Simpler hash table iteration
Hrvoje Niksic <[email protected]> Sat, 27 Aug 2005 15:04:32 +0200
| Newsgroups | gmane.comp.web.wget.patches |
|---|---|
| Message-ID | <[email protected]> |
I noticed that the current interface for iterating over hash tables,
hash_table_map, is a bit tedious to use. Since C doesn't support
anonymous functions with lexical scope, each hash table iteration must
both implement a helper function *and* explicutly maintain the state
between invocations in a "context" structure.
This patch adds a simpler (and no slower) iteration API, used like
this:
hash_table_iterator iter;
for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
... do something with iter.key and iter.value ...
2005-08-27 Hrvoje Niksic <[email protected]>
* hash.c (hash_table_map): Rename to hash_table_for_each and
update callers.
Document the meaning of the callback's return value.
(hash_table_iterate): New function.
(hash_table_iter_next): Likewise.
Update most places that used hash_table_for_each to use the
iteration, which doesn't require a temporary function with
explicit state management.
Index: src/utils.c
===================================================================
--- src/utils.c (revision 2065)
+++ src/utils.c (working copy)
@@ -1165,50 +1165,36 @@
return hash_table_contains (ht, s);
}
-static int
-string_set_to_array_mapper (void *key, void *value_ignored, void *arg)
-{
- char ***arrayptr = (char ***) arg;
- *(*arrayptr)++ = (char *) key;
- return 0;
-}
-
/* Convert the specified string set to array. ARRAY should be large
enough to hold hash_table_count(ht) char pointers. */
void string_set_to_array (struct hash_table *ht, char **array)
{
- hash_table_map (ht, string_set_to_array_mapper, &array);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ *array++ = iter.key;
}
-static int
-string_set_free_mapper (void *key, void *value_ignored, void *arg_ignored)
-{
- xfree (key);
- return 0;
-}
-
void
string_set_free (struct hash_table *ht)
{
- hash_table_map (ht, string_set_free_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ xfree (iter.key);
hash_table_destroy (ht);
}
-static int
-free_keys_and_values_mapper (void *key, void *value, void *arg_ignored)
-{
- xfree (key);
- xfree (value);
- return 0;
-}
+/* Utility function: simply call free() on all keys and values of HT. */
-/* Another utility function: call free() on all keys and values of HT. */
-
void
free_keys_and_values (struct hash_table *ht)
{
- hash_table_map (ht, free_keys_and_values_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ {
+ xfree (iter.key);
+ xfree (iter.value);
+ }
}
/* Get digit grouping data for thousand separors by calling
Index: src/hash.c
===================================================================
--- src/hash.c (revision 2065)
+++ src/hash.c (working copy)
@@ -72,7 +72,9 @@
hash_table_get_pair -- get key/value pair for key.
hash_table_contains -- test whether the table contains key.
hash_table_remove -- remove key->value mapping for given key.
- hash_table_map -- iterate through table entries.
+ hash_table_for_each -- call function for each table entry.
+ hash_table_iterate -- iterate over entries in hash table.
+ hash_table_iter_next -- return next element during iteration.
hash_table_clear -- clear hash table contents.
hash_table_count -- return the number of entries in the table.
@@ -81,21 +83,22 @@
with each resize, which ensures that the amortized time per
operation remains constant.
- By default, tables created by hash_table_new consider the keys to
- be equal if their pointer values are the same. You can use
- make_string_hash_table to create tables whose keys are considered
- equal if their string contents are the same. In the general case,
- the criterion of equality used to compare keys is specified at
- table creation time with two callback functions, "hash" and "test".
- The hash function transforms the key into an arbitrary number that
- must be the same for two equal keys. The test function accepts two
- keys and returns non-zero if they are to be considered equal.
+ If not instructed otherwise, tables created by hash_table_new
+ consider the keys to be equal if their pointer values are the same.
+ You can use make_string_hash_table to create tables whose keys are
+ considered equal if their string contents are the same. In the
+ general case, the criterion of equality used to compare keys is
+ specified at table creation time with two callback functions,
+ "hash" and "test". The hash function transforms the key into an
+ arbitrary number that must be the same for two equal keys. The
+ test function accepts two keys and returns non-zero if they are to
+ be considered equal.
