0
|
1 #include <stdlib.h>
|
|
2 #include <ctype.h>
|
|
3 #include <assert.h>
|
|
4 #include <stdio.h>
|
|
5 #include <string.h>
|
|
6 #include <unistd.h>
|
|
7 #include "bam.h"
|
|
8 #include "ksort.h"
|
|
9
|
|
10 static int g_is_by_qname = 0;
|
|
11
|
|
12 static inline int strnum_cmp(const char *a, const char *b)
|
|
13 {
|
|
14 char *pa, *pb;
|
|
15 pa = (char*)a; pb = (char*)b;
|
|
16 while (*pa && *pb) {
|
|
17 if (isdigit(*pa) && isdigit(*pb)) {
|
|
18 long ai, bi;
|
|
19 ai = strtol(pa, &pa, 10);
|
|
20 bi = strtol(pb, &pb, 10);
|
|
21 if (ai != bi) return ai<bi? -1 : ai>bi? 1 : 0;
|
|
22 } else {
|
|
23 if (*pa != *pb) break;
|
|
24 ++pa; ++pb;
|
|
25 }
|
|
26 }
|
|
27 if (*pa == *pb)
|
|
28 return (pa-a) < (pb-b)? -1 : (pa-a) > (pb-b)? 1 : 0;
|
|
29 return *pa<*pb? -1 : *pa>*pb? 1 : 0;
|
|
30 }
|
|
31
|
|
32 #define HEAP_EMPTY 0xffffffffffffffffull
|
|
33
|
|
34 typedef struct {
|
|
35 int i;
|
|
36 uint64_t pos;
|
|
37 bam1_t *b;
|
|
38 } heap1_t;
|
|
39
|
|
40 static inline int heap_lt(const heap1_t a, const heap1_t b)
|
|
41 {
|
|
42 if (g_is_by_qname) {
|
|
43 int t;
|
|
44 if (a.b == 0 || b.b == 0) return a.b == 0? 1 : 0;
|
|
45 t = strnum_cmp(bam1_qname(a.b), bam1_qname(b.b));
|
|
46 return (t > 0 || (t == 0 && a.pos > b.pos));
|
|
47 } else return (a.pos > b.pos);
|
|
48 }
|
|
49
|
|
50 KSORT_INIT(heap, heap1_t, heap_lt)
|
|
51
|
|
52 static void swap_header_text(bam_header_t *h1, bam_header_t *h2)
|
|
53 {
|
|
54 int tempi;
|
|
55 char *temps;
|
|
56 tempi = h1->l_text, h1->l_text = h2->l_text, h2->l_text = tempi;
|
|
57 temps = h1->text, h1->text = h2->text, h2->text = temps;
|
|
58 }
|
|
59
|
|
60 /*!
|
|
61 @abstract Merge multiple sorted BAM.
|
|
62 @param is_by_qname whether to sort by query name
|
|
63 @param out output BAM file name
|
|
64 @param headers name of SAM file from which to copy '@' header lines,
|
|
65 or NULL to copy them from the first file to be merged
|
|
66 @param n number of files to be merged
|
|
67 @param fn names of files to be merged
|
|
68
|
|
69 @discussion Padding information may NOT correctly maintained. This
|
|
70 function is NOT thread safe.
