X-Git-Url: https://git.donarmstrong.com/?p=samtools.git;a=blobdiff_plain;f=bam_import.c;h=da2bf942ed90573cb3c964b4ced593ba7cbcf3d8;hp=9daa29f42a1676077b29f3f63ab7a7027896dbad;hb=9f118264ea012adc21a46d7c03eaad4f9ce7d4d4;hpb=f1222861f3824c28d35cd143f5565ca09d80f056

diff --git a/bam_import.c b/bam_import.c
index 9daa29f..da2bf94 100644
--- a/bam_import.c
+++ b/bam_import.c
@@ -5,11 +5,16 @@
 #include <stdlib.h>
 #include <unistd.h>
 #include <assert.h>
+#ifdef _WIN32
+#include <fcntl.h>
+#endif
+#include "kstring.h"
 #include "bam.h"
+#include "sam_header.h"
 #include "kseq.h"
 #include "khash.h"
 
-KSTREAM_INIT(gzFile, gzread, 8192)
+KSTREAM_INIT(gzFile, gzread, 16384)
 KHASH_MAP_INIT_STR(ref, uint64_t)
 
 void bam_init_header_hash(bam_header_t *header);
@@ -35,6 +40,25 @@ unsigned char bam_nt16_table[256] = {
 	15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15
 };
 
+unsigned short bam_char2flag_table[256] = {
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0,
+	BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
+	0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0
+};
+
 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
 
 struct __tamFile_t {
@@ -42,6 +66,7 @@ struct __tamFile_t {
 	kstream_t *ks;
 	kstring_t *str;
 	uint64_t n_lines;
+	int is_first;
 };
 
 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
@@ -91,15 +116,16 @@ static bam_header_t *hash2header(const kh_ref_t *hash)
 bam_header_t *sam_header_read2(const char *fn)
 {
 	bam_header_t *header;
-	int c, dret, ret;
+	int c, dret, ret, error = 0;
 	gzFile fp;
 	kstream_t *ks;
 	kstring_t *str;
 	kh_ref_t *hash;
 	khiter_t k;
-	hash = kh_init(ref);
+	if (fn == 0) return 0;
 	fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
-	assert(fp);
+	if (fp == 0) return 0;
+	hash = kh_init(ref);
 	ks = ks_init(fp);
 	str = (kstring_t*)calloc(1, sizeof(kstring_t));
 	while (ks_getuntil(ks, 0, str, &dret) > 0) {
@@ -109,6 +135,10 @@ bam_header_t *sam_header_read2(const char *fn)
 		ks_getuntil(ks, 0, str, &dret);
 		len = atoi(str->s);
 		k = kh_put(ref, hash, s, &ret);
+		if (ret == 0) {
+			fprintf(stderr, "[sam_header_read2] duplicated sequence name: %s\n", s);
+			error = 1;
+		}
 		kh_value(hash, k) = (uint64_t)len<<32 | i;
 		if (dret != '\n')
 			while ((c = ks_getc(ks)) != '\n' && c != -1);
@@ -117,6 +147,7 @@ bam_header_t *sam_header_read2(const char *fn)
 	gzclose(fp);
 	free(str->s); free(str);
 	fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
+	if (error) return 0;
 	header = hash2header(hash);
 	kh_destroy(ref, hash);
 	return header;
@@ -137,35 +168,87 @@ static inline void parse_error(int64_t n_lines, const char * __restrict msg)
 }
 static inline void append_text(bam_header_t *header, kstring_t *str)
 {
-	int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
+	size_t x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
 	kroundup32(x); kroundup32(y);
-	if (x < y) header->text = (char*)realloc(header->text, y);
+	if (x < y) 
+    {
+        header->n_text = y;
+        header->text = (char*)realloc(header->text, y);
+        if ( !header->text ) 
+        {
+            fprintf(stderr,"realloc failed to alloc %ld bytes\n", y);
+            abort();
+        }
+    }
+    // Sanity check
+    if ( header->l_text+str->l+1 >= header->n_text )
+    {
+        fprintf(stderr,"append_text FIXME: %ld>=%ld, x=%ld,y=%ld\n",  header->l_text+str->l+1,(long)header->n_text,x,y);
+        abort();
+    }
 	strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
 	header->l_text += str->l + 1;
 	header->text[header->l_text] = 0;
 }
-int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
+
+int sam_header_parse(bam_header_t *h)
 {
-	int ret, doff, doff0, dret, z = 0;
-	bam1_core_t *c = &b->core;
-	kstring_t *str = fp->str;
-	kstream_t *ks = fp->ks;
+	char **tmp;
+	int i;
+	free(h->target_len); free(h->target_name);
+	h->n_targets = 0; h->target_len = 0; h->target_name = 0;
+	if (h->l_text < 3) return 0;
+	if (h->dict == 0) h->dict = sam_header_parse2(h->text);
+	tmp = sam_header2list(h->dict, "SQ", "SN", &h->n_targets);
+	if (h->n_targets == 0) return 0;
+	h->target_name = calloc(h->n_targets, sizeof(void*));
+	for (i = 0; i < h->n_targets; ++i)
+		h->target_name[i] = strdup(tmp[i]);
+	free(tmp);
+	tmp = sam_header2list(h->dict, "SQ", "LN", &h->n_targets);
+	h->target_len = calloc(h->n_targets, 4);
+	for (i = 0; i < h->n_targets; ++i)
+		h->target_len[i] = atoi(tmp[i]);
+	free(tmp);
+	return h->n_targets;
+}
 
