glib/glib/ghash.c

1462 lines
40 KiB
C
Raw Normal View History

1998-06-10 23:21:14 +00:00
/* GLIB - Library of useful routines for C programming
* Copyright (C) 1995-1997 Peter Mattis, Spencer Kimball and Josh MacDonald
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
1998-06-10 23:21:14 +00:00
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
1998-06-10 23:21:14 +00:00
*
* You should have received a copy of the GNU Lesser General Public
1998-06-10 23:21:14 +00:00
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
/*
* Modified by the GLib Team and others 1997-2000. See the AUTHORS
* file for a list of people on the GLib Team. See the ChangeLog
* files for a list of changes. These files are distributed with
* GLib at ftp://ftp.gtk.org/pub/gtk/.
*/
/*
* MT safe
*/
#include "config.h"
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
#include <string.h> /* memset */
2010-09-03 23:03:14 -04:00
#include "ghash.h"
#include "gatomic.h"
2010-09-03 23:03:14 -04:00
#include "gtestutils.h"
1998-06-10 23:21:14 +00:00
/**
* SECTION: hash_tables
* @title: Hash Tables
* @short_description: associations between keys and values so that
* given a key the value can be found quickly
*
* A #GHashTable provides associations between keys and values which is
* optimized so that given a key, the associated value can be found
* very quickly.
*
* Note that neither keys nor values are copied when inserted into the
* #GHashTable, so they must exist for the lifetime of the #GHashTable.
* This means that the use of static strings is OK, but temporary
* strings (i.e. those created in buffers and those returned by GTK+
* widgets) should be copied with g_strdup() before being inserted.
*
* If keys or values are dynamically allocated, you must be careful to
* ensure that they are freed when they are removed from the
* #GHashTable, and also when they are overwritten by new insertions
* into the #GHashTable. It is also not advisable to mix static strings
* and dynamically-allocated strings in a #GHashTable, because it then
* becomes difficult to determine whether the string should be freed.
*
* To create a #GHashTable, use g_hash_table_new().
*
* To insert a key and value into a #GHashTable, use
* g_hash_table_insert().
*
* To lookup a value corresponding to a given key, use
* g_hash_table_lookup() and g_hash_table_lookup_extended().
*
* To remove a key and value, use g_hash_table_remove().
*
* To call a function for each key and value pair use
* g_hash_table_foreach() or use a iterator to iterate over the
* key/value pairs in the hash table, see #GHashTableIter.
*
* To destroy a #GHashTable use g_hash_table_destroy().
**/
/**
* GHashTable:
*
* The #GHashTable struct is an opaque data structure to represent a
* <link linkend="glib-Hash-Tables">Hash Table</link>. It should only be
* accessed via the following functions.
**/
/**
* GHashFunc:
* @key: a key.
* @Returns: the hash value corresponding to the key.
*
* Specifies the type of the hash function which is passed to
* g_hash_table_new() when a #GHashTable is created.
*
* The function is passed a key and should return a #guint hash value.
* The functions g_direct_hash(), g_int_hash() and g_str_hash() provide
* hash functions which can be used when the key is a #gpointer, #gint,
* and #gchar* respectively.
*
* <!-- FIXME: Need more here. --> The hash values should be evenly
* distributed over a fairly large range? The modulus is taken with the
* hash table size (a prime number) to find the 'bucket' to place each
* key into. The function should also be very fast, since it is called
* for each key lookup.
**/
/**
* GHFunc:
* @key: a key.
* @value: the value corresponding to the key.
* @user_data: user data passed to g_hash_table_foreach().
*
* Specifies the type of the function passed to g_hash_table_foreach().
* It is called with each key/value pair, together with the @user_data
* parameter which is passed to g_hash_table_foreach().
**/
/**
* GHRFunc:
* @key: a key.
* @value: the value associated with the key.
* @user_data: user data passed to g_hash_table_remove().
* @Returns: %TRUE if the key/value pair should be removed from the
* #GHashTable.
*
* Specifies the type of the function passed to
* g_hash_table_foreach_remove(). It is called with each key/value
* pair, together with the @user_data parameter passed to
* g_hash_table_foreach_remove(). It should return %TRUE if the
* key/value pair should be removed from the #GHashTable.
**/
/**
* GEqualFunc:
* @a: a value.
* @b: a value to compare with.
* @Returns: %TRUE if @a = @b; %FALSE otherwise.
*
* Specifies the type of a function used to test two values for
* equality. The function should return %TRUE if both values are equal
* and %FALSE otherwise.
**/
/**
* GHashTableIter:
*
* A GHashTableIter structure represents an iterator that can be used
* to iterate over the elements of a #GHashTable. GHashTableIter
* structures are typically allocated on the stack and then initialized
* with g_hash_table_iter_init().
**/
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
#define HASH_TABLE_MIN_SHIFT 3 /* 1 << 3 == 8 buckets */
1998-06-10 23:21:14 +00:00
typedef struct _GHashNode GHashNode;
struct _GHashNode
{
gpointer key;
gpointer value;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
/* If key_hash == 0, node is not in use
* If key_hash == 1, node is a tombstone
* If key_hash >= 2, node contains data */
guint key_hash;
1998-06-10 23:21:14 +00:00
};
struct _GHashTable
1998-06-10 23:21:14 +00:00
{
gint size;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
gint mod;
guint mask;
gint nnodes;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
gint noccupied; /* nnodes + tombstones */
GHashNode *nodes;
GHashFunc hash_func;
GEqualFunc key_equal_func;
2006-12-18 20:17:35 +00:00
volatile gint ref_count;
#ifndef G_DISABLE_ASSERT
/*
* Tracks the structure of the hash table, not its contents: is only
* incremented when a node is added or removed (is not incremented
* when the key or data of a node is modified).