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
means that e.g. the use of static strings is OK, but objects with a
- shorter life-time need to be copied (with strdup() or the like in
- the case of strings) before being inserted. */
+ shorter life-time probably need to be copied (with strdup() or the
+ like in the case of strings) before being inserted. */
/* IMPLEMENTATION:
@@ -495,20 +498,22 @@
ht->count = 0;
}
-/* Map MAPFUN over all entries in HT. MAPFUN is called with three
- arguments: the key, the value, and MAPARG.
+/* Call FN for each entry in HT. FN is called with three arguments:
+ the key, the value, and ARG. When FN returns a non-zero value, the
+ mapping stops.
It is undefined what happens if you add or remove entries in the
- hash table while hash_table_map is running. The exception is the
- entry you're currently mapping over; you may remove or change that
- entry. */
+ hash table while hash_table_for_each is running. The exception is
+ the entry you're currently mapping over; you may call
+ hash_table_put or hash_table_remove on that entry's key. That is
+ also the reason why this function cannot be implemented in terms of
+ hash_table_iterate. */
void
-hash_table_map (struct hash_table *ht,
- int (*mapfun) (void *, void *, void *),
- void *maparg)
+hash_table_for_each (struct hash_table *ht,
+ int (*fn) (void *, void *, void *), void *arg)
{
- struct cell *c = ht->cells;
+ struct cell *c = ht->cells;
struct cell *end = ht->cells + ht->size;
for (; c < end; c++)
@@ -517,7 +522,7 @@
void *key;
repeat:
key = c->key;
- if (mapfun (key, c->value, maparg))
+ if (fn (key, c->value, arg))
return;
/* hash_table_remove might have moved the adjacent cells. */
if (c->key != key && CELL_OCCUPIED (c))
@@ -525,6 +530,48 @@
}
}
+/* Initiate iteration over HT. Get the next entry using
+ hash_table_iter_next. The typical loop looks like this:
+
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ ... do something with iter.key and iter.value ...
+
+ The iterator does not need to be deallocated after use. The hash
+ table must not be modified while being iterated over. */
+
+void
+hash_table_iterate (struct hash_table *ht, hash_table_iterator *iter)
+{
+ iter->pos = ht->cells;
+ iter->end = ht->cells + ht->size;
+}
+
+/* Get the next hash table entry. ITER is an iterator object
+ initialized using hash_table_iterate. While there are more
+ entries, the key and value pointers are stored to ITER->key and
+ ITER->value respectively and 1 is returned. When there are no more
+ entries, 0 is returned.
+
+ The hash table must not be modified between calls to this
+ function. */
+
+int
+hash_table_iter_next (hash_table_iterator *iter)
+{
+ struct cell *c = iter->pos;
+ struct cell *end = iter->end;
+ for (; c < end; c++)
+ if (CELL_OCCUPIED (c))
+ {
+ iter->key = c->key;
+ iter->value = c->value;
+ iter->pos = c + 1;
+ return 1;
+ }
+ return 0;
+}
+
/* Return the number of elements in the hash table. This is not the
same as the physical size of the hash table, which is always
greater than the number of elements. */
@@ -653,14 +700,15 @@
/* Hashing of numeric values, such as pointers and integers.