|
|
71 */
|
|
72 void bam_merge_core(int by_qname, const char *out, const char *headers, int n, char * const *fn)
|
|
73 {
|
|
74 bamFile fpout, *fp;
|
|
75 heap1_t *heap;
|
|
76 bam_header_t *hout = 0;
|
|
77 bam_header_t *hheaders = NULL;
|
|
78 int i, j;
|
|
79
|
|
80 if (headers) {
|
|
81 tamFile fpheaders = sam_open(headers);
|
|
82 if (fpheaders == 0) {
|
|
83 fprintf(stderr, "[bam_merge_core] Cannot open file `%s'. Continue anyway.\n", headers);
|
|
84 } else {
|
|
85 hheaders = sam_header_read(fpheaders);
|
|
86 sam_close(fpheaders);
|
|
87 }
|
|
88 }
|
|
89
|
|
90 g_is_by_qname = by_qname;
|
|
91 fp = (bamFile*)calloc(n, sizeof(bamFile));
|
|
92 heap = (heap1_t*)calloc(n, sizeof(heap1_t));
|
|
93 for (i = 0; i != n; ++i) {
|
|
94 heap1_t *h;
|
|
95 bam_header_t *hin;
|
|
96 assert(fp[i] = bam_open(fn[i], "r"));
|
|
97 hin = bam_header_read(fp[i]);
|
|
98 if (i == 0) { // the first SAM
|
|
99 hout = hin;
|
|
100 if (hheaders) {
|
|
101 // If the text headers to be swapped in include any @SQ headers,
|
|
102 // check that they are consistent with the existing binary list
|
|
103 // of reference information.
|
|
104 if (hheaders->n_targets > 0) {
|
|
105 if (hout->n_targets != hheaders->n_targets)
|
|
106 fprintf(stderr, "[bam_merge_core] number of @SQ headers in `%s' differs from number of target sequences", headers);
|
|
107 for (j = 0; j < hout->n_targets; ++j)
|
|
108 if (strcmp(hout->target_name[j], hheaders->target_name[j]) != 0)
|
|
109 fprintf(stderr, "[bam_merge_core] @SQ header '%s' in '%s' differs from target sequence", hheaders->target_name[j], headers);
|
|
110 }
|
|
111 swap_header_text(hout, hheaders);
|
|
112 bam_header_destroy(hheaders);
|
|
113 hheaders = NULL;
|
|
114 }
|
|
115 } else { // validate multiple baf
|
|
116 if (hout->n_targets != hin->n_targets) {
|
|
117 fprintf(stderr, "[bam_merge_core] file '%s' has different number of target sequences. Abort!\n", fn[i]);
|
|
118 exit(1);
|
|
119 }
|
|
120 for (j = 0; j < hout->n_targets; ++j) {
|
|
121 if (strcmp(hout->target_name[j], hin->target_name[j])) {
|
|
122 fprintf(stderr, "[bam_merge_core] different target sequence name: '%s' != '%s' in file '%s'. Abort!\n",
|
|
123 hout->target_name[j], hin->target_name[j], fn[i]);
|
|
124 exit(1);
|
|
125 }
|
|
126 }
|
|
127 bam_header_destroy(hin);
|
|
128 }
|
|
129 h = heap + i;
|
|
130 h->i = i;
|
|
131 h->b = (bam1_t*)calloc(1, sizeof(bam1_t));
|
|
132 if (bam_read1(fp[i], h->b) >= 0)
|
|
133 h->pos = ((uint64_t)h->b->core.tid<<32) | (uint32_t)h->b->core.pos<<1 | bam1_strand(h->b);
|
|
134 else h->pos = HEAP_EMPTY;
|
|
135 }
|
|
136 fpout = strcmp(out, "-")? bam_open(out, "w") : bam_dopen(fileno(stdout), "w");
|
|
137 assert(fpout);
|
|
138 bam_header_write(fpout, hout);
|
|
139 bam_header_destroy(hout);
|
|
140
|
|
141 ks_heapmake(heap, n, heap);
|
|
142 while (heap->pos != HEAP_EMPTY) {
|
|
143 bam1_t *b = heap->b;
|
|
144 bam_write1_core(fpout, &b->core, b->data_len, b->data);
|
|
145 if ((j = bam_read1(fp[heap->i], b)) >= 0) {
|
|