*/
int version;
#endif
GDestroyNotify key_destroy_func;
GDestroyNotify value_destroy_func;
1998-06-10 23:21:14 +00:00
};
typedef struct
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashTable *hash_table;
gpointer dummy1;
gpointer dummy2;
int position;
gboolean dummy3;
int version;
} RealIter;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
/* Each table size has an associated prime modulo (the first prime
* lower than the table size) used to find the initial bucket. Probing
* then works modulo 2^n. The prime modulo is necessary to get a
* good distribution with poor hash functions. */
static const gint prime_mod [] =
{
1, /* For 1 << 0 */
2,
3,
7,
13,
31,
61,
127,
251,
509,
1021,
2039,
4093,
8191,
16381,
32749,
65521, /* For 1 << 16 */
131071,
262139,
524287,
1048573,
2097143,
4194301,
8388593,
16777213,
33554393,
67108859,
134217689,
268435399,
536870909,
1073741789,
2147483647 /* For 1 << 31 */
};
static void
g_hash_table_set_shift (GHashTable *hash_table, gint shift)
{
gint i;
guint mask = 0;
hash_table->size = 1 << shift;
hash_table->mod = prime_mod [shift];
for (i = 0; i < shift; i++)
{
mask <<= 1;
mask |= 1;
}
hash_table->mask = mask;
}
static gint
g_hash_table_find_closest_shift (gint n)
{
gint i;
for (i = 0; n; i++)
n >>= 1;
return i;
}
static void
g_hash_table_set_shift_from_size (GHashTable *hash_table, gint size)
{
gint shift;
shift = g_hash_table_find_closest_shift (size);
shift = MAX (shift, HASH_TABLE_MIN_SHIFT);
g_hash_table_set_shift (hash_table, shift);
}
/*
* g_hash_table_lookup_node:
* @hash_table: our #GHashTable
* @key: the key to lookup against
* @hash_return: optional key hash return location
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
* Return value: index of the described #GHashNode
*
* Performs a lookup in the hash table. Virtually all hash operations
* will use this function internally.
*
* This function first computes the hash value of the key using the
* user's hash function.
*
* If an entry in the table matching @key is found then this function
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
* returns the index of that entry in the table, and if not, the
* index of an empty node (never a tombstone).
*/
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
static inline guint
g_hash_table_lookup_node (GHashTable *hash_table,
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
gconstpointer key)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *node;
guint node_index;
guint hash_value;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
guint step = 0;
/* Empty buckets have hash_value set to 0, and for tombstones, it's 1.
* We need to make sure our hash value is not one of these. */
hash_value = (* hash_table->hash_func) (key);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (G_UNLIKELY (hash_value <= 1))
hash_value = 2;
node_index = hash_value % hash_table->mod;
node = &hash_table->nodes [node_index];
while (node->key_hash)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
/* We first check if our full hash values
* are equal so we can avoid calling the full-blown
* key equality function in most cases.
*/
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (node->key_hash == hash_value)
{
if (hash_table->key_equal_func)
{
if (hash_table->key_equal_func (node->key, key))
break;
}
else if (node->key == key)
{
break;
}
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
step++;
node_index += step;
node_index &= hash_table->mask;
node = &hash_table->nodes [node_index];
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
return node_index;
}
/*
* g_hash_table_lookup_node_for_insertion:
* @hash_table: our #GHashTable
* @key: the key to lookup against
* @hash_return: key hash return location
* Return value: index of the described #GHashNode
*
* Performs a lookup in the hash table, preserving extra information
* usually needed for insertion.
*
* This function first computes the hash value of the key using the
* user's hash function.
*
* If an entry in the table matching @key is found then this function
* returns the index of that entry in the table, and if not, the
* index of an unused node (empty or tombstone) where the key can be
* inserted.
*
* The computed hash value is returned in the variable pointed to
* by @hash_return. This is to save insertions from having to compute
* the hash record again for the new record.
*/
static inline guint
g_hash_table_lookup_node_for_insertion (GHashTable *hash_table,
gconstpointer key,
guint *hash_return)
{
GHashNode *node;
guint node_index;
guint hash_value;
guint first_tombstone;
gboolean have_tombstone = FALSE;
guint step = 0;
/* Empty buckets have hash_value set to 0, and for tombstones, it's 1.
* We need to make sure our hash value is not one of these. */
hash_value = (* hash_table->hash_func) (key);
if (G_UNLIKELY (hash_value <= 1))
hash_value = 2;
*hash_return = hash_value;
node_index = hash_value % hash_table->mod;
node = &hash_table->nodes [node_index];
while (node->key_hash)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
/* We first check if our full hash values
* are equal so we can avoid calling the full-blown
* key equality function in most cases.
*/
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (node->key_hash == hash_value)
{
if (hash_table->key_equal_func)
{
if (hash_table->key_equal_func (node->key, key))
return node_index;
}
else if (node->key == key)
{
return node_index;
}
}
else if (node->key_hash == 1 && !have_tombstone)
{
first_tombstone = node_index;
have_tombstone = TRUE;
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
step++;
node_index += step;
node_index &= hash_table->mask;
node = &hash_table->nodes [node_index];
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (have_tombstone)
return first_tombstone;
return node_index;
}
/*
* g_hash_table_remove_node:
* @hash_table: our #GHashTable
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
* @node: pointer to node to remove
* @notify: %TRUE if the destroy notify handlers are to be called
*
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
* Removes a node from the hash table and updates the node count.
