src/api/internal/BgzfStream_p.cpp

   1 // ***************************************************************************
   2 // BgzfStream_p.cpp (c) 2011 Derek Barnett
   3 // Marth Lab, Department of Biology, Boston College
   4 // ---------------------------------------------------------------------------
   5 // Last modified: 10 October 2011(DB)
   6 // ---------------------------------------------------------------------------
   7 // Based on BGZF routines developed at the Broad Institute.
   8 // Provides the basic functionality for reading & writing BGZF files
   9 // Replaces the old BGZF.* files to avoid clashing with other toolkits
  10 // ***************************************************************************
  11
  12 #include "api/BamAux.h"
  13 #include "api/BamConstants.h"
  14 #include "api/internal/BamDeviceFactory_p.h"
  15 #include "api/internal/BamException_p.h"
  16 #include "api/internal/BgzfStream_p.h"
  17 using namespace BamTools;
  18 using namespace BamTools::Internal;
  19
  20 #include "zlib.h"
  21
  22 #include <cstring>
  23 #include <algorithm>
  24 #include <iostream>
  25 #include <sstream>
  26 using namespace std;
  27
  28 // ----------------------------
  29 // RaiiWrapper implementation
  30 // ----------------------------
  31
  32 BgzfStream::RaiiWrapper::RaiiWrapper(void) {
  33     CompressedBlock   = new char[Constants::BGZF_MAX_BLOCK_SIZE];
  34     UncompressedBlock = new char[Constants::BGZF_DEFAULT_BLOCK_SIZE];
  35 }
  36
  37 BgzfStream::RaiiWrapper::~RaiiWrapper(void) {
  38
  39     // clean up buffers
  40     delete[] CompressedBlock;
  41     delete[] UncompressedBlock;
  42     CompressedBlock = 0;
  43     UncompressedBlock = 0;
  44 }
  45
  46 // ---------------------------
  47 // BgzfStream implementation
  48 // ---------------------------
  49
  50 // constructor
  51 BgzfStream::BgzfStream(void)
  52   : m_blockLength(0)
  53   , m_blockOffset(0)
  54   , m_blockAddress(0)
  55   , m_isWriteCompressed(true)
  56   , m_device(0)
  57 { }
  58
  59 // destructor
  60 BgzfStream::~BgzfStream(void) {
  61     Close();
  62 }
  63
  64 // checks BGZF block header
  65 bool BgzfStream::CheckBlockHeader(char* header) {
  66     return (header[0] == Constants::GZIP_ID1 &&
  67             header[1] == Constants::GZIP_ID2 &&
  68             header[2] == Z_DEFLATED &&
  69             (header[3] & Constants::FLG_FEXTRA) != 0 &&
  70             BamTools::UnpackUnsignedShort(&header[10]) == Constants::BGZF_XLEN &&
  71             header[12] == Constants::BGZF_ID1 &&
  72             header[13] == Constants::BGZF_ID2 &&
  73             BamTools::UnpackUnsignedShort(&header[14]) == Constants::BGZF_LEN );
  74 }
  75
  76 // closes BGZF file
  77 void BgzfStream::Close(void) {
  78
  79     // reset state
  80     m_blockLength = 0;
  81     m_blockOffset = 0;
  82     m_blockAddress = 0;
  83     m_isWriteCompressed = true;
  84
  85     // skip if no device open
  86     if ( m_device == 0 ) return;
  87
  88     // if writing to file, flush the current BGZF block,
  89     // then write an empty block (as EOF marker)
  90     if ( m_device->IsOpen() && (m_device->Mode() == IBamIODevice::WriteOnly) ) {
  91         FlushBlock();
  92         const size_t blockLength = DeflateBlock();
  93         m_device->Write(Resources.CompressedBlock, blockLength);
  94     }
  95
  96     // close device
  97     m_device->Close();
  98     delete m_device;
  99     m_device = 0;
 100 }
 101
 102 // compresses the current block
 103 size_t BgzfStream::DeflateBlock(void) {
 104
 105     // initialize the gzip header
 106     char* buffer = Resources.CompressedBlock;
 107     memset(buffer, 0, 18);
 108     buffer[0]  = Constants::GZIP_ID1;
 109     buffer[1]  = Constants::GZIP_ID2;
 110     buffer[2]  = Constants::CM_DEFLATE;
 111     buffer[3]  = Constants::FLG_FEXTRA;
 112     buffer[9]  = Constants::OS_UNKNOWN;
 113     buffer[10] = Constants::BGZF_XLEN;
 114     buffer[12] = Constants::BGZF_ID1;
 115     buffer[13] = Constants::BGZF_ID2;
 116     buffer[14] = Constants::BGZF_LEN;
 117
 118     // set compression level
 119     const int compressionLevel = ( m_isWriteCompressed ? Z_DEFAULT_COMPRESSION : 0 );
 120
 121     // loop to retry for blocks that do not compress enough
 122     int inputLength = m_blockOffset;