146 heap->pos = ((uint64_t)b->core.tid<<32) | (uint32_t)b->core.pos<<1 | bam1_strand(b);
|
|
147 } else if (j == -1) {
|
|
148 heap->pos = HEAP_EMPTY;
|
|
149 free(heap->b->data); free(heap->b);
|
|
150 heap->b = 0;
|
|
151 } else fprintf(stderr, "[bam_merge_core] '%s' is truncated. Continue anyway.\n", fn[heap->i]);
|
|
152 ks_heapadjust(heap, 0, n, heap);
|
|
153 }
|
|
154
|
|
155 for (i = 0; i != n; ++i) bam_close(fp[i]);
|
|
156 bam_close(fpout);
|
|
157 free(fp); free(heap);
|
|
158 }
|
|
159 int bam_merge(int argc, char *argv[])
|
|
160 {
|
|
161 int c, is_by_qname = 0;
|
|
162 char *fn_headers = NULL;
|
|
163
|
|
164 while ((c = getopt(argc, argv, "h:n")) >= 0) {
|
|
165 switch (c) {
|
|
166 case 'h': fn_headers = strdup(optarg); break;
|
|
167 case 'n': is_by_qname = 1; break;
|
|
168 }
|
|
169 }
|
|
170 if (optind + 2 >= argc) {
|
|
171 fprintf(stderr, "\n");
|
|
172 fprintf(stderr, "Usage: samtools merge [-n] [-h inh.sam] <out.bam> <in1.bam> <in2.bam> [...]\n\n");
|
|
173 fprintf(stderr, "Options: -n sort by read names\n");
|
|
174 fprintf(stderr, " -h FILE copy the header in FILE to <out.bam> [in1.bam]\n\n");
|
|
175 fprintf(stderr, "Note: Samtools' merge does not reconstruct the @RG dictionary in the header. Users\n");
|
|
176 fprintf(stderr, " must provide the correct header with -h, or uses Picard which properly maintains\n");
|
|
177 fprintf(stderr, " the header dictionary in merging.\n\n");
|
|
178 return 1;
|
|
179 }
|
|
180 bam_merge_core(is_by_qname, argv[optind], fn_headers, argc - optind - 1, argv + optind + 1);
|
|
181 free(fn_headers);
|
|
182 return 0;
|
|
183 }
|
|
184
|
|
185 typedef bam1_t *bam1_p;
|
|
186
|
|
187 static inline int bam1_lt(const bam1_p a, const bam1_p b)
|
|
188 {
|
|
189 if (g_is_by_qname) {
|
|
190 int t = strnum_cmp(bam1_qname(a), bam1_qname(b));
|
|
191 return (t < 0 || (t == 0 && (((uint64_t)a->core.tid<<32|a->core.pos) < ((uint64_t)b->core.tid<<32|b->core.pos))));
|
|
192 } else return (((uint64_t)a->core.tid<<32|a->core.pos) < ((uint64_t)b->core.tid<<32|b->core.pos));
|
|
193 }
|
|
194 KSORT_INIT(sort, bam1_p, bam1_lt)
|
|
195
|
|
196 static void sort_blocks(int n, int k, bam1_p *buf, const char *prefix, const bam_header_t *h)
|
|
197 {
|
|
198 char *name;
|
|
199 int i;
|
|
200 bamFile fp;
|
|
201 ks_mergesort(sort, k, buf, 0);
|
|
202 name = (char*)calloc(strlen(prefix) + 20, 1);
|
|
203 if (n >= 0) sprintf(name, "%s.%.4d.bam", prefix, n);
|
|
204 else sprintf(name, "%s.bam", prefix);
|
|
205 assert(fp = bam_open(name, "w"));
|
|
206 free(name);
|
|
207 bam_header_write(fp, h);
|
|
208 for (i = 0; i < k; ++i)
|
|
209 bam_write1_core(fp, &buf[i]->core, buf[i]->data_len, buf[i]->data);
|
|
210 bam_close(fp);
|
|
211 }
|
|
212
|
|
213 /*!
|
|
214 @abstract Sort an unsorted BAM file based on the chromosome order
|
|
215 and the leftmost position of an alignment
|
|
216
|
|
217 @param is_by_qname whether to sort by query name
|
|
218 @param fn name of the file to be sorted
|
|
219 @param prefix prefix of the output and the temporary files; upon
|
|
220 sucessess, prefix.bam will be written.