This implementation is the Robert Jenkins' 32 bit Mix Function,
- with a simple adaptation for 64-bit values. It offers excellent
- spreading of values and doesn't need to know the hash table size to
- work (unlike the very popular Knuth's multiplication hash). */
+ with a simple adaptation for 64-bit values. According to Jenkins
+ it should offer excellent spreading of values. Unlike the popular
+ Knuth's multiplication hash, this function doesn't need to know the
+ hash table size to work. */
unsigned long
hash_pointer (const void *ptr)
{
- unsigned long key = (unsigned long)ptr;
+ unsigned long key = (unsigned long) ptr;
key += (key << 12);
key ^= (key >> 22);
key += (key << 4);
@@ -693,20 +741,16 @@
#include <stdio.h>
#include <string.h>
-int
-print_hash_table_mapper (void *key, void *value, void *count)
-{
- ++*(int *)count;
- printf ("%s: %s\n", (const char *)key, (char *)value);
- return 0;
-}
-
void
print_hash (struct hash_table *sht)
{
- int debug_count = 0;
- hash_table_map (sht, print_hash_table_mapper, &debug_count);
- assert (debug_count == sht->count);
+ hash_table_iterator iter;
+ int count = 0;
+
+ for (hash_table_iterate (sht, &iter); hash_table_iter_next (&iter);
+ ++count)
+ printf ("%s: %s\n", iter.key, iter.value);
+ assert (count == sht->count);
}
int
Index: src/hash.h
===================================================================
--- src/hash.h (revision 2065)
+++ src/hash.h (working copy)
@@ -45,8 +45,16 @@
int hash_table_remove (struct hash_table *, const void *);
void hash_table_clear (struct hash_table *);
-void hash_table_map (struct hash_table *,
- int (*) (void *, void *, void *), void *);
+void hash_table_for_each (struct hash_table *,
+ int (*) (void *, void *, void *), void *);
+
+typedef struct {
+ void *key, *value; /* public members */
+ void *pos, *end; /* private members */
+} hash_table_iterator;
+void hash_table_iterate (struct hash_table *, hash_table_iterator *);
+int hash_table_iter_next (hash_table_iterator *);
+
int hash_table_count (const struct hash_table *);
struct hash_table *make_string_hash_table (int);
Index: src/cookies.c
===================================================================
--- src/cookies.c (revision 2065)
+++ src/cookies.c (working copy)
@@ -1421,44 +1421,13 @@
fclose (fp);
}
-/* Mapper for save_cookies callable by hash_table_map. VALUE points
- to the head in a chain of cookies. The function prints the entire
- chain. */
-
-static int
-save_cookies_mapper (void *key, void *value, void *arg)
-{
- FILE *fp = (FILE *)arg;
- char *domain = (char *)key;
- struct cookie *cookie = (struct cookie *)value;
- for (; cookie; cookie = cookie->next)
- {
- if (!cookie->permanent && !opt.keep_session_cookies)
- continue;
- if (cookie_expired_p (cookie))
- continue;
- if (!cookie->domain_exact)
- fputc ('.', fp);
- fputs (domain, fp);
- if (cookie->port != PORT_ANY)
- fprintf (fp, ":%d", cookie->port);
- fprintf (fp, "\t%s\t%s\t%s\t%.0f\t%s\t%s\n",
- cookie->domain_exact ? "FALSE" : "TRUE",
- cookie->path, cookie->secure ? "TRUE" : "FALSE",
- (double)cookie->expiry_time,
- cookie->attr, cookie->value);
- if (ferror (fp))
- return 1; /* stop mapping */
- }
- return 0;
-}
-
/* Save cookies, in format described above, to FILE. */
void
cookie_jar_save (struct cookie_jar *jar, const char *file)
{
FILE *fp;
+ hash_table_iterator iter;
DEBUGP (("Saving cookies to %s.\n", file));
@@ -1476,8 +1445,33 @@
fprintf (fp, "# Generated by Wget on %s.\n", datetime_str (&cookies_now));
fputs ("# Edit at your own risk.\n\n", fp);
- hash_table_map (jar->chains, save_cookies_mapper, fp);
-
+ for (hash_table_iterate (jar->chains, &iter);
+ hash_table_iter_next (&iter);
+ )
+ {
+ char *domain = iter.key;
+ struct cookie *cookie = iter.value;
+ for (; cookie; cookie = cookie->next)
+ {
+ if (!cookie->permanent && !opt.keep_session_cookies)
