#include <snappy.h>
+#include <iostream>
+
#include <assert.h>
#include <string.h>
* The file is composed of a number of chunks, they are:
* chunk {
* uint32 - specifying the length of the compressed data
- * compressed data
+ * compressed data, in little endian
* }
* File can contain any number of such chunks.
* The default size of an uncompressed chunk is specified in
m_cachePtr(0),
m_cacheSize(0)
{
- m_compressedCache = new char[SNAPPY_CHUNK_SIZE];
+ size_t maxCompressedLength =
+ snappy::MaxCompressedLength(SNAPPY_CHUNK_SIZE);
+ m_compressedCache = new char[maxCompressedLength];
}
SnappyFile::~SnappyFile()
{
delete [] m_compressedCache;
+ delete [] m_cache;
}
bool SnappyFile::rawOpen(const std::string &filename, File::Mode mode)
return m_stream.is_open();
}
-bool SnappyFile::rawWrite(const void *buffer, int length)
+bool SnappyFile::rawWrite(const void *buffer, size_t length)
{
if (freeCacheSize() > length) {
memcpy(m_cachePtr, buffer, length);
while (sizeToWrite >= freeCacheSize()) {
int endSize = freeCacheSize();
int offset = length - sizeToWrite;
- memcpy(m_cachePtr, (char*)buffer + offset, endSize);
+ memcpy(m_cachePtr, (const char*)buffer + offset, endSize);
sizeToWrite -= endSize;
m_cachePtr += endSize;
flushCache();
}
if (sizeToWrite) {
int offset = length - sizeToWrite;
- memcpy(m_cachePtr, (char*)buffer + offset, sizeToWrite);
+ memcpy(m_cachePtr, (const char*)buffer + offset, sizeToWrite);
m_cachePtr += sizeToWrite;
}
}
return true;
}
-bool SnappyFile::rawRead(void *buffer, int length)
+bool SnappyFile::rawRead(void *buffer, size_t length)
{
- if (m_stream.eof()) {
+ if (endOfData()) {
return false;
}
- if (freeCacheSize() > length) {
- memcpy(buffer, m_cachePtr, length);
- m_cachePtr += length;
- } else if (freeCacheSize() == length) {
+
+ if (freeCacheSize() >= length) {
memcpy(buffer, m_cachePtr, length);
m_cachePtr += length;
- flushCache();
} else {
- int sizeToRead = length;
- int offset = 0;
+ size_t sizeToRead = length;
+ size_t offset = 0;
while (sizeToRead) {
- int chunkSize = std::min(freeCacheSize(), sizeToRead);
+ size_t chunkSize = std::min(freeCacheSize(), sizeToRead);
offset = length - sizeToRead;
memcpy((char*)buffer + offset, m_cachePtr, chunkSize);
m_cachePtr += chunkSize;
void SnappyFile::flushCache()
{
if (m_mode == File::Write) {
- size_t compressedLength;
+ size_t inputLength = usedCacheSize();
+
+ if (inputLength) {
+ size_t compressedLength;
- ::snappy::RawCompress(m_cache, SNAPPY_CHUNK_SIZE - freeCacheSize(),
- m_compressedCache, &compressedLength);
+ ::snappy::RawCompress(m_cache, inputLength,
+ m_compressedCache, &compressedLength);
- writeCompressedLength(compressedLength);
- m_stream.write(m_compressedCache, compressedLength);
- m_cachePtr = m_cache;
+ writeCompressedLength(compressedLength);
+ m_stream.write(m_compressedCache, compressedLength);
+ m_cachePtr = m_cache;
+ }
+ assert(m_cachePtr == m_cache);
} else if (m_mode == File::Read) {
- if (m_stream.eof())
- return;
//assert(m_cachePtr == m_cache + m_cacheSize);
+ m_currentOffset.chunk = m_stream.tellg();
size_t compressedLength;
compressedLength = readCompressedLength();
- m_stream.read((char*)m_compressedCache, compressedLength);
- ::snappy::GetUncompressedLength(m_compressedCache, compressedLength,
- &m_cacheSize);
- if (m_cache)
- delete [] m_cache;
- createCache(m_cacheSize);
- ::snappy::RawUncompress(m_compressedCache, compressedLength,
- m_cache);
+
+ if (compressedLength) {
+ m_stream.read((char*)m_compressedCache, compressedLength);
+ ::snappy::GetUncompressedLength(m_compressedCache, compressedLength,
+ &m_cacheSize);
+ createCache(m_cacheSize);
+ ::snappy::RawUncompress(m_compressedCache, compressedLength,
+ m_cache);
+ } else {
+ createCache(0);
+ }
}
}
void SnappyFile::createCache(size_t size)
{
- m_cache = new char[size];
+ // TODO: only re-allocate if the current buffer is not big enough
+
+ if (m_cache) {
+ delete [] m_cache;
+ }
+
+ if (size) {
+ m_cache = new char[size];
+ } else {
+ m_cache = NULL;
+ }
+
m_cachePtr = m_cache;
m_cacheSize = size;
}
-void SnappyFile::writeCompressedLength(uint32_t value)
+void SnappyFile::writeCompressedLength(size_t length)
+{
+ unsigned char buf[4];
+ buf[0] = length & 0xff; length >>= 8;
+ buf[1] = length & 0xff; length >>= 8;
+ buf[2] = length & 0xff; length >>= 8;
+ buf[3] = length & 0xff; length >>= 8;
+ assert(length == 0);
+ m_stream.write((const char *)buf, sizeof buf);
+}
+
+size_t SnappyFile::readCompressedLength()
+{
+ unsigned char buf[4];
+ size_t length;
+ m_stream.read((char *)buf, sizeof buf);
+ if (m_stream.fail()) {
+ length = 0;
+ } else {
+ length = (size_t)buf[0];
+ length |= ((size_t)buf[1] << 8);
+ length |= ((size_t)buf[2] << 16);
+ length |= ((size_t)buf[3] << 24);
+ }
+ return length;
+}
+
+bool SnappyFile::supportsOffsets() const
{
- m_stream.write((const char*)&value, sizeof value);
+ return true;
}
-uint32_t SnappyFile::readCompressedLength()
+File::Offset SnappyFile::currentOffset()
{
- uint32_t len;
- m_stream.read((char*)&len, sizeof len);
- return len;
+ m_currentOffset.offsetInChunk = m_cachePtr - m_cache;
+ return m_currentOffset;
+}
+
+void SnappyFile::setCurrentOffset(const File::Offset &offset)
+{
+ // to remove eof bit
+ m_stream.clear();
+ // seek to the start of a chunk
+ m_stream.seekg(offset.chunk, std::ios::beg);
+ // load the chunk
+ flushCache();
+ assert(m_cacheSize >= offset.offsetInChunk);
+ // seek within our cache to the correct location within the chunk
+ m_cachePtr = m_cache + offset.offsetInChunk;
+
+}
+
+bool SnappyFile::rawSkip(unsigned length)
+{
+ if (endOfData()) {
+ return false;
+ }
+
+ if (freeCacheSize() >= length) {
+ m_cachePtr += length;
+ } else {
+ size_t sizeToRead = length;
+ while (sizeToRead) {
+ size_t chunkSize = std::min(freeCacheSize(), sizeToRead);
+ m_cachePtr += chunkSize;
+ sizeToRead -= chunkSize;
+ if (sizeToRead > 0)
+ flushCache();
+ if (!m_cacheSize)
+ break;
+ }
+ }
+
+ return true;
}