 123     size_t compressedLength = 0;
 124     const unsigned int bufferSize = Constants::BGZF_MAX_BLOCK_SIZE;
 125
 126     while ( true ) {
 127
 128         // initialize zstream values
 129         z_stream zs;
 130         zs.zalloc    = NULL;
 131         zs.zfree     = NULL;
 132         zs.next_in   = (Bytef*)Resources.UncompressedBlock;
 133         zs.avail_in  = inputLength;
 134         zs.next_out  = (Bytef*)&buffer[Constants::BGZF_BLOCK_HEADER_LENGTH];
 135         zs.avail_out = bufferSize -
 136                        Constants::BGZF_BLOCK_HEADER_LENGTH -
 137                        Constants::BGZF_BLOCK_FOOTER_LENGTH;
 138
 139         // initialize the zlib compression algorithm
 140         int status = deflateInit2(&zs,
 141                                   compressionLevel,
 142                                   Z_DEFLATED,
 143                                   Constants::GZIP_WINDOW_BITS,
 144                                   Constants::Z_DEFAULT_MEM_LEVEL,
 145                                   Z_DEFAULT_STRATEGY);
 146         if ( status != Z_OK )
 147             throw BamException("BgzfStream::DeflateBlock", "zlib deflateInit2 failed");
 148
 149         // compress the data
 150         status = deflate(&zs, Z_FINISH);
 151
 152         // if not at stream end
 153         if ( status != Z_STREAM_END ) {
 154
 155             deflateEnd(&zs);
 156
 157             // if error status
 158             if ( status != Z_OK )
 159                 throw BamException("BgzfStream::DeflateBlock", "zlib deflate failed");
 160
 161             // not enough space available in buffer
 162             // try to reduce the input length & re-start loop
 163             inputLength -= 1024;
 164             if ( inputLength <= 0 )
 165                 throw BamException("BgzfStream::DeflateBlock", "input reduction failed");
 166             continue;
 167         }
 168
 169         // finalize the compression routine
 170         status = deflateEnd(&zs);
 171         if ( status != Z_OK )
 172             throw BamException("BgzfStream::DeflateBlock", "zlib deflateEnd failed");
 173
 174         // update compressedLength
 175         compressedLength = zs.total_out +
 176                            Constants::BGZF_BLOCK_HEADER_LENGTH +
 177                            Constants::BGZF_BLOCK_FOOTER_LENGTH;
 178         if ( compressedLength > Constants::BGZF_MAX_BLOCK_SIZE )
 179             throw BamException("BgzfStream::DeflateBlock", "deflate overflow");
 180
 181         // quit while loop
 182         break;
 183     }
 184
 185     // store the compressed length
 186     BamTools::PackUnsignedShort(&buffer[16], static_cast<uint16_t>(compressedLength - 1));
 187
 188     // store the CRC32 checksum
 189     uint32_t crc = crc32(0, NULL, 0);
 190     crc = crc32(crc, (Bytef*)Resources.UncompressedBlock, inputLength);
 191     BamTools::PackUnsignedInt(&buffer[compressedLength - 8], crc);
 192     BamTools::PackUnsignedInt(&buffer[compressedLength - 4], inputLength);
 193
 194     // ensure that we have less than a block of data left
 195     int remaining = m_blockOffset - inputLength;
 196     if ( remaining > 0 ) {
 197         if ( remaining > inputLength )
 198             throw BamException("BgzfStream::DeflateBlock", "after deflate, remainder too large");
 199         memcpy(Resources.UncompressedBlock, Resources.UncompressedBlock + inputLength, remaining);
 200     }
 201
 202     // update block data
 203     m_blockOffset = remaining;
 204
 205     // return result
 206     return compressedLength;
 207 }
 208
 209 // flushes the data in the BGZF block
 210 void BgzfStream::FlushBlock(void) {
 211
 212     BT_ASSERT_X( m_device, "BgzfStream::FlushBlock() - attempting to flush to null device" );
 213
 214     // flush all of the remaining blocks
 215     while ( m_blockOffset > 0 ) {
 216
 217         // compress the data block
 218         const size_t blockLength = DeflateBlock();
 219
 220         // flush the data to our output device
 221         const size_t numBytesWritten = m_device->Write(Resources.CompressedBlock, blockLength);
 222         if ( numBytesWritten != blockLength ) {
 223             stringstream s("");
 224             s << "expected to write " << blockLength
 225               << " bytes during flushing, but wrote " << numBytesWritten;