+ continue;
+ if (cookie_expired_p (cookie))
+ continue;
+ if (!cookie->domain_exact)
+ fputc ('.', fp);
+ fputs (domain, fp);
+ if (cookie->port != PORT_ANY)
+ fprintf (fp, ":%d", cookie->port);
+ fprintf (fp, "\t%s\t%s\t%s\t%.0f\t%s\t%s\n",
+ cookie->domain_exact ? "FALSE" : "TRUE",
+ cookie->path, cookie->secure ? "TRUE" : "FALSE",
+ (double)cookie->expiry_time,
+ cookie->attr, cookie->value);
+ if (ferror (fp))
+ goto out;
+ }
+ }
+ out:
if (ferror (fp))
logprintf (LOG_NOTQUIET, _("Error writing to `%s': %s\n"),
file, strerror (errno));
@@ -1489,9 +1483,9 @@
}
/* Destroy all the elements in the chain and unhook it from the cookie
- jar. This is written in the form of a callback to hash_table_map
- and used by cookie_jar_delete to delete all the cookies in a
- jar. */
+ jar. This is written in the form of a callback to
+ hash_table_for_each and used by cookie_jar_delete to delete all the
+ cookies in a jar. */
static int
nuke_cookie_chain (void *value, void *key, void *arg)
@@ -1521,7 +1515,7 @@
void
cookie_jar_delete (struct cookie_jar *jar)
{
- hash_table_map (jar->chains, nuke_cookie_chain, jar);
+ hash_table_for_each (jar->chains, nuke_cookie_chain, jar);
hash_table_destroy (jar->chains);
xfree (jar);
}
Index: src/host.c
===================================================================
--- src/host.c (revision 2065)
+++ src/host.c (working copy)
@@ -857,26 +857,22 @@
return false;
}
-static int
-host_cleanup_mapper (void *key, void *value, void *arg_ignored)
-{
- struct address_list *al;
-
- xfree (key); /* host */
-
- al = (struct address_list *)value;
- assert (al->refcount == 1);
- address_list_delete (al);
-
- return 0;
-}
-
void
host_cleanup (void)
{
if (host_name_addresses_map)
{
- hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (host_name_addresses_map, &iter);
+ hash_table_iter_next (&iter);
+ )
+ {
+ char *host = iter.key;
+ struct address_list *al = iter.value;
+ xfree (host);
+ assert (al->refcount == 1);
+ address_list_delete (al);
+ }
hash_table_destroy (host_name_addresses_map);
host_name_addresses_map = NULL;
}
Index: src/res.c
===================================================================
--- src/res.c (revision 2065)
+++ src/res.c (working copy)
@@ -552,20 +552,19 @@
return err == RETROK;
}
-static int
-cleanup_hash_table_mapper (void *key, void *value, void *arg_ignored)
-{
- xfree (key);
- free_specs (value);
- return 0;
-}
-
void
res_cleanup (void)
{
if (registered_specs)
{
- hash_table_map (registered_specs, cleanup_hash_table_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (registered_specs, &iter);
+ hash_table_iter_next (&iter);
+ )
+ {
+ xfree (iter.key);
+ free_specs (iter.value);
+ }
hash_table_destroy (registered_specs);
registered_specs = NULL;
}
Index: src/convert.c
===================================================================
--- src/convert.c (revision 2065)
+++ src/convert.c (working copy)
@@ -700,8 +700,9 @@
static void
dissociate_urls_from_file (const char *file)
{
- hash_table_map (dl_url_file_map, dissociate_urls_from_file_mapper,
- (char *)file);
+ /* Can't use hash_table_iter_* because the table mutates while mapping. */
+ hash_table_for_each (dl_url_file_map, dissociate_urls_from_file_mapper,
+ (char *) file);
}
/* Register that URL has been successfully downloaded to FILE. This
@@ -937,19 +938,16 @@
return FILE_NOT_ALREADY_DOWNLOADED;
}
-static int
-df_free_mapper (void *key, void *value, void *ignored)
-{
- xfree (key);
- return 0;
-}
-
static void
downloaded_files_free (void)
{
if (downloaded_files_hash)
{
- hash_table_map (downloaded_files_hash, df_free_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (downloaded_files_hash, &iter);
+ hash_table_iter_next (&iter);
+ )
+ xfree (iter.key);
hash_table_destroy (downloaded_files_hash);
downloaded_files_hash = NULL;
}