forked from samtools/htslib
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbgzf.c
1943 lines (1709 loc) · 57.5 KB
/
bgzf.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* The MIT License
Copyright (c) 2008 Broad Institute / Massachusetts Institute of Technology
2011, 2012 Attractive Chaos <[email protected]>
Copyright (C) 2009, 2013-2017 Genome Research Ltd
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <unistd.h>
#include <assert.h>
#include <pthread.h>
#include <sys/types.h>
#include <inttypes.h>
#include "htslib/hts.h"
#include "htslib/bgzf.h"
#include "htslib/hfile.h"
#include "htslib/thread_pool.h"
#include "htslib/hts_endian.h"
#include "cram/pooled_alloc.h"
#define BGZF_CACHE
#define BGZF_MT
#define BLOCK_HEADER_LENGTH 18
#define BLOCK_FOOTER_LENGTH 8
/* BGZF/GZIP header (speciallized from RFC 1952; little endian):
+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
| 31|139| 8| 4| 0| 0|255| 6| 66| 67| 2|BLK_LEN|
+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
BGZF extension:
^ ^ ^ ^
| | | |
FLG.EXTRA XLEN B C
BGZF format is compatible with GZIP. It limits the size of each compressed
block to 2^16 bytes and adds and an extra "BC" field in the gzip header which
records the size.
*/
static const uint8_t g_magic[19] = "\037\213\010\4\0\0\0\0\0\377\6\0\102\103\2\0\0\0";
#ifdef BGZF_CACHE
typedef struct {
int size;
uint8_t *block;
int64_t end_offset;
} cache_t;
#include "htslib/khash.h"
KHASH_MAP_INIT_INT64(cache, cache_t)
#endif
struct bgzf_cache_t {
khash_t(cache) *h;
khint_t last_pos;
};
#ifdef BGZF_MT
typedef struct bgzf_job {
BGZF *fp;
unsigned char comp_data[BGZF_MAX_BLOCK_SIZE];
size_t comp_len;
unsigned char uncomp_data[BGZF_MAX_BLOCK_SIZE];
size_t uncomp_len;
int errcode;
int64_t block_address;
int hit_eof;
} bgzf_job;
enum mtaux_cmd {
NONE = 0,
SEEK,
HAS_EOF,
CLOSE,
};
typedef struct bgzf_mtaux_t {
// Memory pool for bgzf_job structs, to avoid many malloc/free
pool_alloc_t *job_pool;
bgzf_job *curr_job;
// Thread pool
int n_threads;
int own_pool;
hts_tpool *pool;
// Output queue holding completed bgzf_jobs
hts_tpool_process *out_queue;
// I/O thread.
pthread_t io_task;
pthread_mutex_t job_pool_m;
int jobs_pending; // number of jobs waiting
int flush_pending;
void *free_block;
int hit_eof; // r/w entirely within main thread
// Message passing to the reader thread; eg seek requests
int errcode;
uint64_t block_address;
int eof;
pthread_mutex_t command_m; // Set whenever fp is being updated
pthread_cond_t command_c;
enum mtaux_cmd command;
} mtaux_t;
#endif
typedef struct
{
uint64_t uaddr; // offset w.r.t. uncompressed data
uint64_t caddr; // offset w.r.t. compressed data
}
bgzidx1_t;
struct __bgzidx_t
{
int noffs, moffs; // the size of the index, n:used, m:allocated
bgzidx1_t *offs; // offsets
uint64_t ublock_addr; // offset of the current block (uncompressed data)
};
void bgzf_index_destroy(BGZF *fp);
int bgzf_index_add_block(BGZF *fp);
static void mt_destroy(mtaux_t *mt);
static inline void packInt16(uint8_t *buffer, uint16_t value)
{
buffer[0] = value;
buffer[1] = value >> 8;
}
static inline int unpackInt16(const uint8_t *buffer)
{
return buffer[0] | buffer[1] << 8;
}
static inline void packInt32(uint8_t *buffer, uint32_t value)
{
buffer[0] = value;
buffer[1] = value >> 8;
buffer[2] = value >> 16;
buffer[3] = value >> 24;
}
static const char *bgzf_zerr(int errnum, z_stream *zs)
{
static char buffer[32];
/* Return zs->msg if available.