* The node is replaced by a tombstone. No table resize is performed.
*
* If @notify is %TRUE then the destroy notify functions are called
* for the key and value of the hash node.
*/
static void
g_hash_table_remove_node (GHashTable *hash_table,
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *node,
gboolean notify)
{
if (notify && hash_table->key_destroy_func)
hash_table->key_destroy_func (node->key);
if (notify && hash_table->value_destroy_func)
hash_table->value_destroy_func (node->value);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
/* Erect tombstone */
node->key_hash = 1;
/* Be GC friendly */
node->key = NULL;
node->value = NULL;
hash_table->nnodes--;
}
/*
* g_hash_table_remove_all_nodes:
* @hash_table: our #GHashTable
* @notify: %TRUE if the destroy notify handlers are to be called
*
* Removes all nodes from the table. Since this may be a precursor to
* freeing the table entirely, no resize is performed.
*
* If @notify is %TRUE then the destroy notify functions are called
* for the key and value of the hash node.
*/
static void
g_hash_table_remove_all_nodes (GHashTable *hash_table,
gboolean notify)
{
int i;
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1)
{
if (notify && hash_table->key_destroy_func)
hash_table->key_destroy_func (node->key);
if (notify && hash_table->value_destroy_func)
hash_table->value_destroy_func (node->value);
}
}
/* We need to set node->key_hash = 0 for all nodes - might as well be GC
* friendly and clear everything */
memset (hash_table->nodes, 0, hash_table->size * sizeof (GHashNode));
hash_table->nnodes = 0;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
hash_table->noccupied = 0;
}
/*
* g_hash_table_resize:
* @hash_table: our #GHashTable
*
* Resizes the hash table to the optimal size based on the number of
* nodes currently held. If you call this function then a resize will
* occur, even if one does not need to occur. Use
* g_hash_table_maybe_resize() instead.
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
*
* This function may "resize" the hash table to its current size, with
* the side effect of cleaning up tombstones and otherwise optimizing
* the probe sequences.
*/
static void
g_hash_table_resize (GHashTable *hash_table)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *new_nodes;
gint old_size;
gint i;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
old_size = hash_table->size;
g_hash_table_set_shift_from_size (hash_table, hash_table->nnodes * 2);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
new_nodes = g_new0 (GHashNode, hash_table->size);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
for (i = 0; i < old_size; i++)
{
GHashNode *node = &hash_table->nodes [i];
GHashNode *new_node;
guint hash_val;
guint step = 0;
if (node->key_hash <= 1)
continue;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
hash_val = node->key_hash % hash_table->mod;
new_node = &new_nodes [hash_val];
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
while (new_node->key_hash)
{
step++;
hash_val += step;
hash_val &= hash_table->mask;
new_node = &new_nodes [hash_val];
}
*new_node = *node;
}
g_free (hash_table->nodes);
hash_table->nodes = new_nodes;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
hash_table->noccupied = hash_table->nnodes;
}
1998-06-10 23:21:14 +00:00
/*
* g_hash_table_maybe_resize:
* @hash_table: our #GHashTable
*
* Resizes the hash table, if needed.
*
* Essentially, calls g_hash_table_resize() if the table has strayed
* too far from its ideal size for its number of nodes.
*/
static inline void
g_hash_table_maybe_resize (GHashTable *hash_table)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
gint noccupied = hash_table->noccupied;
gint size = hash_table->size;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if ((size > hash_table->nnodes * 4 && size > 1 << HASH_TABLE_MIN_SHIFT) ||
(size <= noccupied + (noccupied / 16)))
g_hash_table_resize (hash_table);
}
1998-06-10 23:21:14 +00:00
/**
* g_hash_table_new:
* @hash_func: a function to create a hash value from a key.
* Hash values are used to determine where keys are stored within the
* #GHashTable data structure. The g_direct_hash(), g_int_hash(),
* g_int64_hash(), g_double_hash() and g_str_hash() functions are provided
* for some common types of keys.
* If hash_func is %NULL, g_direct_hash() is used.
* @key_equal_func: a function to check two keys for equality. This is
* used when looking up keys in the #GHashTable. The g_direct_equal(),
* g_int_equal(), g_int64_equal(), g_double_equal() and g_str_equal()
* functions are provided for the most common types of keys.
* If @key_equal_func is %NULL, keys are compared directly in a similar
* fashion to g_direct_equal(), but without the overhead of a function call.
*
* Creates a new #GHashTable with a reference count of 1.
*
* Return value: a new #GHashTable.
**/
1998-06-10 23:21:14 +00:00
GHashTable*
g_hash_table_new (GHashFunc hash_func,
GEqualFunc key_equal_func)
{
return g_hash_table_new_full (hash_func, key_equal_func, NULL, NULL);
}
/**
* g_hash_table_new_full:
* @hash_func: a function to create a hash value from a key.
* @key_equal_func: a function to check two keys for equality.
* @key_destroy_func: a function to free the memory allocated for the key
* used when removing the entry from the #GHashTable or %NULL if you
* don't want to supply such a function.
* @value_destroy_func: a function to free the memory allocated for the
* value used when removing the entry from the #GHashTable or %NULL if
* you don't want to supply such a function.
*
* Creates a new #GHashTable like g_hash_table_new() with a reference count
* of 1 and allows to specify functions to free the memory allocated for the
* key and value that get called when removing the entry from the #GHashTable.
