deletes it.
+ hash_table_put establishes a mapping between a key and a value.
+ hash_table_get retrieves the value that corresponds to a key.
+ hash_table_contains queries whether a key is stored in a table at
+ all. hash_table_remove removes a mapping that corresponds to a
+ key. hash_table_map allows you to map through all the entries in a
+ hash table. hash_table_clear clears all the entries from the hash
+ table.
The number of mappings in a table is not limited, except by the
amount of memory. As you add new elements to a table, it regrows
as necessary. If you have an idea about how many elements you will
store, you can provide a hint to hash_table_new().
- The hashing and equality functions are normally provided by the
- user. For the special (and frequent) case of hashing strings, you
- can use the pre-canned make_string_hash_table(), which provides an
- efficient string hashing function, and a string equality wrapper
- around strcmp().
+ The hashing and equality functions depend on the type of key and
+ are normally provided by the user. For the special (and frequent)
+ case of using string keys, you can use the pre-canned
+ make_string_hash_table(), which provides an efficient string
+ hashing function, and a string equality wrapper around strcmp().
- When specifying your own hash and test functions, make sure the
+ When specifying your hash and test functions, make sure the
following holds true:
- The test function returns non-zero for keys that are considered
before calculating a hash. That way you have easily guaranteed
that case differences will not result in a different hash.
- - (optional) Choose the hash function to get as good "spreading" as
- possible. A good hash function will react to even a small change
- in its input with a completely different resulting hash.
- Finally, don't make your hash function extremely slow, because
- you're then defeating the purpose of hashing.
+ - If you care about performance, choose a hash function with as
+ good "spreading" as possible. A good hash function will react to
+ even a small change in its input with a completely different
+ resulting hash. Finally, don't make the hash function itself
+ overly slow, because you'll be incurring a non-negligible
+ overhead to reads and writes to the hash table.
Note that neither keys nor values are copied when inserted into the
hash table, so they must exist for the lifetime of the table. This
All the hash mappings (key-value pairs of pointers) are stored in a
contiguous array. The position of each mapping is determined by
- applying the hash function to the key: location = hash(key) % size.
- If two different keys end up on the same position, the collision is
- resolved by placing the second mapping at the next empty place in
- the array following the occupied place. This method of collision
- resolution is called "linear probing".
+ the hash value of its key and the size of the table: location :=
+ hash(key) % size. If two different keys end up on the same
+ position (hash collision), the one that came second is placed at
+ the next empty position following the occupied place. This
+ collision resolution technique is called "linear probing".
There are more advanced collision resolution mechanisms (quadratic
- probing, double hashing), but we don't use them because they
- involve more non-sequential access to the array, and therefore
- worse cache behavior. Linear probing works well as long as the
+ probing, double hashing), but we don't use them because they incur
+ more non-sequential access to the array, which results in worse
+ cache behavior. Linear probing works well as long as the
fullness/size ratio is kept below 75%. We make sure to regrow or
rehash the hash table whenever this threshold is exceeded.
};
struct hash_table {
- unsigned long (*hash_function) (const void *);
- int (*test_function) (const void *, const void *);
+ unsigned long (*hash_function) PARAMS ((const void *));
+ int (*test_function) PARAMS ((const void *, const void *));
int size; /* size of the array */
int count; /* number of non-empty, non-deleted
int resize_threshold; /* after size exceeds this number of
entries, resize the table. */
+ int prime_offset; /* the offset of the current prime in
+ the prime table. */
- struct mapping *mappings;
+ struct mapping *mappings; /* the array of mapping pairs. */
};
#define EMPTY_MAPPING_P(mp) ((mp)->key == NULL)
#define LOOP_NON_EMPTY(mp, mappings, size) \
for (; !EMPTY_MAPPING_P (mp); mp = NEXT_MAPPING (mp, mappings, size))
+/* #### We might want to multiply with the "golden ratio" here to get
+ better randomness for keys that do not result from a good hash
+ function. This is currently not a problem in Wget because we only
+ use the string hash tables. */
+
#define HASH_POSITION(ht, key) (ht->hash_function (key) % ht->size)
-/* Find a prime near, but greather than or equal to SIZE. */
+/* Find a prime near, but greather than or equal to SIZE. Of course,
+ the primes are not calculated, but looked up from a table. The
+ table does not contain all primes in range, just a selection useful
+ for this purpose.
