/* GRAPHITE2 LICENSING
Copyright 2010, SIL International
All rights reserved.
This library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published
by the Free Software Foundation; either version 2.1 of License, or
(at your option) any later version.
This program 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.
You should also have received a copy of the GNU Lesser General Public
License along with this library in the file named "LICENSE".
If not, write to the Free Software Foundation, 51 Franklin Street,
Suite 500, Boston, MA 02110-1335, USA or visit their web page on the
internet at http://www.fsf.org/licenses/lgpl.html.
Alternatively, the contents of this file may be used under the terms of the
Mozilla Public License (http://mozilla.org/MPL) or the GNU General Public
License, as published by the Free Software Foundation, either version 2
of the License or (at your option) any later version.
*/
#pragma once
#ifndef GRAPHITE2_NSEGCACHE
#include <graphite2/Segment.h>
#include "inc/Main.h"
#include "inc/Slot.h"
#include "inc/FeatureVal.h"
#include "inc/SegCacheEntry.h"
#include "inc/Segment.h"
namespace graphite2 {
class SegCache;
class SegCacheEntry;
class SegCacheStore;
/**
* SegPrefixEntry stores lists of word/syllable segments
* with one list for each word length. The prefix size should be chosen so that
* these list sizes stay small since they will be searched iteratively.
*/
class SegCachePrefixEntry
{
SegCachePrefixEntry(const SegCachePrefixEntry &);
SegCachePrefixEntry & operator = (const SegCachePrefixEntry &);
public:
SegCachePrefixEntry() : m_lastPurge(0)
{
memset(m_entryCounts, 0, sizeof m_entryCounts);
memset(m_entryBSIndex, 0, sizeof m_entryBSIndex);
memset(m_entries, 0, sizeof m_entries);
}
~SegCachePrefixEntry()
{
for (size_t j = 0; j < eMaxSpliceSize; j++)
{
if (m_entryCounts[j])
{
assert(m_entries[j]);
for (size_t k = 0; k < m_entryCounts[j]; k++)
m_entries[j][k].clear();
free(m_entries[j]);
}
}
}
const SegCacheEntry * find(const uint16 * cmapGlyphs, size_t length) const
{
if (length <= ePrefixLength)
{
assert(m_entryCounts[length-1] <= 1);
if (m_entries[length-1])
return m_entries[length-1];
return NULL;
}
SegCacheEntry * entry = NULL;
findPosition(cmapGlyphs, length, &entry);
return entry;
}
SegCacheEntry * cache(const uint16* cmapGlyphs, size_t length, Segment * seg, size_t charOffset, unsigned long long totalAccessCount)
{
size_t listSize = m_entryBSIndex[length-1]? (m_entryBSIndex[length-1] << 1) - 1 : 0;
SegCacheEntry * newEntries = NULL;
if (m_entryCounts[length-1] + 1u > listSize)
{
if (m_entryCounts[length-1] == 0)
listSize = 1;
else
{
// the problem comes when you get incremental numeric ids in a large doc
if (listSize >= eMaxSuffixCount)
return NULL;
listSize = (m_entryBSIndex[length-1] << 2) - 1;
}
newEntries = gralloc<SegCacheEntry>(listSize);
if (!newEntries)
return NULL;
}
uint16 insertPos = 0;
if (m_entryCounts[length-1] > 0)
{
insertPos = findPosition(cmapGlyphs, length, NULL);
if (!newEntries)
{
// same buffer, shift entries up
memmove(m_entries[length-1] + insertPos + 1, m_entries[length-1] + insertPos,
sizeof(SegCacheEntry) * (m_entryCounts[length-1] - insertPos));
}
else
{
memcpy(newEntries, m_entries[length-1], sizeof(SegCacheEntry) * (insertPos));
memcpy(newEntries + insertPos + 1, m_entries[length-1] + insertPos,
sizeof(SegCacheEntry) * (m_entryCounts[length-1] - insertPos));
free(m_entries[length-1]);
m_entries[length-1] = newEntries;
assert (m_entryBSIndex[length-1]);
m_entryBSIndex[length-1] <<= 1;
}
}
else
{
m_entryBSIndex[length-1] = 1;
m_entries[length-1] = newEntries;
}
m_entryCounts[length-1] += 1;
new (m_entries[length-1] + insertPos)
SegCacheEntry(cmapGlyphs, length, seg, charOffset, totalAccessCount);
return m_entries[length-1] + insertPos;
}
uint32 purge(unsigned long long minAccessCount, unsigned long long oldAccessTime,