*
* Return value: a new #GHashTable.
**/
GHashTable*
g_hash_table_new_full (GHashFunc hash_func,
GEqualFunc key_equal_func,
GDestroyNotify key_destroy_func,
GDestroyNotify value_destroy_func)
1998-06-10 23:21:14 +00:00
{
GHashTable *hash_table;
prepared deprecation of GMemChunk and GAllocator. added g_slice_*() API to Tue Nov 1 16:24:20 2005 Tim Janik <timj@imendio.com> * glib/gmem.[hc]: prepared deprecation of GMemChunk and GAllocator. added g_slice_*() API to allocate and cache small bits of memory. an actuall allocator implementation for g_slice_*() is still pending. * glib/gthread.[hc]: changes from a patch by Matthias Clasen. changed GRealThread list to use in-structure *next; fields instead of GSList, in order for thread iteration to not depenend on g_slice_*() indirectly. _g_thread_mem_private_get(): _g_thread_mem_private_set(): added accessors for private memory, needed because the ordinary GPrivate implementation relies on GArray and GSList and therefore indirectly on working g_slice_*() allocations. * glib/gthread.[hc]: g_thread_foreach(): new public API function to loop over all existing threads. * glib/gdataset.c: * glib/gstring.c: * glib/gcache.c: * glib/garray.c: * glib/gqueue.c: * glib/gslist.c: * glib/glist.c: * glib/ghash.c: * glib/gtree.c: * glib/ghook.c: * glib/gmain.c: * glib/gnode.c: removed GAllocator and free list usages and accompanying locks. use g_slice_*() API to allocate and cache small bits of memory. * glib/ghook.h: removed GMemChunk field from public API. * glib/gslist.h: * glib/glist.h: deprecate allocator API, provide _free1() for consistency. * glib/gnode.h: deprecate allocator API. * glib/gmain.c: reordered GPollRec fields so g_slice_free_chain() can be used for poll rec lists. * glib/grel.c: removed mem chunk usage, and allocated tuples via g_slice_*(). g_relation_destroy(): free all tuples from the all_tuples hash table, this effectively maintains the life time track keeping of tuples. g_relation_delete_tuple(): free tuples which are removed from the all_tuples hash table. this fixes a temporary leak that was present in the memchunk code until the destruction of the relation.
2005-11-01 18:10:31 +00:00
hash_table = g_slice_new (GHashTable);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_hash_table_set_shift (hash_table, HASH_TABLE_MIN_SHIFT);
hash_table->nnodes = 0;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
hash_table->noccupied = 0;
hash_table->hash_func = hash_func ? hash_func : g_direct_hash;
hash_table->key_equal_func = key_equal_func;
hash_table->ref_count = 1;
#ifndef G_DISABLE_ASSERT
hash_table->version = 0;
#endif
hash_table->key_destroy_func = key_destroy_func;
hash_table->value_destroy_func = value_destroy_func;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
hash_table->nodes = g_new0 (GHashNode, hash_table->size);
return hash_table;
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_iter_init:
* @iter: an uninitialized #GHashTableIter.
* @hash_table: a #GHashTable.
*
* Initializes a key/value pair iterator and associates it with
* @hash_table. Modifying the hash table after calling this function
* invalidates the returned iterator.
* |[
* GHashTableIter iter;
* gpointer key, value;
*
* g_hash_table_iter_init (&iter, hash_table);
* while (g_hash_table_iter_next (&iter, &key, &value))
* {
* /&ast; do something with key and value &ast;/
* }
* ]|
*
* Since: 2.16
**/
void
g_hash_table_iter_init (GHashTableIter *iter,
GHashTable *hash_table)
{
RealIter *ri = (RealIter *) iter;
g_return_if_fail (iter != NULL);
g_return_if_fail (hash_table != NULL);
ri->hash_table = hash_table;
ri->position = -1;
#ifndef G_DISABLE_ASSERT
ri->version = hash_table->version;
#endif
}
/**
* g_hash_table_iter_next:
* @iter: an initialized #GHashTableIter.
* @key: a location to store the key, or %NULL.
* @value: a location to store the value, or %NULL.
*
* Advances @iter and retrieves the key and/or value that are now
* pointed to as a result of this advancement. If %FALSE is returned,
* @key and @value are not set, and the iterator becomes invalid.
*
* Return value: %FALSE if the end of the #GHashTable has been reached.
*
* Since: 2.16
**/
gboolean
g_hash_table_iter_next (GHashTableIter *iter,
gpointer *key,
gpointer *value)
{
RealIter *ri = (RealIter *) iter;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *node;
gint position;
g_return_val_if_fail (iter != NULL, FALSE);
#ifndef G_DISABLE_ASSERT
g_return_val_if_fail (ri->version == ri->hash_table->version, FALSE);
#endif
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_return_val_if_fail (ri->position < ri->hash_table->size, FALSE);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
position = ri->position;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
do
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
position++;
if (position >= ri->hash_table->size)
{
ri->position = position;
return FALSE;
}
node = &ri->hash_table->nodes [position];
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
while (node->key_hash <= 1);
if (key != NULL)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
*key = node->key;
if (value != NULL)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
*value = node->value;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
ri->position = position;
return TRUE;
}
/**
* g_hash_table_iter_get_hash_table:
* @iter: an initialized #GHashTableIter.
*
* Returns the #GHashTable associated with @iter.
*
* Return value: the #GHashTable associated with @iter.
