libs-base/Source/HashTable.m
netc 5f7871270a Remove dependency upon config.h by headers files and include
directly in source files because the config.h file is system
dependent, used just for compiling the source, and should
not be installed.
Some minor bug fixes.


git-svn-id: svn+ssh://svn.gna.org/svn/gnustep/libs/base/trunk@2619 72102866-910b-0410-8b05-ffd578937521
1997-11-06 00:51:23 +00:00

362 lines
7.7 KiB
Objective-C

/* Implementation of Objective C NeXT-compatible HashTable object
Copyright (C) 1993,1994 Free Software Foundation, Inc.
Written by: Andrew Kachites McCallum <mccallum@gnu.ai.mit.edu>
Date: May 1993
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public
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
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with this library; if not, write to the Free
Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <config.h>
#include <gnustep/base/preface.h>
#include <objc/HashTable.h>
#define DEFAULT_HASH_CAPACITY 32
/* Some useful hash and compare functions not provided by hash.h */
static inline unsigned int
hash_object (cache_ptr cache, const void *key)
{
return (([((id)key) hash]) & cache->mask);
}
static inline int
compare_objects (const void *k1, const void *k2)
{
return (int)[(id)k1 isEqual:(id)k2];
}
static inline unsigned int
hash_int (cache_ptr cache, const void *key)
{
return ((unsigned int)key & cache->mask);
}
static inline int
compare_ints (const void *k1, const void *k2)
{
return !((int)k1 - (int)k2);
}
static inline int
compare_long_ints (const void *k1, const void *k2)
{
return !((long int)k1 - (long int)k2);
}
@implementation HashTable
+ initialize
{
if (self == [HashTable class])
[self setVersion:0]; /* beta release */
return self;
}
- initKeyDesc: (const char *)aKeyDesc
valueDesc: (const char *)aValueDesc
capacity: (unsigned) aCapacity
{
hash_func_type hf;
compare_func_type cf;
if (!aKeyDesc)
[self error:"in %s, NULL keyDesc\n", sel_get_name(_cmd)];
if (!aValueDesc)
[self error:"in %s, NULL valueDesc\n", sel_get_name(_cmd)];
count = 0;
keyDesc = aKeyDesc;
valueDesc = aValueDesc;
switch (*aKeyDesc)
{
case _C_ATOM:
case _C_CHARPTR :
hf = (hash_func_type)hash_string;
cf = (compare_func_type)compare_strings;
break;
case _C_ID:
case _C_CLASS:
hf = (hash_func_type)hash_object;
cf = (compare_func_type)compare_objects;
break;
case _C_PTR:
hf = (hash_func_type)hash_ptr;
cf = (compare_func_type)compare_ptrs;
break;
case _C_INT:
case _C_SEL:
case _C_UINT:
hf = (hash_func_type)hash_int;
cf = (compare_func_type)compare_ints;
break;
case _C_LNG:
case _C_ULNG:
hf = (hash_func_type)hash_int;
cf = (compare_func_type)compare_long_ints;
break;
case _C_FLT:
/* Fix this. Do something better with floats. */
hf = (hash_func_type)hash_int;
cf = (compare_func_type)compare_ints;
break;
default:
hf = (hash_func_type)hash_int;
cf = (compare_func_type)compare_ints;
break;
}
_buckets = hash_new(aCapacity, hf, cf);
_nbBuckets = _buckets->size;
return self;
}
- initKeyDesc:(const char *)aKeyDesc
valueDesc:(const char *)aValueDesc
{
return [self initKeyDesc:aKeyDesc
valueDesc:aValueDesc
capacity:DEFAULT_HASH_CAPACITY];
}
- initKeyDesc: (const char *)aKeyDesc
{
return [self initKeyDesc:aKeyDesc
valueDesc:@encode(id)];
}
- init
{
return [self initKeyDesc:@encode(id)];
}
- free
{
hash_delete(_buckets);
return [super free];
}
- freeObjects
{
node_ptr node;
void *val;
while ((node = hash_next(_buckets, 0)))
{
val = node->value;
hash_remove(_buckets, node->key);
if (*valueDesc == _C_ID)
[(id)val free];
}
count = 0;
_nbBuckets = _buckets->size;
