Keep order of nodes when wxHashMap is resized; this ensures
that wxHashTable has the same behaviour when wxUSE_STL=1 as when wxUSE_STL=0. git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/trunk@23052 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775
This commit is contained in:
parent
6f2ec3c3de
commit
8ae94823a3
@ -70,6 +70,7 @@ protected:
|
||||
static void CopyHashTable( _wxHashTable_NodeBase** srcTable,
|
||||
size_t srcBuckets, _wxHashTableBase2* dst,
|
||||
_wxHashTable_NodeBase** dstTable,
|
||||
size_t dstBuckets,
|
||||
BucketFromNode func, ProcessNode proc );
|
||||
|
||||
static void** AllocTable( size_t sz )
|
||||
@ -357,7 +358,7 @@ protected: \
|
||||
m_tableBuckets = newSize; \
|
||||
\
|
||||
CopyHashTable( (_wxHashTable_NodeBase**)srcTable, srcBuckets, \
|
||||
this, (_wxHashTable_NodeBase**)m_table, \
|
||||
this, (_wxHashTable_NodeBase**)m_table, newSize, \
|
||||
(BucketFromNode)GetBucketForNode,\
|
||||
(ProcessNode)&DummyProcessNode ); \
|
||||
free(srcTable); \
|
||||
@ -369,7 +370,7 @@ protected: \
|
||||
ResizeTable( ht.size() ); \
|
||||
CopyHashTable( (_wxHashTable_NodeBase**)ht.m_table, ht.m_tableBuckets,\
|
||||
(_wxHashTableBase2*)this, \
|
||||
(_wxHashTable_NodeBase**)m_table, \
|
||||
(_wxHashTable_NodeBase**)m_table, m_tableBuckets, \
|
||||
(BucketFromNode)GetBucketForNode, \
|
||||
(ProcessNode)CopyNode ); \
|
||||
} \
|
||||
|
@ -1151,7 +1151,7 @@ static void TestHash()
|
||||
wxPuts(_T("*** Testing wxHashTable ***\n"));
|
||||
|
||||
{
|
||||
wxHashTable hash(wxKEY_INTEGER), hash2(wxKEY_STRING);
|
||||
wxHashTable hash(wxKEY_INTEGER, 10), hash2(wxKEY_STRING);
|
||||
wxObject o;
|
||||
int i;
|
||||
|
||||
|
@ -128,9 +128,20 @@ void _wxHashTableBase2::CopyHashTable( _wxHashTable_NodeBase** srcTable,
|
||||
size_t srcBuckets,
|
||||
_wxHashTableBase2* dst,
|
||||
_wxHashTable_NodeBase** dstTable,
|
||||
size_t dstBuckets,
|
||||
BucketFromNode func, ProcessNode proc )
|
||||
{
|
||||
for( size_t i = 0; i < srcBuckets; ++i )
|
||||
// for compatibility with wxHashTable (to avoid reimplementig it
|
||||
// from scratch), we need to preserve the order of nodes in a
|
||||
// source bucket when copying the table, hence, to avoid
|
||||
// allocating an auxiliary table we use a circular list for each
|
||||
// bucket, and we keep the *tail* of each list in dstTable[i], to
|
||||
// be able to append nodes in O(1) time. Wen we're done copying,
|
||||
// we adjust dstTable[i] to point at the head of the list and we
|
||||
// break the circular list into a linear one.
|
||||
size_t i;
|
||||
|
||||
for( i = 0; i < srcBuckets; ++i )
|
||||
{
|
||||
_wxHashTable_NodeBase* nextnode;
|
||||
|
||||
@ -140,8 +151,24 @@ void _wxHashTableBase2::CopyHashTable( _wxHashTable_NodeBase** srcTable,
|
||||
|
||||
nextnode = node->m_nxt;
|
||||
_wxHashTable_NodeBase* newnode = proc( node );
|
||||
newnode->m_nxt = dstTable[bucket];
|
||||
dstTable[bucket] = newnode;
|
||||
if( dstTable[bucket] )
|
||||
{
|
||||
newnode->m_nxt = dstTable[bucket]->m_nxt; // head of the list
|
||||
dstTable[bucket]->m_nxt = newnode;
|
||||
dstTable[bucket] = newnode;
|
||||
}
|
||||
else
|
||||
dstTable[bucket] = newnode->m_nxt = newnode;
|
||||
}
|
||||
}
|
||||
|
||||
for( i = 0; i < dstBuckets; ++i )
|
||||
{
|
||||
if( dstTable[i] )
|
||||
{
|
||||
_wxHashTable_NodeBase* tmp = dstTable[i];
|
||||
dstTable[i] = dstTable[i]->m_nxt;
|
||||
tmp->m_nxt = NULL;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user