*
* Since: 2.16
**/
GHashTable *
g_hash_table_iter_get_hash_table (GHashTableIter *iter)
{
g_return_val_if_fail (iter != NULL, NULL);
return ((RealIter *) iter)->hash_table;
}
static void
iter_remove_or_steal (RealIter *ri, gboolean notify)
{
g_return_if_fail (ri != NULL);
#ifndef G_DISABLE_ASSERT
g_return_if_fail (ri->version == ri->hash_table->version);
#endif
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_return_if_fail (ri->position >= 0);
g_return_if_fail (ri->position < ri->hash_table->size);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_hash_table_remove_node (ri->hash_table, &ri->hash_table->nodes [ri->position], notify);
#ifndef G_DISABLE_ASSERT
ri->version++;
ri->hash_table->version++;
#endif
}
/**
2010-03-21 14:04:07 -05:00
* g_hash_table_iter_remove:
* @iter: an initialized #GHashTableIter.
*
* Removes the key/value pair currently pointed to by the iterator
* from its associated #GHashTable. Can only be called after
* g_hash_table_iter_next() returned %TRUE, and cannot be called more
* than once for the same key/value pair.
*
* If the #GHashTable was created using g_hash_table_new_full(), the
* key and value are freed using the supplied destroy functions, otherwise
* you have to make sure that any dynamically allocated values are freed
* yourself.
*
* Since: 2.16
**/
void
g_hash_table_iter_remove (GHashTableIter *iter)
{
iter_remove_or_steal ((RealIter *) iter, TRUE);
}
/**
2010-03-21 14:04:07 -05:00
* g_hash_table_iter_steal:
* @iter: an initialized #GHashTableIter.
*
* Removes the key/value pair currently pointed to by the iterator
* from its associated #GHashTable, without calling the key and value
* destroy functions. Can only be called after
* g_hash_table_iter_next() returned %TRUE, and cannot be called more
* than once for the same key/value pair.
*
* Since: 2.16
**/
void
g_hash_table_iter_steal (GHashTableIter *iter)
{
iter_remove_or_steal ((RealIter *) iter, FALSE);
}
/**
* g_hash_table_ref:
* @hash_table: a valid #GHashTable.
*
* Atomically increments the reference count of @hash_table by one.
* This function is MT-safe and may be called from any thread.
*
* Return value: the passed in #GHashTable.
*
2005-11-23 04:29:43 +00:00
* Since: 2.10
**/
GHashTable*
g_hash_table_ref (GHashTable *hash_table)
{
g_return_val_if_fail (hash_table != NULL, NULL);
g_return_val_if_fail (hash_table->ref_count > 0, hash_table);
g_atomic_int_add (&hash_table->ref_count, 1);
return hash_table;
}
/**
* g_hash_table_unref:
* @hash_table: a valid #GHashTable.
*
* Atomically decrements the reference count of @hash_table by one.
* If the reference count drops to 0, all keys and values will be
* destroyed, and all memory allocated by the hash table is released.
* This function is MT-safe and may be called from any thread.
*
2005-11-23 04:29:43 +00:00
* Since: 2.10
**/
void
g_hash_table_unref (GHashTable *hash_table)
{
g_return_if_fail (hash_table != NULL);
g_return_if_fail (hash_table->ref_count > 0);
if (g_atomic_int_exchange_and_add (&hash_table->ref_count, -1) - 1 == 0)
{
g_hash_table_remove_all_nodes (hash_table, TRUE);
g_free (hash_table->nodes);
g_slice_free (GHashTable, hash_table);
}
}
/**
* g_hash_table_destroy:
* @hash_table: a #GHashTable.
*
2005-11-25 23:20:22 +00:00
* Destroys all keys and values in the #GHashTable and decrements its
* reference count by 1. If keys and/or values are dynamically allocated,
* you should either free them first or create the #GHashTable with destroy
* notifiers using g_hash_table_new_full(). In the latter case the destroy
* functions you supplied will be called on all keys and values during the
* destruction phase.
**/
1998-06-10 23:21:14 +00:00
void
g_hash_table_destroy (GHashTable *hash_table)
{
g_return_if_fail (hash_table != NULL);
g_return_if_fail (hash_table->ref_count > 0);
g_hash_table_remove_all (hash_table);
g_hash_table_unref (hash_table);
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_lookup:
* @hash_table: a #GHashTable.
* @key: the key to look up.
*
* Looks up a key in a #GHashTable. Note that this function cannot
* distinguish between a key that is not present and one which is present
* and has the value %NULL. If you need this distinction, use
* g_hash_table_lookup_extended().
*
* Return value: the associated value, or %NULL if the key is not found.
**/
gpointer
g_hash_table_lookup (GHashTable *hash_table,
gconstpointer key)
{
GHashNode *node;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
guint node_index;
g_return_val_if_fail (hash_table != NULL, NULL);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
node_index = g_hash_table_lookup_node (hash_table, key);
node = &hash_table->nodes [node_index];
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
return node->key_hash ? node->value : NULL;
}
/**
* g_hash_table_lookup_extended:
* @hash_table: a #GHashTable
* @lookup_key: the key to look up
* @orig_key: return location for the original key, or %NULL
* @value: return location for the value associated with the key, or %NULL
*
* Looks up a key in the #GHashTable, returning the original key and the
* associated value and a #gboolean which is %TRUE if the key was found. This
* is useful if you need to free the memory allocated for the original key,
* for example before calling g_hash_table_remove().