return self;
}
- freeKeys:(void (*) (void *))keyFunc
values:(void (*) (void *))valueFunc
{
/* What exactly is this supposed to do? */
[self notImplemented:_cmd];
return self;
}
- empty
{
node_ptr node;
while ((node = hash_next(_buckets, 0)))
hash_remove(_buckets, node->key);
count = 0;
_nbBuckets = _buckets->size;
return self;
}
- shallowCopy
{
HashTable *c;
node_ptr node;
c = [super shallowCopy];
c->_buckets = hash_new(_buckets->size,
_buckets->hash_func,
_buckets->compare_func);
/* copy nodes to new copy */
node = 0;
while ((node = hash_next(_buckets, node)))
[c insertKey:node->key value:node->value];
return c;
}
- deepen
{
node_ptr node = 0;
if (*valueDesc == _C_ID)
{
while ((node = hash_next(_buckets, node)))
{
node->value = [(id)(node->value) deepCopy];
}
}
/* If the hashtable contains strings should we copy them too??
But we definitely shouldn't copy "%" keys. */
return self;
}
- (unsigned) count
{
return count;
}
- (BOOL) isKey:(const void *)aKey
{
return hash_is_key_in_hash(_buckets, aKey);
}
- (void *) valueForKey:(const void *)aKey
{
return hash_value_for_key(_buckets, aKey);
}
- (void *) insertKey:(const void *)aKey value:(void *)aValue
{
void *prevValue;
prevValue = hash_value_for_key(_buckets, aKey);
if (prevValue)
hash_remove(_buckets, aKey);
hash_add(&_buckets, aKey, aValue);
count = _buckets->used;
_nbBuckets = _buckets->size;
return prevValue;
}
- (void *) removeKey:(const void *)aKey
{
if (hash_value_for_key(_buckets, aKey))
{
hash_remove(_buckets, aKey);
count = _buckets->used;
_nbBuckets = _buckets->size;
}
return nil;
}
- (NXHashState) initState
{
return (NXHashState) 0;
}
- (BOOL) nextState:(NXHashState *)aState
key:(const void **)aKey
value:(void **)aValue
{
*aState = hash_next(_buckets, *aState);
if (*aState)
{
*aKey = (*aState)->key;
*aValue = (*aState)->value;
return YES;
}
else
return NO;
}
- write: (TypedStream*)aStream
{
NXHashState state = [self initState];
const void *k;
void *v;
if (!strcmp(keyDesc, "%"))
[self error:"Archiving atom strings, @encode()=\"%\", not yet handled"];
[super write: aStream];
objc_write_types(aStream, "II**",
[self count], _nbBuckets, keyDesc, valueDesc);
while ([self nextState:&state key:&k value:&v])
{
objc_write_type(aStream, keyDesc, &k);
objc_write_type(aStream, valueDesc, &v);
}
return self;
}
- read: (TypedStream*)aStream
{
unsigned cnt, capacity;
int i;
const void *k;
void *v;
[super read:aStream];
objc_read_types(aStream, "II**",
&cnt, &capacity, &keyDesc, &valueDesc);
if (!strcmp(keyDesc, "%"))
[self error:"Archiving atom strings, @encode()=\"%\", not yet handled"];
[self initKeyDesc:keyDesc valueDesc:valueDesc capacity:capacity];
for (i = 0; i < cnt; i++)
{
objc_read_type(aStream, keyDesc, &k);
objc_read_type(aStream, valueDesc, &v);
[self insertKey:k value:v];
}
return self;
}
+ newKeyDesc: (const char *)aKeyDesc
{
return [[[self class] alloc] initKeyDesc:aKeyDesc];
}
+ newKeyDesc:(const char *)aKeyDesc
valueDesc:(const char *)aValueDesc
{
return [[self alloc]
initKeyDesc:aKeyDesc
valueDesc:aValueDesc];
}
+ newKeyDesc:(const char *)aKeyDesc
valueDesc:(const char *)aValueDesc
capacity:(unsigned)aCapacity
{
return [[self alloc]
initKeyDesc:aKeyDesc
valueDesc:aValueDesc
capacity:aCapacity];
}
- makeObjectsPerform:(SEL)aSel
{
node_ptr node = 0;
while ((node = hash_next(_buckets, node)))
[(id)(node->value) perform:aSel];
return self;
}
- makeObjectsPerform:(SEL)aSel with:anObject
{
node_ptr node = 0;
while ((node = hash_next(_buckets, node)))
[(id)(node->value) perform:aSel with:anObject];
return self;
}
@end