zlib doesn't set this very reliably. Looking at the source suggests
that it may get set to a useful message for deflateInit2, inflateInit2
and inflate when it returns Z_DATA_ERROR. For inflate with other
return codes, deflate, deflateEnd and inflateEnd it doesn't appear
to be useful. For the likely non-useful cases, the caller should
pass NULL into zs. */
if (zs && zs->msg) return zs->msg;
// gzerror OF((gzFile file, int *errnum)
switch (errnum) {
case Z_ERRNO:
return strerror(errno);
case Z_STREAM_ERROR:
return "invalid parameter/compression level, or inconsistent stream state";
case Z_DATA_ERROR:
return "invalid or incomplete IO";
case Z_MEM_ERROR:
return "out of memory";
case Z_BUF_ERROR:
return "progress temporarily not possible, or in() / out() returned an error";
case Z_VERSION_ERROR:
return "zlib version mismatch";
case Z_OK: // 0: maybe gzgets error Z_NULL
default:
snprintf(buffer, sizeof(buffer), "[%d] unknown", errnum);
return buffer; // FIXME: Not thread-safe.
}
}
static BGZF *bgzf_read_init(hFILE *hfpr)
{
BGZF *fp;
uint8_t magic[18];
ssize_t n = hpeek(hfpr, magic, 18);
if (n < 0) return NULL;
fp = (BGZF*)calloc(1, sizeof(BGZF));
if (fp == NULL) return NULL;
fp->is_write = 0;
fp->uncompressed_block = malloc(2 * BGZF_MAX_BLOCK_SIZE);
if (fp->uncompressed_block == NULL) { free(fp); return NULL; }
fp->compressed_block = (char *)fp->uncompressed_block + BGZF_MAX_BLOCK_SIZE;
fp->is_compressed = (n==18 && magic[0]==0x1f && magic[1]==0x8b);
fp->is_gzip = ( !fp->is_compressed || ((magic[3]&4) && memcmp(&magic[12], "BC\2\0",4)==0) ) ? 0 : 1;
#ifdef BGZF_CACHE
if (!(fp->cache = malloc(sizeof(*fp->cache)))) {
free(fp);
return NULL;
}
if (!(fp->cache->h = kh_init(cache))) {
free(fp->cache);
free(fp);
return NULL;
}
fp->cache->last_pos = 0;
#endif
return fp;
}
// get the compress level from the mode string: compress_level==-1 for the default level, -2 plain uncompressed
static int mode2level(const char *mode)
{
int i, compress_level = -1;
for (i = 0; mode[i]; ++i)
if (mode[i] >= '0' && mode[i] <= '9') break;
if (mode[i]) compress_level = (int)mode[i] - '0';
if (strchr(mode, 'u')) compress_level = -2;
return compress_level;
}
static BGZF *bgzf_write_init(const char *mode)
{
BGZF *fp;
fp = (BGZF*)calloc(1, sizeof(BGZF));
if (fp == NULL) goto mem_fail;
fp->is_write = 1;
int compress_level = mode2level(mode);
if ( compress_level==-2 )
{
fp->is_compressed = 0;
return fp;
}
fp->is_compressed = 1;
fp->uncompressed_block = malloc(2 * BGZF_MAX_BLOCK_SIZE);
if (fp->uncompressed_block == NULL) goto mem_fail;
fp->compressed_block = (char *)fp->uncompressed_block + BGZF_MAX_BLOCK_SIZE;
fp->compress_level = compress_level < 0? Z_DEFAULT_COMPRESSION : compress_level; // Z_DEFAULT_COMPRESSION==-1
if (fp->compress_level > 9) fp->compress_level = Z_DEFAULT_COMPRESSION;
if ( strchr(mode,'g') )
{
// gzip output
fp->is_gzip = 1;
fp->gz_stream = (z_stream*)calloc(1,sizeof(z_stream));
if (fp->gz_stream == NULL) goto mem_fail;
