3 * SOFTWARE COPYRIGHT NOTICE AGREEMENT
4 * This software and its documentation are copyright 2008 by the
5 * Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
7 * This software is supplied without any warranty or guaranteed support whatsoever.
8 * Neither the Broad Institute nor MIT can be responsible for its use, misuse,
17 #include <sys/types.h>
21 extern off_t ftello(FILE *stream);
22 extern int fseeko(FILE *stream, off_t offset, int whence);
26 static const int DEFAULT_BLOCK_SIZE = 64 * 1024;
27 static const int MAX_BLOCK_SIZE = 64 * 1024;
29 static const int BLOCK_HEADER_LENGTH = 18;
30 static const int BLOCK_FOOTER_LENGTH = 8;
32 static const int GZIP_ID1 = 31;
33 static const int GZIP_ID2 = 139;
34 static const int CM_DEFLATE = 8;
35 static const int FLG_FEXTRA = 4;
36 static const int OS_UNKNOWN = 255;
37 static const int BGZF_ID1 = 66; // 'B'
38 static const int BGZF_ID2 = 67; // 'C'
39 static const int BGZF_LEN = 2;
40 static const int BGZF_XLEN = 6; // BGZF_LEN+4
42 static const int GZIP_WINDOW_BITS = -15; // no zlib header
43 static const int Z_DEFAULT_MEM_LEVEL = 8;
48 packInt16(uint8_t* buffer, uint16_t value)
51 buffer[1] = value >> 8;
56 unpackInt16(const uint8_t* buffer)
58 return (buffer[0] | (buffer[1] << 8));
63 packInt32(uint8_t* buffer, uint32_t value)
66 buffer[1] = value >> 8;
67 buffer[2] = value >> 16;
68 buffer[3] = value >> 24;
75 return (x < y) ? x : y;
80 report_error(BGZF* fp, const char* message) {
88 FILE* file = fdopen(fd, "r");
89 BGZF* fp = malloc(sizeof(BGZF));
90 fp->file_descriptor = fd;
94 fp->uncompressed_block_size = MAX_BLOCK_SIZE;
95 fp->uncompressed_block = malloc(MAX_BLOCK_SIZE);
96 fp->compressed_block_size = MAX_BLOCK_SIZE;
97 fp->compressed_block = malloc(MAX_BLOCK_SIZE);
98 fp->block_address = 0;
100 fp->block_length = 0;
109 FILE* file = fdopen(fd, "w");
110 BGZF* fp = malloc(sizeof(BGZF));
111 fp->file_descriptor = fd;
115 fp->uncompressed_block_size = DEFAULT_BLOCK_SIZE;
116 fp->uncompressed_block = NULL;
117 fp->compressed_block_size = MAX_BLOCK_SIZE;
118 fp->compressed_block = malloc(MAX_BLOCK_SIZE);
119 fp->block_address = 0;
120 fp->block_offset = 0;
121 fp->block_length = 0;
127 bgzf_open(const char* __restrict path, const char* __restrict mode)
130 if (strcasecmp(mode, "r") == 0) {
131 int oflag = O_RDONLY;
132 int fd = open(path, oflag);
134 } else if (strcasecmp(mode, "w") == 0) {
135 int oflag = O_WRONLY | O_CREAT | O_TRUNC;
136 int fd = open(path, oflag, 0644);
146 bgzf_fdopen(int fd, const char * __restrict mode)
148 if (strcasecmp(mode, "r") == 0) {
149 return open_read(fd);
150 } else if (strcasecmp(mode, "w") == 0) {
151 return open_write(fd);
159 deflate_block(BGZF* fp, int block_length)
161 // Deflate the block in fp->uncompressed_block into fp->compressed_block.
162 // Also adds an extra field that stores the compressed block length.