 226             throw BamException("BgzfStream::FlushBlock", s.str());
 227         }
 228
 229         // update block data
 230         m_blockAddress += blockLength;
 231     }
 232 }
 233
 234 // decompresses the current block
 235 size_t BgzfStream::InflateBlock(const size_t& blockLength) {
 236
 237     // setup zlib stream object
 238     z_stream zs;
 239     zs.zalloc    = NULL;
 240     zs.zfree     = NULL;
 241     zs.next_in   = (Bytef*)Resources.CompressedBlock + 18;
 242     zs.avail_in  = blockLength - 16;
 243     zs.next_out  = (Bytef*)Resources.UncompressedBlock;
 244     zs.avail_out = Constants::BGZF_DEFAULT_BLOCK_SIZE;
 245
 246     // initialize
 247     int status = inflateInit2(&zs, Constants::GZIP_WINDOW_BITS);
 248     if ( status != Z_OK )
 249         throw BamException("BgzfStream::InflateBlock", "zlib inflateInit failed");
 250
 251     // decompress
 252     status = inflate(&zs, Z_FINISH);
 253     if ( status != Z_STREAM_END ) {
 254         inflateEnd(&zs);
 255         throw BamException("BgzfStream::InflateBlock", "zlib inflate failed");
 256     }
 257
 258     // finalize
 259     status = inflateEnd(&zs);
 260     if ( status != Z_OK ) {
 261         inflateEnd(&zs);
 262         throw BamException("BgzfStream::InflateBlock", "zlib inflateEnd failed");
 263     }
 264
 265     // return result
 266     return zs.total_out;
 267 }
 268
 269 bool BgzfStream::IsOpen(void) const {
 270     if ( m_device == 0 )
 271         return false;
 272     return m_device->IsOpen();
 273 }
 274
 275 void BgzfStream::Open(const string& filename, const IBamIODevice::OpenMode mode) {
 276
 277     // close current device if necessary
 278     Close();
 279     BT_ASSERT_X( (m_device == 0), "BgzfStream::Open() - unable to properly close previous IO device" );
 280
 281     // retrieve new IO device depending on filename
 282     m_device = BamDeviceFactory::CreateDevice(filename);
 283     BT_ASSERT_X( m_device, "BgzfStream::Open() - unable to create IO device from filename" );
 284
 285     // if device fails to open
 286     if ( !m_device->Open(mode) ) {
 287         const string deviceError = m_device->GetErrorString();
 288         const string message = string("could not open BGZF stream: \n\t") + deviceError;
 289         throw BamException("BgzfStream::Open", message);
 290     }
 291 }
 292
 293 // reads BGZF data into a byte buffer
 294 size_t BgzfStream::Read(char* data, const size_t dataLength) {
 295
 296     if ( dataLength == 0 )
 297         return 0;
 298
 299     // if stream not open for reading
 300     BT_ASSERT_X( m_device, "BgzfStream::Read() - trying to read from null device");
 301     if ( !m_device->IsOpen() || (m_device->Mode() != IBamIODevice::ReadOnly) )
 302         return 0;
 303
 304     // read blocks as needed until desired data length is retrieved
 305     size_t numBytesRead = 0;
 306     while ( numBytesRead < dataLength ) {
 307
 308         // determine bytes available in current block
 309         int bytesAvailable = m_blockLength - m_blockOffset;
 310
 311         // read (and decompress) next block if needed
 312         if ( bytesAvailable <= 0 ) {
 313             ReadBlock();
 314             bytesAvailable = m_blockLength - m_blockOffset;
 315             if ( bytesAvailable <= 0 )
 316                 break;
 317         }
 318
 319         // copy data from uncompressed source buffer into data destination buffer
 320         const size_t copyLength = min( (dataLength-numBytesRead), (size_t)bytesAvailable );
 321         memcpy(data, Resources.UncompressedBlock + m_blockOffset, copyLength);
 322
 323         // update counters
 324         m_blockOffset  += copyLength;
 325         data         += copyLength;
 326         numBytesRead += copyLength;
 327     }
 328
 329     // update block data
 330     if ( m_blockOffset == m_blockLength ) {
 331         m_blockAddress = m_device->Tell();
 332         m_blockOffset  = 0;
 333         m_blockLength  = 0;
 334
 335     }
 336
 337     // return actual number of bytes read
 338     return numBytesRead;
 339 }
 340
 341 // reads a BGZF block
 342 void BgzfStream::ReadBlock(void) {
 343
 344     BT_ASSERT_X( m_device, "BgzfStream::ReadBlock() - trying to read from null IO device");
 345
 346     // store block's starting address