fp->gz_stream->zalloc = NULL;
fp->gz_stream->zfree = NULL;
fp->gz_stream->msg = NULL;
int ret = deflateInit2(fp->gz_stream, fp->compress_level, Z_DEFLATED, 15|16, 8, Z_DEFAULT_STRATEGY);
if (ret!=Z_OK) {
hts_log_error("Call to deflateInit2 failed: %s", bgzf_zerr(ret, fp->gz_stream));
goto fail;
}
}
return fp;
mem_fail:
hts_log_error("%s", strerror(errno));
fail:
if (fp != NULL) {
free(fp->uncompressed_block);
free(fp->gz_stream);
free(fp);
}
return NULL;
}
BGZF *bgzf_open(const char *path, const char *mode)
{
BGZF *fp = 0;
assert(compressBound(BGZF_BLOCK_SIZE) < BGZF_MAX_BLOCK_SIZE);
if (strchr(mode, 'r')) {
hFILE *fpr;
if ((fpr = hopen(path, mode)) == 0) return 0;
fp = bgzf_read_init(fpr);
if (fp == 0) { hclose_abruptly(fpr); return NULL; }
fp->fp = fpr;
} else if (strchr(mode, 'w') || strchr(mode, 'a')) {
hFILE *fpw;
if ((fpw = hopen(path, mode)) == 0) return 0;
fp = bgzf_write_init(mode);
if (fp == NULL) return NULL;
fp->fp = fpw;
}
else { errno = EINVAL; return 0; }
fp->is_be = ed_is_big();
return fp;
}
BGZF *bgzf_dopen(int fd, const char *mode)
{
BGZF *fp = 0;
assert(compressBound(BGZF_BLOCK_SIZE) < BGZF_MAX_BLOCK_SIZE);
if (strchr(mode, 'r')) {
hFILE *fpr;
if ((fpr = hdopen(fd, mode)) == 0) return 0;
fp = bgzf_read_init(fpr);
if (fp == 0) { hclose_abruptly(fpr); return NULL; } // FIXME this closes fd
fp->fp = fpr;
} else if (strchr(mode, 'w') || strchr(mode, 'a')) {
hFILE *fpw;
if ((fpw = hdopen(fd, mode)) == 0) return 0;
fp = bgzf_write_init(mode);
if (fp == NULL) return NULL;
fp->fp = fpw;
}
else { errno = EINVAL; return 0; }
fp->is_be = ed_is_big();
return fp;
}
BGZF *bgzf_hopen(hFILE *hfp, const char *mode)
{
BGZF *fp = NULL;
assert(compressBound(BGZF_BLOCK_SIZE) < BGZF_MAX_BLOCK_SIZE);
if (strchr(mode, 'r')) {
fp = bgzf_read_init(hfp);
if (fp == NULL) return NULL;
} else if (strchr(mode, 'w') || strchr(mode, 'a')) {
fp = bgzf_write_init(mode);
if (fp == NULL) return NULL;
}
else { errno = EINVAL; return 0; }
fp->fp = hfp;
fp->is_be = ed_is_big();
return fp;
}
int bgzf_compress(void *_dst, size_t *dlen, const void *src, size_t slen, int level)
{
uint32_t crc;
z_stream zs;
uint8_t *dst = (uint8_t*)_dst;
// compress the body
zs.zalloc = NULL; zs.zfree = NULL;
zs.msg = NULL;
zs.next_in = (Bytef*)src;
zs.avail_in = slen;
zs.next_out = dst + BLOCK_HEADER_LENGTH;
zs.avail_out = *dlen - BLOCK_HEADER_LENGTH - BLOCK_FOOTER_LENGTH;
int ret = deflateInit2(&zs, level, Z_DEFLATED, -15, 8, Z_DEFAULT_STRATEGY); // -15 to disable zlib header/footer
if (ret!=Z_OK) {
hts_log_error("Call to deflateInit2 failed: %s", bgzf_zerr(ret, &zs));
return -1;
}
if ((ret = deflate(&zs, Z_FINISH)) != Z_STREAM_END) {
hts_log_error("Deflate operation failed: %s", bgzf_zerr(ret, ret == Z_DATA_ERROR ? &zs : NULL));
return -1;
}
if ((ret = deflateEnd(&zs)) != Z_OK) {
hts_log_error("Call to deflateEnd failed: %s", bgzf_zerr(ret, NULL));
return -1;
}
*dlen = zs.total_out + BLOCK_HEADER_LENGTH + BLOCK_FOOTER_LENGTH;