164 byte* buffer = fp->compressed_block;
165 int buffer_size = fp->compressed_block_size;
168 buffer[0] = GZIP_ID1;
169 buffer[1] = GZIP_ID2;
170 buffer[2] = CM_DEFLATE;
171 buffer[3] = FLG_FEXTRA;
172 buffer[4] = 0; // mtime
177 buffer[9] = OS_UNKNOWN;
178 buffer[10] = BGZF_XLEN;
180 buffer[12] = BGZF_ID1;
181 buffer[13] = BGZF_ID2;
182 buffer[14] = BGZF_LEN;
184 buffer[16] = 0; // placeholder for block length
187 // loop to retry for blocks that do not compress enough
188 int input_length = block_length;
189 int compressed_length = 0;
195 zs.next_in = fp->uncompressed_block;
196 zs.avail_in = input_length;
197 zs.next_out = (void*)&buffer[BLOCK_HEADER_LENGTH];
198 zs.avail_out = buffer_size - BLOCK_HEADER_LENGTH - BLOCK_FOOTER_LENGTH;
200 int status = deflateInit2(&zs, Z_DEFAULT_COMPRESSION, Z_DEFLATED,
201 GZIP_WINDOW_BITS, Z_DEFAULT_MEM_LEVEL, Z_DEFAULT_STRATEGY);
202 if (status != Z_OK) {
203 report_error(fp, "deflate init failed");
206 status = deflate(&zs, Z_FINISH);
207 if (status != Z_STREAM_END) {
209 if (status == Z_OK) {
210 // Not enough space in buffer.
211 // Can happen in the rare case the input doesn't compress enough.
212 // Reduce the amount of input until it fits.
213 input_length -= 1024;
214 if (input_length <= 0) {
215 // should never happen
216 report_error(fp, "input reduction failed");
221 report_error(fp, "deflate failed");
224 status = deflateEnd(&zs);
225 if (status != Z_OK) {
226 report_error(fp, "deflate end failed");
229 compressed_length = zs.total_out;
230 compressed_length += BLOCK_HEADER_LENGTH + BLOCK_FOOTER_LENGTH;
231 if (compressed_length > MAX_BLOCK_SIZE) {
232 // should never happen
233 report_error(fp, "deflate overflow");
239 packInt16((uint8_t*)&buffer[16], compressed_length-1);
240 uint32_t crc = crc32(0L, NULL, 0L);
241 crc = crc32(crc, fp->uncompressed_block, input_length);
242 packInt32((uint8_t*)&buffer[compressed_length-8], crc);
243 packInt32((uint8_t*)&buffer[compressed_length-4], input_length);
245 int remaining = block_length - input_length;
247 if (remaining > input_length) {
248 // should never happen (check so we can use memcpy)
249 report_error(fp, "remainder too large");
252 memcpy(fp->uncompressed_block,
253 fp->uncompressed_block + input_length,
256 fp->block_offset = remaining;
257 return compressed_length;
262 inflate_block(BGZF* fp, int block_length)
264 // Inflate the block in fp->compressed_block into fp->uncompressed_block
269 zs.next_in = fp->compressed_block + 18;
270 zs.avail_in = block_length - 16;
271 zs.next_out = fp->uncompressed_block;
272 zs.avail_out = fp->uncompressed_block_size;
274 int status = inflateInit2(&zs, GZIP_WINDOW_BITS);
275 if (status != Z_OK) {
276 report_error(fp, "inflate init failed");
279 status = inflate(&zs, Z_FINISH);
280 if (status != Z_STREAM_END) {
282 report_error(fp, "inflate failed");
285 status = inflateEnd(&zs);
286 if (status != Z_OK) {
287 report_error(fp, "inflate failed");
295 check_header(const byte* header)
297 return (header[0] == GZIP_ID1 &&
298 header[1] == (byte) GZIP_ID2 &&
299 header[2] == Z_DEFLATED &&
300 (header[3] & FLG_FEXTRA) != 0 &&
301 unpackInt16((uint8_t*)&header[10]) == BGZF_XLEN &&
302 header[12] == BGZF_ID1 &&
303 header[13] == BGZF_ID2 &&
304 unpackInt16((uint8_t*)&header[14]) == BGZF_LEN);
311 byte header[BLOCK_HEADER_LENGTH];
312 int64_t block_address = ftello(fp->file);
313 int count = fread(header, 1, sizeof(header), fp->file);
315 fp->block_length = 0;
318 if (count != sizeof(header)) {
319 report_error(fp, "read failed");
322 if (!check_header(header)) {
323 report_error(fp, "invalid block header");
326 int block_length = unpackInt16((uint8_t*)&header[16]) + 1;
327 byte* compressed_block = (byte*) fp->compressed_block;
328 memcpy(compressed_block, header, BLOCK_HEADER_LENGTH);
329 int remaining = block_length - BLOCK_HEADER_LENGTH;
330 count = fread(&compressed_block[BLOCK_HEADER_LENGTH], 1, remaining, fp->file);
331 if (count != remaining) {
332 report_error(fp, "read failed");
335 count = inflate_block(fp, block_length);
339 if (fp->block_length != 0) {
340 // Do not reset offset if this read follows a seek.