+
+ PRIME_OFFSET is a micro-optimization: if specified, it starts the
+ search for the prime number beginning with the specific offset in
+ the prime number table. The final offset is stored in the same
+ variable. */
static int
-prime_size (int size)
+prime_size (int size, int *prime_offset)
{
static const unsigned long primes [] = {
- 19, 29, 41, 59, 79, 107, 149, 197, 263, 347, 457, 599, 787, 1031,
+ 13, 19, 29, 41, 59, 79, 107, 149, 197, 263, 347, 457, 599, 787, 1031,
1361, 1777, 2333, 3037, 3967, 5167, 6719, 8737, 11369, 14783,
19219, 24989, 32491, 42257, 54941, 71429, 92861, 120721, 156941,
204047, 265271, 344857, 448321, 582821, 757693, 985003, 1280519,
10445899, 13579681, 17653589, 22949669, 29834603, 38784989,
50420551, 65546729, 85210757, 110774011, 144006217, 187208107,
243370577, 316381771, 411296309, 534685237, 695090819, 903618083,
- 1174703521, 1527114613, 1985248999, 2580823717UL, 3355070839UL
+ 1174703521, 1527114613, 1985248999,
+ (unsigned long)0x99d43ea5, (unsigned long)0xc7fa5177
};
- int i;
- for (i = 0; i < ARRAY_SIZE (primes); i++)
+ int i = *prime_offset;
+
+ for (; i < countof (primes); i++)
if (primes[i] >= size)
- return primes[i];
- /* huh? */
- return size;
+ {
+ /* Set the offset to the next prime. That is safe because,
+ next time we are called, it will be with a larger SIZE,
+ which means we could never return the same prime anyway.
+ (If that is not the case, the caller can simply reset
+ *prime_offset.) */
+ *prime_offset = i + 1;
+ return primes[i];
+ }
+
+ abort ();
+ return 0;
}
/* Create a hash table of INITIAL_SIZE with hash function
being a prime number.
Consequently, if you wish to start out with a "small" table which
- will be regrown as needed, specify INITIAL_SIZE 0. */
+ will be regrown as needed, specify INITIAL_SIZE 0.
+
+ If HASH_FUNCTION is not provided, identity table is assumed,
+ i.e. key pointers are compared as keys. If you want strings with
+ equal contents to hash the same, use make_string_hash_table. */
struct hash_table *
hash_table_new (int initial_size,
struct hash_table *ht
= (struct hash_table *)xmalloc (sizeof (struct hash_table));
- ht->hash_function = hash_function;
- ht->test_function = test_function;
+ ht->hash_function = hash_function ? hash_function : ptrhash;
+ ht->test_function = test_function ? test_function : ptrcmp;
- ht->size = prime_size (initial_size);
+ ht->prime_offset = 0;
+ ht->size = prime_size (initial_size, &ht->prime_offset);
ht->resize_threshold = ht->size * 3 / 4;
- ht->count = 0;
+ ht->count = 0;
ht->mappings = xmalloc (ht->size * sizeof (struct mapping));
memset (ht->mappings, '\0', ht->size * sizeof (struct mapping));
mapping that matches KEY, or NULL if none matches. */
static inline struct mapping *
-find_mapping (struct hash_table *ht, const void *key)
+find_mapping (const struct hash_table *ht, const void *key)
{
struct mapping *mappings = ht->mappings;
int size = ht->size;
struct mapping *mp = mappings + HASH_POSITION (ht, key);
- int (*equals) (const void *, const void *) = ht->test_function;
+ int (*equals) PARAMS ((const void *, const void *)) = ht->test_function;
LOOP_NON_EMPTY (mp, mappings, size)
if (equals (key, mp->key))
/* Get the value that corresponds to the key KEY in the hash table HT.
If no value is found, return NULL. Note that NULL is a legal value
for value; if you are storing NULLs in your hash table, you can use
- hash_table_exists to be sure that a (possibly NULL) value exists in
- the table. Or, you can use hash_table_get_pair instead of this
+ hash_table_contains to be sure that a (possibly NULL) value exists
+ in the table. Or, you can use hash_table_get_pair instead of this
function. */
void *
-hash_table_get (struct hash_table *ht, const void *key)
+hash_table_get (const struct hash_table *ht, const void *key)
{
struct mapping *mp = find_mapping (ht, key);
if (mp)
value. Returns non-zero on success. */
int
-hash_table_get_pair (struct hash_table *ht, const void *lookup_key,
+hash_table_get_pair (const struct hash_table *ht, const void *lookup_key,
void *orig_key, void *value)
{
struct mapping *mp = find_mapping (ht, lookup_key);
return 0;
}
-/* Return 1 if KEY exists in HT, 0 otherwise. */
+/* Return 1 if HT contains KEY, 0 otherwise. */
int
-hash_table_exists (struct hash_table *ht, const void *key)
+hash_table_contains (const struct hash_table *ht, const void *key)
{
return find_mapping (ht, key) != NULL;
}
struct mapping *mp, *mappings;
int newsize;
- newsize = prime_size (ht->size * 2);
+ newsize = prime_size (ht->size * 2, &ht->prime_offset);
#if 0
- printf ("growing from %d to %d\n", ht->size, newsize);
+ printf ("growing from %d to %d; fullness %.2f%% to %.2f%%\n",
+ ht->size, newsize,
+ (double)100 * ht->count / ht->size,
+ (double)100 * ht->count / newsize);
#endif
ht->size = newsize;
{
struct mapping *mappings = ht->mappings;
int size = ht->size;
- int (*equals) (const void *, const void *) = ht->test_function;
+ int (*equals) PARAMS ((const void *, const void *)) = ht->test_function;
struct mapping *mp = mappings + HASH_POSITION (ht, key);
greater than the number of elements. */
int
-hash_table_count (struct hash_table *ht)
+hash_table_count (const struct hash_table *ht)
{
return ht->count;
}
\f
-/* Support for hash tables whose keys are strings. */
-
+/* Functions from this point onward are meant for convenience and
+ don't strictly belong to this file. However, this is as good a
+ place for them as any. */
+
+/*
+ * Support for hash tables whose keys are strings.