// write the header
memcpy(dst, g_magic, BLOCK_HEADER_LENGTH); // the last two bytes are a place holder for the length of the block
packInt16(&dst[16], *dlen - 1); // write the compressed length; -1 to fit 2 bytes
// write the footer
crc = crc32(crc32(0L, NULL, 0L), (Bytef*)src, slen);
packInt32((uint8_t*)&dst[*dlen - 8], crc);
packInt32((uint8_t*)&dst[*dlen - 4], slen);
return 0;
}
static int bgzf_gzip_compress(BGZF *fp, void *_dst, size_t *dlen, const void *src, size_t slen, int level)
{
uint8_t *dst = (uint8_t*)_dst;
z_stream *zs = fp->gz_stream;
int flush = slen ? Z_PARTIAL_FLUSH : Z_FINISH;
zs->next_in = (Bytef*)src;
zs->avail_in = slen;
zs->next_out = dst;
zs->avail_out = *dlen;
int ret = deflate(zs, flush);
if (ret == Z_STREAM_ERROR) {
hts_log_error("Deflate operation failed: %s", bgzf_zerr(ret, NULL));
return -1;
}
if (zs->avail_in != 0) {
hts_log_error("Deflate block too large for output buffer");
return -1;
}
*dlen = *dlen - zs->avail_out;
return 0;
}
// Deflate the block in fp->uncompressed_block into fp->compressed_block. Also adds an extra field that stores the compressed block length.
static int deflate_block(BGZF *fp, int block_length)
{
size_t comp_size = BGZF_MAX_BLOCK_SIZE;
int ret;
if ( !fp->is_gzip )
ret = bgzf_compress(fp->compressed_block, &comp_size, fp->uncompressed_block, block_length, fp->compress_level);
else
ret = bgzf_gzip_compress(fp, fp->compressed_block, &comp_size, fp->uncompressed_block, block_length, fp->compress_level);
if ( ret != 0 )
{
hts_log_debug("Compression error %d", ret);
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
fp->block_offset = 0;
return comp_size;
}
static int bgzf_uncompress(uint8_t *dst, size_t *dlen, const uint8_t *src, size_t slen) {
z_stream zs;
zs.zalloc = NULL;
zs.zfree = NULL;
zs.msg = NULL;
zs.next_in = (Bytef*)src;
zs.avail_in = slen;
zs.next_out = (Bytef*)dst;
zs.avail_out = *dlen;
int ret = inflateInit2(&zs, -15);
if (ret != Z_OK) {
hts_log_error("Call to inflateInit2 failed: %s", bgzf_zerr(ret, &zs));
return -1;
}
if ((ret = inflate(&zs, Z_FINISH)) != Z_STREAM_END) {
hts_log_error("Inflate operation failed: %s", bgzf_zerr(ret, ret == Z_DATA_ERROR ? &zs : NULL));
if ((ret = inflateEnd(&zs)) != Z_OK) {
hts_log_warning("Call to inflateEnd failed: %s", bgzf_zerr(ret, NULL));
}
return -1;
}
if ((ret = inflateEnd(&zs)) != Z_OK) {
hts_log_error("Call to inflateEnd failed: %s", bgzf_zerr(ret, NULL));
return -1;
}
*dlen = *dlen - zs.avail_out;
return 0;
}
// Inflate the block in fp->compressed_block into fp->uncompressed_block
static int inflate_block(BGZF* fp, int block_length)
{
size_t dlen = BGZF_MAX_BLOCK_SIZE;
int ret = bgzf_uncompress(fp->uncompressed_block, &dlen,
(Bytef*)fp->compressed_block + 18, block_length - 18);
if (ret < 0) {
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
// Check CRC of uncompressed block matches the gzip header.
// NB: we may wish to switch out the zlib crc32 for something more performant.