341 fp->block_offset = 0;
343 fp->block_address = block_address;
344 fp->block_length = count;
349 bgzf_read(BGZF* fp, void* data, int length)
354 if (fp->open_mode != 'r') {
355 report_error(fp, "file not open for reading");
361 while (bytes_read < length) {
362 int available = fp->block_length - fp->block_offset;
363 if (available <= 0) {
364 if (read_block(fp) != 0) {
367 available = fp->block_length - fp->block_offset;
368 if (available <= 0) {
372 int copy_length = min(length-bytes_read, available);
373 byte* buffer = fp->uncompressed_block;
374 memcpy(output, buffer + fp->block_offset, copy_length);
375 fp->block_offset += copy_length;
376 output += copy_length;
377 bytes_read += copy_length;
379 if (fp->block_offset == fp->block_length) {
380 fp->block_address = ftello(fp->file);
381 fp->block_offset = 0;
382 fp->block_length = 0;
389 flush_block(BGZF* fp)
391 while (fp->block_offset > 0) {
392 int block_length = deflate_block(fp, fp->block_offset);
393 if (block_length < 0) {
396 int count = fwrite(fp->compressed_block, 1, block_length, fp->file);
397 if (count != block_length) {
398 report_error(fp, "write failed");
401 fp->block_address += block_length;
407 bgzf_write(BGZF* fp, const void* data, int length)
409 if (fp->open_mode != 'w') {
410 report_error(fp, "file not open for writing");
414 if (fp->uncompressed_block == NULL) {
415 fp->uncompressed_block = malloc(fp->uncompressed_block_size);
418 const byte* input = data;
419 int block_length = fp->uncompressed_block_size;
420 int bytes_written = 0;
421 while (bytes_written < length) {
422 int copy_length = min(block_length - fp->block_offset, length - bytes_written);
423 byte* buffer = fp->uncompressed_block;
424 memcpy(buffer + fp->block_offset, input, copy_length);
425 fp->block_offset += copy_length;
426 input += copy_length;
427 bytes_written += copy_length;
428 if (fp->block_offset == block_length) {
429 if (flush_block(fp) != 0) {
434 return bytes_written;
440 if (fp->open_mode == 'w') {
441 if (flush_block(fp) != 0) {
444 if (fflush(fp->file) != 0) {
445 report_error(fp, "flush failed");
449 if (fp->owned_file) {
450 if (fclose(fp->file) != 0) {
454 free(fp->uncompressed_block);
455 free(fp->compressed_block);
463 return ((fp->block_address << 16) | (fp->block_offset & 0xFFFF));
467 bgzf_seek(BGZF* fp, int64_t pos, int where)
469 if (fp->open_mode != 'r') {
470 report_error(fp, "file not open for read");
473 if (where != SEEK_SET) {
474 report_error(fp, "unimplemented seek option");
477 int block_offset = pos & 0xFFFF;
478 int64_t block_address = (pos >> 16) & 0xFFFFFFFFFFFFLL;
479 if (fseeko(fp->file, block_address, SEEK_SET) != 0) {
480 report_error(fp, "seek failed");
483 fp->block_length = 0; // indicates current block is not loaded
484 fp->block_address = block_address;
485 fp->block_offset = block_offset;