unsigned long long currentTime);
CLASS_NEW_DELETE
private:
uint16 findPosition(const uint16 * cmapGlyphs, uint16 length, SegCacheEntry ** entry) const
{
int dir = 0;
if (m_entryCounts[length-1] == 0)
{
if (entry) *entry = NULL;
return 0;
}
else if (m_entryCounts[length-1] == 1)
{
// optimize single entry case
for (int i = ePrefixLength; i < length; i++)
{
if (cmapGlyphs[i] > m_entries[length-1][0].m_unicode[i])
{
return 1;
}
else if (cmapGlyphs[i] < m_entries[length-1][0].m_unicode[i])
{
return 0;
}
}
if (entry)
*entry = m_entries[length-1];
return 0;
}
uint16 searchIndex = m_entryBSIndex[length-1] - 1;
uint16 stepSize = m_entryBSIndex[length-1] >> 1;
size_t prevIndex = searchIndex;
do
{
dir = 0;
if (searchIndex >= m_entryCounts[length-1])
{
dir = -1;
searchIndex -= stepSize;
stepSize >>= 1;
}
else
{
for (int i = ePrefixLength; i < length; i++)
{
if (cmapGlyphs[i] > m_entries[length-1][searchIndex].m_unicode[i])
{
dir = 1;
searchIndex += stepSize;
stepSize >>= 1;
break;
}
else if (cmapGlyphs[i] < m_entries[length-1][searchIndex].m_unicode[i])
{
dir = -1;
searchIndex -= stepSize;
stepSize >>= 1;
break;
}
}
}
if (prevIndex == searchIndex)
break;
prevIndex = searchIndex;
} while (dir != 0);
if (entry)
{
if (dir == 0)
*entry = m_entries[length-1] + searchIndex;
else
*entry = NULL;
}
else
{
// if entry is null, then this is for inserting a new value, which
// shouldn't already be in the cache
assert(dir != 0);
if (dir > 0)
++searchIndex;
}
return searchIndex;
}
/** m_entries is a null terminated list of entries */
uint16 m_entryCounts[eMaxSpliceSize];
uint16 m_entryBSIndex[eMaxSpliceSize];
SegCacheEntry * m_entries[eMaxSpliceSize];
unsigned long long m_lastPurge;
};
#define SEG_CACHE_MIN_INDEX (store->maxCmapGid())
#define SEG_CACHE_MAX_INDEX (store->maxCmapGid()+1u)
#define SEG_CACHE_UNSET_INDEX (store->maxCmapGid()+2u)
union SegCachePrefixArray
{
void ** raw;
SegCachePrefixArray * array;
SegCachePrefixEntry ** prefixEntries;
uintptr * range;
};
class SegCache
{
public:
SegCache(const SegCacheStore * store, const Features& features);
~SegCache();
const SegCacheEntry * find(const uint16 * cmapGlyphs, size_t length) const;
SegCacheEntry * cache(SegCacheStore * store, const uint16 * cmapGlyphs, size_t length, Segment * seg, size_t charOffset);
void purge(SegCacheStore * store);
long long totalAccessCount() const { return m_totalAccessCount; }
size_t segmentCount() const { return m_segmentCount; }
const Features & features() const { return m_features; }
void clear(SegCacheStore * store);
CLASS_NEW_DELETE
private:
void freeLevel(SegCacheStore * store, SegCachePrefixArray prefixes, size_t level);
void purgeLevel(SegCacheStore * store, SegCachePrefixArray prefixes, size_t level,
unsigned long long minAccessCount, unsigned long long oldAccessTime);
uint16 m_prefixLength;
// uint16 m_maxCachedSegLength;
size_t m_segmentCount;
SegCachePrefixArray m_prefixes;
Features m_features;
mutable unsigned long long m_totalAccessCount;
mutable unsigned long long m_totalMisses;
float m_purgeFactor;
};
inline const SegCacheEntry * SegCache::find(const uint16 * cmapGlyphs, size_t length) const
{
uint16 pos = 0;
if (!length || length > eMaxSpliceSize) return NULL;
SegCachePrefixArray pEntry = m_prefixes.array[cmapGlyphs[0]];
while (++pos < m_prefixLength - 1)
{
if (!pEntry.raw)
{
++m_totalMisses;
return NULL;
}
pEntry = pEntry.array[(pos < length)? cmapGlyphs[pos] : 0];
}
if (!pEntry.raw)
{
++m_totalMisses;
return NULL;
}
SegCachePrefixEntry * prefixEntry = pEntry.prefixEntries[(pos < length)? cmapGlyphs[pos] : 0];
if (!prefixEntry)
{
++m_totalMisses;
return NULL;
}
const SegCacheEntry * entry = prefixEntry->find(cmapGlyphs, length);
if (entry)
{
++m_totalAccessCount;
entry->accessed(m_totalAccessCount);
}
else
{
++m_totalMisses;
}
return entry;
}
} // namespace graphite2
#endif