*
* You can actually pass %NULL for @lookup_key to test
* whether the %NULL key exists.
*
* Return value: %TRUE if the key was found in the #GHashTable.
**/
gboolean
g_hash_table_lookup_extended (GHashTable *hash_table,
gconstpointer lookup_key,
gpointer *orig_key,
gpointer *value)
{
GHashNode *node;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
guint node_index;
g_return_val_if_fail (hash_table != NULL, FALSE);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
node_index = g_hash_table_lookup_node (hash_table, lookup_key);
node = &hash_table->nodes [node_index];
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (!node->key_hash)
return FALSE;
if (orig_key)
*orig_key = node->key;
if (value)
*value = node->value;
return TRUE;
}
/*
* g_hash_table_insert_internal:
* @hash_table: our #GHashTable
* @key: the key to insert
* @value: the value to insert
* @keep_new_key: if %TRUE and this key already exists in the table
* then call the destroy notify function on the old key. If %FALSE
* then call the destroy notify function on the new key.
*
* Implements the common logic for the g_hash_table_insert() and
* g_hash_table_replace() functions.
*
* Do a lookup of @key. If it is found, replace it with the new
* @value (and perhaps the new @key). If it is not found, create a
* new node.
*/
static void
g_hash_table_insert_internal (GHashTable *hash_table,
gpointer key,
gpointer value,
gboolean keep_new_key)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *node;
guint node_index;
guint key_hash;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
guint old_hash;
g_return_if_fail (hash_table != NULL);
g_return_if_fail (hash_table->ref_count > 0);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
node_index = g_hash_table_lookup_node_for_insertion (hash_table, key, &key_hash);
node = &hash_table->nodes [node_index];
old_hash = node->key_hash;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (old_hash > 1)
{
if (keep_new_key)
{
if (hash_table->key_destroy_func)
hash_table->key_destroy_func (node->key);
node->key = key;
}
else
{
if (hash_table->key_destroy_func)
hash_table->key_destroy_func (key);
}
if (hash_table->value_destroy_func)
hash_table->value_destroy_func (node->value);
node->value = value;
}
else
{
node->key = key;
node->value = value;
node->key_hash = key_hash;
hash_table->nnodes++;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
if (old_hash == 0)
{
/* We replaced an empty node, and not a tombstone */
hash_table->noccupied++;
g_hash_table_maybe_resize (hash_table);
}
#ifndef G_DISABLE_ASSERT
hash_table->version++;
#endif
}
}
/**
* g_hash_table_insert:
* @hash_table: a #GHashTable.
* @key: a key to insert.
* @value: the value to associate with the key.
*
* Inserts a new key and value into a #GHashTable.
*
* If the key already exists in the #GHashTable its current value is replaced
* with the new value. If you supplied a @value_destroy_func when creating the
* #GHashTable, the old value is freed using that function. If you supplied
* a @key_destroy_func when creating the #GHashTable, the passed key is freed
* using that function.
**/
void
g_hash_table_insert (GHashTable *hash_table,
gpointer key,
gpointer value)
{
g_hash_table_insert_internal (hash_table, key, value, FALSE);
}
/**
* g_hash_table_replace:
* @hash_table: a #GHashTable.
* @key: a key to insert.
* @value: the value to associate with the key.
*
* Inserts a new key and value into a #GHashTable similar to
* g_hash_table_insert(). The difference is that if the key already exists
* in the #GHashTable, it gets replaced by the new key. If you supplied a
* @value_destroy_func when creating the #GHashTable, the old value is freed
* using that function. If you supplied a @key_destroy_func when creating the
* #GHashTable, the old key is freed using that function.
**/
void
g_hash_table_replace (GHashTable *hash_table,
gpointer key,
gpointer value)
{
g_hash_table_insert_internal (hash_table, key, value, TRUE);
1998-06-10 23:21:14 +00:00
}
/*
* g_hash_table_remove_internal:
* @hash_table: our #GHashTable
* @key: the key to remove
* @notify: %TRUE if the destroy notify handlers are to be called
* Return value: %TRUE if a node was found and removed, else %FALSE
*
* Implements the common logic for the g_hash_table_remove() and
* g_hash_table_steal() functions.
*
* Do a lookup of @key and remove it if it is found, calling the
* destroy notify handlers only if @notify is %TRUE.
*/
static gboolean
g_hash_table_remove_internal (GHashTable *hash_table,
gconstpointer key,
gboolean notify)
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
GHashNode *node;
guint node_index;
g_return_val_if_fail (hash_table != NULL, FALSE);
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
node_index = g_hash_table_lookup_node (hash_table, key);
node = &hash_table->nodes [node_index];
/* g_hash_table_lookup_node() never returns a tombstone, so this is safe */
if (!node->key_hash)
return FALSE;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_hash_table_remove_node (hash_table, node, notify);
g_hash_table_maybe_resize (hash_table);
#ifndef G_DISABLE_ASSERT
hash_table->version++;
#endif
return TRUE;
}
/**
* g_hash_table_remove:
* @hash_table: a #GHashTable.
* @key: the key to remove.
*
* Removes a key and its associated value from a #GHashTable.
*
* If the #GHashTable was created using g_hash_table_new_full(), the
* key and value are freed using the supplied destroy functions, otherwise
* you have to make sure that any dynamically allocated values are freed
* yourself.
*
* Return value: %TRUE if the key was found and removed from the #GHashTable.
**/
gboolean
g_hash_table_remove (GHashTable *hash_table,
gconstpointer key)
1998-06-10 23:21:14 +00:00
{
return g_hash_table_remove_internal (hash_table, key, TRUE);
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_steal:
* @hash_table: a #GHashTable.