// See PR#361 and issue#467
uint32_t c1 = crc32(0L, (unsigned char *)fp->uncompressed_block, dlen);
uint32_t c2 = le_to_u32((uint8_t *)fp->compressed_block + block_length-8);
if (c1 != c2) {
fp->errcode |= BGZF_ERR_CRC;
return -1;
}
return dlen;
}
static int inflate_gzip_block(BGZF *fp, int cached)
{
int ret = Z_OK;
do
{
if ( !cached && fp->gz_stream->avail_out!=0 )
{
fp->gz_stream->avail_in = hread(fp->fp, fp->compressed_block, BGZF_BLOCK_SIZE);
if ( fp->gz_stream->avail_in<=0 ) return fp->gz_stream->avail_in;
if ( fp->gz_stream->avail_in==0 ) break;
fp->gz_stream->next_in = fp->compressed_block;
}
else cached = 0;
do
{
fp->gz_stream->next_out = (Bytef*)fp->uncompressed_block + fp->block_offset;
fp->gz_stream->avail_out = BGZF_MAX_BLOCK_SIZE - fp->block_offset;
fp->gz_stream->msg = NULL;
ret = inflate(fp->gz_stream, Z_NO_FLUSH);
if (ret < 0 && ret != Z_BUF_ERROR) {
hts_log_error("Inflate operation failed: %s", bgzf_zerr(ret, ret == Z_DATA_ERROR ? fp->gz_stream : NULL));
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
unsigned int have = BGZF_MAX_BLOCK_SIZE - fp->gz_stream->avail_out;
if ( have ) return have;
}
while ( fp->gz_stream->avail_out == 0 );
}
while (ret != Z_STREAM_END);
return BGZF_MAX_BLOCK_SIZE - fp->gz_stream->avail_out;
}
// Returns: 0 on success (BGZF header); -1 on non-BGZF GZIP header; -2 on error
static int check_header(const uint8_t *header)
{
if ( header[0] != 31 || header[1] != 139 || header[2] != 8 ) return -2;
return ((header[3] & 4) != 0
&& unpackInt16((uint8_t*)&header[10]) == 6
&& header[12] == 'B' && header[13] == 'C'
&& unpackInt16((uint8_t*)&header[14]) == 2) ? 0 : -1;
}
#ifdef BGZF_CACHE
static void free_cache(BGZF *fp)
{
khint_t k;
if (fp->is_write) return;
khash_t(cache) *h = fp->cache->h;
for (k = kh_begin(h); k < kh_end(h); ++k)
if (kh_exist(h, k)) free(kh_val(h, k).block);
kh_destroy(cache, h);
free(fp->cache);
}
static int load_block_from_cache(BGZF *fp, int64_t block_address)
{
khint_t k;
cache_t *p;
khash_t(cache) *h = fp->cache->h;
k = kh_get(cache, h, block_address);
if (k == kh_end(h)) return 0;
p = &kh_val(h, k);
if (fp->block_length != 0) fp->block_offset = 0;
fp->block_address = block_address;
fp->block_length = p->size;
memcpy(fp->uncompressed_block, p->block, p->size);
if ( hseek(fp->fp, p->end_offset, SEEK_SET) < 0 )
{
// todo: move the error up
hts_log_error("Could not hseek to %"PRId64"", p->end_offset);
exit(1);
}
return p->size;
}
static void cache_block(BGZF *fp, int size)
{
int ret;
khint_t k, k_orig;
uint8_t *block = NULL;
cache_t *p;
//fprintf(stderr, "Cache block at %llx\n", (int)fp->block_address);
khash_t(cache) *h = fp->cache->h;
if (BGZF_MAX_BLOCK_SIZE >= fp->cache_size) return;
if (fp->block_length < 0 || fp->block_length > BGZF_MAX_BLOCK_SIZE) return;
if ((kh_size(h) + 1) * BGZF_MAX_BLOCK_SIZE > (uint32_t)fp->cache_size) {
/* Remove uniformly from any position in the hash by a simple
* round-robin approach. An alternative strategy would be to
* remove the least recently accessed block, but the round-robin
* removal is simpler and is not expected to have a big impact
* on performance */
if (fp->cache->last_pos >= kh_end(h)) fp->cache->last_pos = kh_begin(h);
k_orig = k = fp->cache->last_pos;
if (++k >= kh_end(h)) k = kh_begin(h);
while (k != k_orig) {
if (kh_exist(h, k))
break;
if (++k == kh_end(h))
k = kh_begin(h);
}
fp->cache->last_pos = k;
if (k != k_orig) {
block = kh_val(h, k).block;
kh_del(cache, h, k);
}
} else {
block = (uint8_t*)malloc(BGZF_MAX_BLOCK_SIZE);
}
if (!block) return;
k = kh_put(cache, h, fp->block_address, &ret);
if (ret <= 0) { // kh_put failed, or in there already (shouldn't happen)
free(block);
return;
}
p = &kh_val(h, k);
p->size = fp->block_length;
p->end_offset = fp->block_address + size;
p->block = block;
memcpy(p->block, fp->uncompressed_block, p->size);
}
#else
static void free_cache(BGZF *fp) {}
static int load_block_from_cache(BGZF *fp, int64_t block_address) {return 0;}
static void cache_block(BGZF *fp, int size) {}
#endif
/*
* Absolute htell in this compressed file.