+ *
+ */
+
/* 31 bit hash function. Taken from Gnome's glib, modified to use
standard C types.
return h;
}
-#if 0
-/* If I ever need it: hashing of integers. */
+/* Frontend for strcmp usable for hash tables. */
-unsigned int
-inthash (unsigned int key)
+int
+string_cmp (const void *s1, const void *s2)
+{
+ return !strcmp ((const char *)s1, (const char *)s2);
+}
+
+/* Return a hash table of initial size INITIAL_SIZE suitable to use
+ strings as keys. */
+
+struct hash_table *
+make_string_hash_table (int initial_size)
+{
+ return hash_table_new (initial_size, string_hash, string_cmp);
+}
+
+/*
+ * Support for hash tables whose keys are strings, but which are
+ * compared case-insensitively.
+ *
+ */
+
+/* Like string_hash, but produce the same hash regardless of the case. */
+
+static unsigned long
+string_hash_nocase (const void *key)
+{
+ const char *p = key;
+ unsigned int h = TOLOWER (*p);
+
+ if (h)
+ for (p += 1; *p != '\0'; p++)
+ h = (h << 5) - h + TOLOWER (*p);
+
+ return h;
+}
+
+/* Like string_cmp, but doing case-insensitive compareison. */
+
+static int
+string_cmp_nocase (const void *s1, const void *s2)
+{
+ return !strcasecmp ((const char *)s1, (const char *)s2);
+}
+
+/* Like make_string_hash_table, but uses string_hash_nocase and
+ string_cmp_nocase. */
+
+struct hash_table *
+make_nocase_string_hash_table (int initial_size)
{
+ return hash_table_new (initial_size, string_hash_nocase, string_cmp_nocase);
+}
+
+/* Hashing of pointers. Used for hash tables that are keyed by
+ pointer identity. (Common Lisp calls them EQ hash tables, and Java
+ calls them IdentityHashMaps.) */
+
+unsigned long
+ptrhash (const void *ptr)
+{
+ unsigned long key = (unsigned long)ptr;
key += (key << 12);
key ^= (key >> 22);
key += (key << 4);
key ^= (key >> 2);
key += (key << 7);
key ^= (key >> 12);
+#if SIZEOF_LONG > 4
+ key += (key << 44);
+ key ^= (key >> 54);
+ key += (key << 36);
+ key ^= (key >> 41);
+ key += (key << 42);
+ key ^= (key >> 34);
+ key += (key << 39);
+ key ^= (key >> 44);
+#endif
return key;
}
-#endif
int
-string_cmp (const void *s1, const void *s2)
+ptrcmp (const void *ptr1, const void *ptr2)
{
- return !strcmp ((const char *)s1, (const char *)s2);
+ return ptr1 == ptr2;
}
-/* Return a hash table of initial size INITIAL_SIZE suitable to use
- strings as keys. */
+#if 0
+/* Currently unused: hashing of integers. */
-struct hash_table *
-make_string_hash_table (int initial_size)
+unsigned long
+inthash (unsigned int key)
{
- return hash_table_new (initial_size, string_hash, string_cmp);
+ key += (key << 12);
+ key ^= (key >> 22);
+ key += (key << 4);
+ key ^= (key >> 9);
+ key += (key << 10);
+ key ^= (key >> 2);
+ key += (key << 7);
+ key ^= (key >> 12);
+ return key;
}
-
+#endif
\f
#ifdef STANDALONE
if (len <= 1)
continue;
line[--len] = '\0';
- if (!hash_table_exists (ht, line))
+ if (!hash_table_contains (ht, line))
hash_table_put (ht, strdup (line), "here I am!");
#if 1
if (len % 5 == 0)
projects / wget / blobdiff