* @key: the key to remove.
*
* Removes a key and its associated value from a #GHashTable without
* calling the key and value destroy functions.
*
* Return value: %TRUE if the key was found and removed from the #GHashTable.
**/
gboolean
g_hash_table_steal (GHashTable *hash_table,
gconstpointer key)
{
return g_hash_table_remove_internal (hash_table, key, FALSE);
}
/**
* g_hash_table_remove_all:
* @hash_table: a #GHashTable
*
* Removes all keys and their associated values from a #GHashTable.
*
* If the #GHashTable was created using g_hash_table_new_full(), the keys
* and values are freed using the supplied destroy functions, otherwise you
* have to make sure that any dynamically allocated values are freed
* yourself.
*
* Since: 2.12
**/
void
g_hash_table_remove_all (GHashTable *hash_table)
{
g_return_if_fail (hash_table != NULL);
#ifndef G_DISABLE_ASSERT
if (hash_table->nnodes != 0)
hash_table->version++;
#endif
g_hash_table_remove_all_nodes (hash_table, TRUE);
g_hash_table_maybe_resize (hash_table);
}
/**
* g_hash_table_steal_all:
* @hash_table: a #GHashTable.
*
* Removes all keys and their associated values from a #GHashTable
* without calling the key and value destroy functions.
*
* Since: 2.12
**/
void
g_hash_table_steal_all (GHashTable *hash_table)
{
g_return_if_fail (hash_table != NULL);
#ifndef G_DISABLE_ASSERT
if (hash_table->nnodes != 0)
hash_table->version++;
#endif
g_hash_table_remove_all_nodes (hash_table, FALSE);
g_hash_table_maybe_resize (hash_table);
}
/*
* g_hash_table_foreach_remove_or_steal:
* @hash_table: our #GHashTable
* @func: the user's callback function
* @user_data: data for @func
* @notify: %TRUE if the destroy notify handlers are to be called
*
* Implements the common logic for g_hash_table_foreach_remove() and
* g_hash_table_foreach_steal().
*
* Iterates over every node in the table, calling @func with the key
* and value of the node (and @user_data). If @func returns %TRUE the
* node is removed from the table.
*
* If @notify is true then the destroy notify handlers will be called
* for each removed node.
*/
static guint
g_hash_table_foreach_remove_or_steal (GHashTable *hash_table,
GHRFunc func,
gpointer user_data,
gboolean notify)
{
guint deleted = 0;
gint i;
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1 && (* func) (node->key, node->value, user_data))
{
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
g_hash_table_remove_node (hash_table, node, notify);
deleted++;
}
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
}
g_hash_table_maybe_resize (hash_table);
#ifndef G_DISABLE_ASSERT
if (deleted > 0)
hash_table->version++;
#endif
return deleted;
}
/**
* g_hash_table_foreach_remove:
* @hash_table: a #GHashTable.
* @func: the function to call for each key/value pair.
* @user_data: user data to pass to the function.
*
* Calls the given function for each key/value pair in the #GHashTable.
* If the function returns %TRUE, then the key/value pair is removed from the
* #GHashTable. If you supplied key or value destroy functions when creating
* the #GHashTable, they are used to free the memory allocated for the removed
* keys and values.
*
* See #GHashTableIter for an alternative way to loop over the
* key/value pairs in the hash table.
*
* Return value: the number of key/value pairs removed.
**/
guint
g_hash_table_foreach_remove (GHashTable *hash_table,
GHRFunc func,
gpointer user_data)
1998-06-10 23:21:14 +00:00
{
g_return_val_if_fail (hash_table != NULL, 0);
g_return_val_if_fail (func != NULL, 0);
return g_hash_table_foreach_remove_or_steal (hash_table, func, user_data, TRUE);
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_foreach_steal:
* @hash_table: a #GHashTable.
* @func: the function to call for each key/value pair.
* @user_data: user data to pass to the function.
*
* Calls the given function for each key/value pair in the #GHashTable.
* If the function returns %TRUE, then the key/value pair is removed from the
* #GHashTable, but no key or value destroy functions are called.
*
* See #GHashTableIter for an alternative way to loop over the
* key/value pairs in the hash table.
*
* Return value: the number of key/value pairs removed.
**/
guint
g_hash_table_foreach_steal (GHashTable *hash_table,
GHRFunc func,
gpointer user_data)
1998-06-10 23:21:14 +00:00
{
g_return_val_if_fail (hash_table != NULL, 0);
g_return_val_if_fail (func != NULL, 0);
return g_hash_table_foreach_remove_or_steal (hash_table, func, user_data, FALSE);
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_foreach:
* @hash_table: a #GHashTable.
* @func: the function to call for each key/value pair.
* @user_data: user data to pass to the function.
*
* Calls the given function for each of the key/value pairs in the
* #GHashTable. The function is passed the key and value of each
* pair, and the given @user_data parameter. The hash table may not
* be modified while iterating over it (you can't add/remove
* items). To remove all items matching a predicate, use
* g_hash_table_foreach_remove().
*
* See g_hash_table_find() for performance caveats for linear
* order searches in contrast to g_hash_table_lookup().
**/
1998-06-10 23:21:14 +00:00
void
g_hash_table_foreach (GHashTable *hash_table,
GHFunc func,
gpointer user_data)
1998-06-10 23:21:14 +00:00
{
gint i;
g_return_if_fail (hash_table != NULL);
g_return_if_fail (func != NULL);
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1)
(* func) (node->key, node->value, user_data);
}
1998-06-10 23:21:14 +00:00
}
/**
* g_hash_table_find:
* @hash_table: a #GHashTable.