|
|
221 @param max_mem approxiate maximum memory (very inaccurate)
|
|
222
|
|
223 @discussion It may create multiple temporary subalignment files
|
|
224 and then merge them by calling bam_merge_core(). This function is
|
|
225 NOT thread safe.
|
|
226 */
|
|
227 void bam_sort_core(int is_by_qname, const char *fn, const char *prefix, size_t max_mem)
|
|
228 {
|
|
229 int n, ret, k, i;
|
|
230 size_t mem;
|
|
231 bam_header_t *header;
|
|
232 bamFile fp;
|
|
233 bam1_t *b, **buf;
|
|
234
|
|
235 g_is_by_qname = is_by_qname;
|
|
236 n = k = 0; mem = 0;
|
|
237 fp = strcmp(fn, "-")? bam_open(fn, "r") : bam_dopen(fileno(stdin), "r");
|
|
238 assert(fp);
|
|
239 header = bam_header_read(fp);
|
|
240 buf = (bam1_t**)calloc(max_mem / BAM_CORE_SIZE, sizeof(bam1_t*));
|
|
241 // write sub files
|
|
242 for (;;) {
|
|
243 if (buf[k] == 0) buf[k] = (bam1_t*)calloc(1, sizeof(bam1_t));
|
|
244 b = buf[k];
|
|
245 if ((ret = bam_read1(fp, b)) < 0) break;
|
|
246 mem += ret;
|
|
247 ++k;
|
|
248 if (mem >= max_mem) {
|
|
249 sort_blocks(n++, k, buf, prefix, header);
|
|
250 mem = 0; k = 0;
|
|
251 }
|
|
252 }
|
|
253 if (ret != -1)
|
|
254 fprintf(stderr, "[bam_sort_core] truncated file. Continue anyway.\n");
|
|
255 if (n == 0) sort_blocks(-1, k, buf, prefix, header);
|
|
256 else { // then merge
|
|
257 char **fns, *fnout;
|
|
258 fprintf(stderr, "[bam_sort_core] merging from %d files...\n", n+1);
|
|
259 sort_blocks(n++, k, buf, prefix, header);
|
|
260 fnout = (char*)calloc(strlen(prefix) + 20, 1);
|
|
261 sprintf(fnout, "%s.bam", prefix);
|
|
262 fns = (char**)calloc(n, sizeof(char*));
|
|
263 for (i = 0; i < n; ++i) {
|
|
264 fns[i] = (char*)calloc(strlen(prefix) + 20, 1);
|
|
265 sprintf(fns[i], "%s.%.4d.bam", prefix, i);
|
|
266 }
|
|
267 bam_merge_core(is_by_qname, fnout, 0, n, fns);
|
|
268 free(fnout);
|
|
269 for (i = 0; i < n; ++i) {
|
|
270 unlink(fns[i]);
|
|
271 free(fns[i]);
|
|
272 }
|
|
273 free(fns);
|
|
274 }
|
|
275 for (k = 0; k < max_mem / BAM_CORE_SIZE; ++k) {
|
|
276 if (buf[k]) {
|
|
277 free(buf[k]->data);
|
|
278 free(buf[k]);
|
|
279 }
|
|
280 }
|
|
281 free(buf);
|
|
282 bam_header_destroy(header);
|
|
283 bam_close(fp);
|
|
284 }
|
|
285
|
|
286 int bam_sort(int argc, char *argv[])
|
|
287 {
|
|
288 size_t max_mem = 500000000;
|
|
289 int c, is_by_qname = 0;
|
|
290 while ((c = getopt(argc, argv, "nm:")) >= 0) {
|
|
291 switch (c) {
|
|
292 case 'n': is_by_qname = 1; break;
|
|
293 case 'm': max_mem = atol(optarg); break;
|
|
294 }
|
|
295 }
|
|
296 if (optind + 2 > argc) {
|
|
297 fprintf(stderr, "Usage: samtools sort [-n] [-m <maxMem>] <in.bam> <out.prefix>\n");
|
|
298 return 1;
|
|
299 }
|
|
300 bam_sort_core(is_by_qname, argv[optind], argv[optind+1], max_mem);
|
|
301 return 0;
|
|
302 }
|