 347     int64_t blockAddress = m_device->Tell();
 348
 349     // read block header from file
 350     char header[Constants::BGZF_BLOCK_HEADER_LENGTH];
 351     size_t numBytesRead = m_device->Read(header, Constants::BGZF_BLOCK_HEADER_LENGTH);
 352
 353     // if block header empty
 354     if ( numBytesRead == 0 ) {
 355         m_blockLength = 0;
 356         return;
 357     }
 358
 359     // if block header invalid size
 360     if ( numBytesRead != Constants::BGZF_BLOCK_HEADER_LENGTH )
 361         throw BamException("BgzfStream::ReadBlock", "invalid block header size");
 362
 363     // validate block header contents
 364     if ( !BgzfStream::CheckBlockHeader(header) )
 365         throw BamException("BgzfStream::ReadBlock", "invalid block header contents");
 366
 367     // copy header contents to compressed buffer
 368     const size_t blockLength = BamTools::UnpackUnsignedShort(&header[16]) + 1;
 369     memcpy(Resources.CompressedBlock, header, Constants::BGZF_BLOCK_HEADER_LENGTH);
 370
 371     // read remainder of block
 372     const size_t remaining = blockLength - Constants::BGZF_BLOCK_HEADER_LENGTH;
 373     numBytesRead = m_device->Read(&Resources.CompressedBlock[Constants::BGZF_BLOCK_HEADER_LENGTH], remaining);
 374     if ( numBytesRead != remaining )
 375         throw BamException("BgzfStream::ReadBlock", "could not read data from block");
 376
 377     // decompress block data
 378     numBytesRead = InflateBlock(blockLength);
 379
 380     // update block data
 381     if ( m_blockLength != 0 )
 382         m_blockOffset = 0;
 383     m_blockAddress = blockAddress;
 384     m_blockLength  = numBytesRead;
 385 }
 386
 387 // seek to position in BGZF file
 388 void BgzfStream::Seek(const int64_t& position) {
 389
 390     BT_ASSERT_X( m_device, "BgzfStream::Seek() - trying to seek on null IO device");
 391
 392     // skip if device is not open
 393     if ( !IsOpen() ) return;
 394
 395     // determine adjusted offset & address
 396     int     blockOffset  = (position & 0xFFFF);
 397     int64_t blockAddress = (position >> 16) & 0xFFFFFFFFFFFFLL;
 398
 399     // attempt seek in file
 400     if ( m_device->IsRandomAccess() && m_device->Seek(blockAddress) ) {
 401
 402         // update block data & return success
 403         m_blockLength  = 0;
 404         m_blockAddress = blockAddress;
 405         m_blockOffset  = blockOffset;
 406     }
 407     else {
 408         stringstream s("");
 409         s << "unable to seek to position: " << position;
 410         throw BamException("BgzfStream::Seek", s.str());
 411     }
 412 }
 413
 414 void BgzfStream::SetWriteCompressed(bool ok) {
 415     m_isWriteCompressed = ok;
 416 }
 417
 418 // get file position in BGZF file
 419 int64_t BgzfStream::Tell(void) const {
 420     if ( !IsOpen() )
 421         return 0;
 422     return ( (m_blockAddress << 16) | (m_blockOffset & 0xFFFF) );
 423 }
 424
 425 // writes the supplied data into the BGZF buffer
 426 size_t BgzfStream::Write(const char* data, const size_t dataLength) {
 427
 428     BT_ASSERT_X( m_device, "BgzfStream::Write() - trying to write to null IO device");
 429     BT_ASSERT_X( (m_device->Mode() == IBamIODevice::WriteOnly),
 430                  "BgzfStream::Write() - trying to write to non-writable IO device");
 431
 432     // skip if file not open for writing
 433     if ( !IsOpen() )
 434         return 0;
 435
 436     // write blocks as needed til all data is written
 437     size_t numBytesWritten = 0;
 438     const char* input = data;
 439     const size_t blockLength = Constants::BGZF_DEFAULT_BLOCK_SIZE;
 440     while ( numBytesWritten < dataLength ) {
 441
 442         // copy data contents to uncompressed output buffer
 443         unsigned int copyLength = min(blockLength - m_blockOffset, dataLength - numBytesWritten);
 444         char* buffer = Resources.UncompressedBlock;
 445         memcpy(buffer + m_blockOffset, input, copyLength);
 446
 447         // update counter
 448         m_blockOffset   += copyLength;
 449         input           += copyLength;
 450         numBytesWritten += copyLength;
 451
 452         // flush (& compress) output buffer when full
 453         if ( m_blockOffset == blockLength )
 454             FlushBlock();
 455     }
 456
 457     // return actual number of bytes written
 458     return numBytesWritten;
 459 }