* @predicate: function to test the key/value pairs for a certain property.
* @user_data: user data to pass to the function.
*
* Calls the given function for key/value pairs in the #GHashTable until
* @predicate returns %TRUE. The function is passed the key and value of
2004-02-27 21:11:51 +00:00
* each pair, and the given @user_data parameter. The hash table may not
* be modified while iterating over it (you can't add/remove items).
*
* Note, that hash tables are really only optimized for forward lookups,
* i.e. g_hash_table_lookup().
* So code that frequently issues g_hash_table_find() or
* g_hash_table_foreach() (e.g. in the order of once per every entry in a
* hash table) should probably be reworked to use additional or different
* data structures for reverse lookups (keep in mind that an O(n) find/foreach
* operation issued for all n values in a hash table ends up needing O(n*n)
* operations).
2004-02-27 21:11:51 +00:00
*
* Return value: The value of the first key/value pair is returned, for which
* func evaluates to %TRUE. If no pair with the requested property is found,
2004-02-27 21:11:51 +00:00
* %NULL is returned.
*
* Since: 2.4
**/
gpointer
g_hash_table_find (GHashTable *hash_table,
GHRFunc predicate,
gpointer user_data)
{
gint i;
g_return_val_if_fail (hash_table != NULL, NULL);
g_return_val_if_fail (predicate != NULL, NULL);
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1 && predicate (node->key, node->value, user_data))
return node->value;
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
}
return NULL;
}
/**
* g_hash_table_size:
* @hash_table: a #GHashTable.
*
* Returns the number of elements contained in the #GHashTable.
*
* Return value: the number of key/value pairs in the #GHashTable.
**/
guint
g_hash_table_size (GHashTable *hash_table)
{
g_return_val_if_fail (hash_table != NULL, 0);
return hash_table->nnodes;
}
1998-06-10 23:21:14 +00:00
/**
* g_hash_table_get_keys:
* @hash_table: a #GHashTable
*
* Retrieves every key inside @hash_table. The returned data is valid
* until @hash_table is modified.
*
* Return value: a #GList containing all the keys inside the hash
* table. The content of the list is owned by the hash table and
* should not be modified or freed. Use g_list_free() when done
* using the list.
*
* Since: 2.14
*/
GList *
g_hash_table_get_keys (GHashTable *hash_table)
{
gint i;
GList *retval;
g_return_val_if_fail (hash_table != NULL, NULL);
retval = NULL;
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1)
retval = g_list_prepend (retval, node->key);
}
return retval;
}
/**
* g_hash_table_get_values:
* @hash_table: a #GHashTable
*
* Retrieves every value inside @hash_table. The returned data is
* valid until @hash_table is modified.
*
* Return value: a #GList containing all the values inside the hash
* table. The content of the list is owned by the hash table and
* should not be modified or freed. Use g_list_free() when done
* using the list.
*
* Since: 2.14
*/
GList *
g_hash_table_get_values (GHashTable *hash_table)
{
gint i;
GList *retval;
g_return_val_if_fail (hash_table != NULL, NULL);
retval = NULL;
for (i = 0; i < hash_table->size; i++)
Rewrite most of GHashTable to use open addressing with quadratic probing 2008-09-19 Hans Petter Jansson <hpj@novell.com> Rewrite most of GHashTable to use open addressing with quadratic probing instead of chaining. This has the potential to reduce memory fragmentation significantly, while being slightly faster due to better locality and no need to call alloc/free functions for nodes. Benchmarks suggest it also uses less memory overall. * glib/ghash.c (prime_mod): Table of suitable primes for initial-probe distribution. (g_hash_table_set_shift): New function. (g_hash_table_find_closest_shift): New function. (g_hash_table_set_shift_from_size): New function. (g_hash_table_lookup_node_for_insertion): New function. (g_hash_table_lookup_node): Rewritten to return node index instead of pointer, use quadratic probe on flat table, and not return insertion data. The latter saves some computation for read-only lookups. (g_hash_table_remove_node): Rewrite to take a pointer directly to the node structure to remove, and clear that. Remove unlinking code. (g_hash_table_remove_all_nodes): Rewrite to not clear nodes individually, but en masse using memset () after potentially calling notify functions. (iter_remove_or_steal): Use new data structure and algorithm. Vastly simplified - now just a call to g_hash_table_remove_node (). (g_hash_table_resize): New resize code, re-indexing with new prime and cleaning up tombstones. (g_hash_table_maybe_resize): Table may hold 8 buckets minimum, no less than 1/4 load excluding tombstones, and no more than 15/16 load including tombstones. These numbers are the results of a lot of benchmarking with multiple complex applications, and should not be changed lightly. (g_hash_table_iter_next) (g_hash_table_lookup) (g_hash_table_lookup_extended) (g_hash_table_insert_internal) (g_hash_table_remove_internal) (g_hash_table_foreach_remove_or_steal) (g_hash_table_foreach) (g_hash_table_find) (g_hash_table_get_keys) (g_hash_table_get_values): Use new data structure and algorithm, fairly trivial changes. svn path=/trunk/; revision=7518
2008-09-20 04:05:11 +00:00
{
GHashNode *node = &hash_table->nodes [i];
if (node->key_hash > 1)
retval = g_list_prepend (retval, node->value);
}
return retval;
}