*
* Do not confuse with the external bgzf_tell macro which returns the virtual
* offset.
*/
static off_t bgzf_htell(BGZF *fp) {
if (fp->mt) {
pthread_mutex_lock(&fp->mt->job_pool_m);
off_t pos = fp->block_address + fp->block_clength;
pthread_mutex_unlock(&fp->mt->job_pool_m);
return pos;
} else {
return htell(fp->fp);
}
}
int bgzf_read_block(BGZF *fp)
{
hts_tpool_result *r;
if (fp->mt) {
again:
if (fp->mt->hit_eof) {
// Further reading at EOF will always return 0
fp->block_length = 0;
return 0;
}
r = hts_tpool_next_result_wait(fp->mt->out_queue);
bgzf_job *j = r ? (bgzf_job *)hts_tpool_result_data(r) : NULL;
if (!j || j->errcode == BGZF_ERR_MT) {
if (!fp->mt->free_block) {
fp->uncompressed_block = malloc(2 * BGZF_MAX_BLOCK_SIZE);
if (fp->uncompressed_block == NULL) return -1;
fp->compressed_block = (char *)fp->uncompressed_block + BGZF_MAX_BLOCK_SIZE;
} // else it's already allocated with malloc, maybe even in-use.
mt_destroy(fp->mt);
fp->mt = NULL;
hts_tpool_delete_result(r, 0);
goto single_threaded;
}
if (j->errcode) {
fp->errcode = j->errcode;
return -1;
}
if (j->hit_eof) {
if (!fp->last_block_eof && !fp->no_eof_block) {
fp->no_eof_block = 1;
hts_log_warning("EOF marker is absent. The input is probably truncated");
}
fp->mt->hit_eof = 1;
}
// Zero length blocks in the middle of a file are (wrongly)
// considered as EOF by many callers. We work around this by
// trying again to see if we hit a genuine EOF.
if (!j->hit_eof && j->uncomp_len == 0) {
fp->last_block_eof = 1;
hts_tpool_delete_result(r, 0);
goto again;
}
// block_length=0 and block_offset set by bgzf_seek.
if (fp->block_length != 0) fp->block_offset = 0;
if (!j->hit_eof) fp->block_address = j->block_address;
fp->block_clength = j->comp_len;
fp->block_length = j->uncomp_len;
// bgzf_read() can change fp->block_length
fp->last_block_eof = (fp->block_length == 0);
if ( j->uncomp_len && j->fp->idx_build_otf )
{
bgzf_index_add_block(j->fp);
j->fp->idx->ublock_addr += j->uncomp_len;
}
// Steal the data block as it's quicker than a memcpy.
// We just need to make sure we delay the pool free.
if (fp->mt->curr_job) {
pthread_mutex_lock(&fp->mt->job_pool_m);
pool_free(fp->mt->job_pool, fp->mt->curr_job);
pthread_mutex_unlock(&fp->mt->job_pool_m);
}
fp->uncompressed_block = j->uncomp_data;
fp->mt->curr_job = j;
if (fp->mt->free_block) {
free(fp->mt->free_block); // clear up last non-mt block
fp->mt->free_block = NULL;
}
hts_tpool_delete_result(r, 0);
return 0;
}
uint8_t header[BLOCK_HEADER_LENGTH], *compressed_block;
int count, size, block_length, remaining;
single_threaded:
size = 0;
// Reading an uncompressed file
if ( !fp->is_compressed )
{
count = hread(fp->fp, fp->uncompressed_block, BGZF_MAX_BLOCK_SIZE);
if (count < 0) // Error
{
fp->errcode |= BGZF_ERR_IO;
return -1;
}
else if (count == 0) // EOF
{
fp->block_length = 0;
return 0;
}
if (fp->block_length != 0) fp->block_offset = 0;
fp->block_address += count;
fp->block_length = count;
return 0;
}
// Reading compressed file
int64_t block_address;
block_address = bgzf_htell(fp);
if ( fp->is_gzip && fp->gz_stream ) // is this is an initialized gzip stream?
{
count = inflate_gzip_block(fp, 0);
if ( count<0 )
{
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
fp->block_length = count;
fp->block_address = block_address;
return 0;
}
if (fp->cache_size && load_block_from_cache(fp, block_address)) return 0;
// loop to skip empty bgzf blocks
while (1)
{
count = hread(fp->fp, header, sizeof(header));
if (count == 0) { // no data read
if (!fp->last_block_eof && !fp->no_eof_block && !fp->is_gzip) {
fp->no_eof_block = 1;
hts_log_warning("EOF marker is absent. The input is probably truncated");
}
fp->block_length = 0;
return 0;
}
int ret;
if ( count != sizeof(header) || (ret=check_header(header))==-2 )
{
fp->errcode |= BGZF_ERR_HEADER;
return -1;
}
if ( ret==-1 )
{
// GZIP, not BGZF
uint8_t *cblock = (uint8_t*)fp->compressed_block;
memcpy(cblock, header, sizeof(header));
count = hread(fp->fp, cblock+sizeof(header), BGZF_BLOCK_SIZE - sizeof(header)) + sizeof(header);
int nskip = 10;
// Check optional fields to skip: FLG.FNAME,FLG.FCOMMENT,FLG.FHCRC,FLG.FEXTRA
// Note: Some of these fields are untested, I did not have appropriate data available
if ( header[3] & 0x4 ) // FLG.FEXTRA
{
nskip += unpackInt16(&cblock[nskip]) + 2;
}
if ( header[3] & 0x8 ) // FLG.FNAME
{
while ( nskip<count && cblock[nskip] ) nskip++;
nskip++;
}
if ( header[3] & 0x10 ) // FLG.FCOMMENT
{
while ( nskip<count && cblock[nskip] ) nskip++;
nskip++;
}
if ( header[3] & 0x2 ) nskip += 2; // FLG.FHCRC
/* FIXME: Should handle this better. There's no reason why
someone shouldn't include a massively long comment in their
gzip stream. */
if ( nskip >= count )
{
fp->errcode |= BGZF_ERR_HEADER;
return -1;
}
fp->is_gzip = 1;
fp->gz_stream = (z_stream*) calloc(1,sizeof(z_stream));
int ret = inflateInit2(fp->gz_stream, -15);
if (ret != Z_OK)
{
hts_log_error("Call to inflateInit2 failed: %s", bgzf_zerr(ret, fp->gz_stream));
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
fp->gz_stream->avail_in = count - nskip;
fp->gz_stream->next_in = cblock + nskip;
count = inflate_gzip_block(fp, 1);
if ( count<0 )
{
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
fp->block_length = count;
fp->block_address = block_address;
if ( fp->idx_build_otf ) return -1; // cannot build index for gzip
return 0;
}
size = count;
block_length = unpackInt16((uint8_t*)&header[16]) + 1; // +1 because when writing this number, we used "-1"
if (block_length < BLOCK_HEADER_LENGTH)
{
fp->errcode |= BGZF_ERR_HEADER;
return -1;
}
compressed_block = (uint8_t*)fp->compressed_block;
memcpy(compressed_block, header, BLOCK_HEADER_LENGTH);
remaining = block_length - BLOCK_HEADER_LENGTH;
count = hread(fp->fp, &compressed_block[BLOCK_HEADER_LENGTH], remaining);
if (count != remaining) {
fp->errcode |= BGZF_ERR_IO;
return -1;
}
size += count;
if ((count = inflate_block(fp, block_length)) < 0) {
hts_log_debug("Inflate block operation failed: %s", bgzf_zerr(count, NULL));
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
fp->last_block_eof = (count == 0);
if ( count ) break; // otherwise an empty bgzf block
}
if (fp->block_length != 0) fp->block_offset = 0; // Do not reset offset if this read follows a seek.
fp->block_address = block_address;
fp->block_length = count;
if ( fp->idx_build_otf )
{
bgzf_index_add_block(fp);
fp->idx->ublock_addr += count;
}
cache_block(fp, size);
return 0;
}
ssize_t bgzf_read(BGZF *fp, void *data, size_t length)
{
ssize_t bytes_read = 0;
uint8_t *output = (uint8_t*)data;
if (length <= 0) return 0;
assert(fp->is_write == 0);
while (bytes_read < length) {
int copy_length, available = fp->block_length - fp->block_offset;
uint8_t *buffer;
if (available <= 0) {
int ret = bgzf_read_block(fp);
if (ret != 0) {
hts_log_error("Read block operation failed with error %d after %zd of %zu bytes", ret, bytes_read, length);
fp->errcode |= BGZF_ERR_ZLIB;
return -1;
}
available = fp->block_length - fp->block_offset;
if (available <= 0) break;
}
copy_length = length - bytes_read < available? length - bytes_read : available;
buffer = (uint8_t*)fp->uncompressed_block;
memcpy(output, buffer + fp->block_offset, copy_length);
fp->block_offset += copy_length;
output += copy_length;
bytes_read += copy_length;
// For raw gzip streams this avoids short reads.
if (fp->block_offset == fp->block_length) {
fp->block_address = bgzf_htell(fp);
fp->block_offset = fp->block_length = 0;
}
}
fp->uncompressed_address += bytes_read;
return bytes_read;
}
ssize_t bgzf_raw_read(BGZF *fp, void *data, size_t length)
{
ssize_t ret = hread(fp->fp, data, length);
if (ret < 0) fp->errcode |= BGZF_ERR_IO;
return ret;
}
#ifdef BGZF_MT
void *bgzf_encode_func(void *arg) {
bgzf_job *j = (bgzf_job *)arg;
j->comp_len = BGZF_MAX_BLOCK_SIZE;
int ret = bgzf_compress(j->comp_data, &j->comp_len,
j->uncomp_data, j->uncomp_len,
j->fp->compress_level);
if (ret != 0)
j->errcode |= BGZF_ERR_ZLIB;
return arg;
}
// Our input block has already been decoded by bgzf_mt_read_block().
// We need to split that into a fetch block (compressed) and make this
// do the actual decompression step.
void *bgzf_decode_func(void *arg) {
bgzf_job *j = (bgzf_job *)arg;
j->uncomp_len = BGZF_MAX_BLOCK_SIZE;
int ret = bgzf_uncompress(j->uncomp_data, &j->uncomp_len,
j->comp_data+18, j->comp_len-18);
if (ret != 0)
j->errcode |= BGZF_ERR_ZLIB;
return arg;
}
/*
* Nul function so we can dispatch a job with the correct serial
* to mark failure or to indicate an empty read (EOF).
*/
void *bgzf_nul_func(void *arg) { return arg; }
/*
* Takes compressed blocks off the results queue and calls hwrite to
* punt them to the output stream.
*
* Returns NULL when no more are left, or -1 on error
*/
static void *bgzf_mt_writer(void *vp) {
BGZF *fp = (BGZF *)vp;
mtaux_t *mt = fp->mt;
hts_tpool_result *r;
// Iterates until result queue is shutdown, where it returns NULL.
while ((r = hts_tpool_next_result_wait(mt->out_queue))) {
bgzf_job *j = (bgzf_job *)hts_tpool_result_data(r);
assert(j);
if (hwrite(fp->fp, j->comp_data, j->comp_len) != j->comp_len) {
fp->errcode |= BGZF_ERR_IO;
goto err;
}
/*
* Periodically call hflush (which calls fsync when on a file).
* This avoids the fsync being done at the bgzf_close stage,
* which can sometimes cause signficant delays. As this is in
* a separate thread, spreading the sync delays throughout the
* program execution seems better.
* Frequency of 1/512 has been chosen by experimentation
* across local XFS, NFS and Lustre tests.
*/
if (++mt->flush_pending % 512 == 0)
if (hflush(fp->fp) != 0)
goto err;
hts_tpool_delete_result(r, 0);
// Also updated by main thread
pthread_mutex_lock(&mt->job_pool_m);
pool_free(mt->job_pool, j);
mt->jobs_pending--;
pthread_mutex_unlock(&mt->job_pool_m);
}