+bam_header_t *sam_header_read(tamFile fp)
+{
+	int ret, dret;
+	bam_header_t *header = bam_header_init();
+	kstring_t *str = fp->str;
 	while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
-		z += str->l + 1;
 		str->s[str->l] = dret; // note that str->s is NOT null terminated!!
 		append_text(header, str);
 		if (dret != '\n') {
 			ret = ks_getuntil(fp->ks, '\n', str, &dret);
-			z += str->l + 1;
 			str->s[str->l] = '\n'; // NOT null terminated!!
 			append_text(header, str);
 		}
 		++fp->n_lines;
 	}
-	while (ret == 0) { // special consideration for "\r\n" and empty lines
-		ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
-		if (ret >= 0) z += str->l + 1;
+	sam_header_parse(header);
+	bam_init_header_hash(header);
+	fp->is_first = 1;
+	return header;
+}
+
+int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
+{
+	int ret, doff, doff0, dret, z = 0;
+	bam1_core_t *c = &b->core;
+	kstring_t *str = fp->str;
+	kstream_t *ks = fp->ks;
+
+	if (fp->is_first) {
+		fp->is_first = 0;
+		ret = str->l;
+	} else {
+		do { // special consideration for empty lines
+			ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
+			if (ret >= 0) z += str->l + 1;
+		} while (ret == 0);
 	}
 	if (ret < 0) return -1;
 	++fp->n_lines;
@@ -176,9 +259,26 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 		memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
 		doff += c->l_qname;
 	}
-	{ // flag, tid, pos, qual
-		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->flag = atoi(str->s);
+	{ // flag
+		long flag;
+		char *s;
+		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
+		flag = strtol((char*)str->s, &s, 0);
+		if (*s) { // not the end of the string
+			flag = 0;
+			for (s = str->s; *s; ++s)
+				flag |= bam_char2flag_table[(int)*s];
+		}
+		c->flag = flag;
+	}
+	{ // tid, pos, qual
 		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
+		if (c->tid < 0 && strcmp(str->s, "*")) {
+			if (header->n_targets == 0) {
+				fprintf(stderr, "[sam_read1] missing header? Abort!\n");
+				exit(1);
+			} else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
+		}
 		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
 		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
 		if (ret < 0) return -2;
@@ -191,11 +291,13 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 		if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
 		z += str->l + 1;
 		if (str->s[0] != '*') {
+			uint32_t *cigar;
 			for (s = str->s; *s; ++s) {
-				if (isalpha(*s)) ++c->n_cigar;
+				if ((isalpha(*s)) || (*s=='=')) ++c->n_cigar;
 				else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
 			}
 			b->data = alloc_data(b, doff + c->n_cigar * 4);
+			cigar = bam1_cigar(b);
 			for (i = 0, s = str->s; i != c->n_cigar; ++i) {
 				x = strtol(s, &t, 10);
 				op = toupper(*t);
@@ -206,14 +308,23 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 				else if (op == 'S') op = BAM_CSOFT_CLIP;
 				else if (op == 'H') op = BAM_CHARD_CLIP;
 				else if (op == 'P') op = BAM_CPAD;
+				else if (op == '=') op = BAM_CEQUAL;
+				else if (op == 'X') op = BAM_CDIFF;
+				else if (op == 'B') op = BAM_CBACK;
 				else parse_error(fp->n_lines, "invalid CIGAR operation");
 				s = t + 1;
-				bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
+				cigar[i] = bam_cigar_gen(x, op);
 			}
 			if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
-			c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
+			c->bin = bam_reg2bin(c->pos, bam_calend(c, cigar));
 			doff += c->n_cigar * 4;
-		} else c->bin = bam_reg2bin(c->pos, c->pos + 1);
+		} else {
+			if (!(c->flag&BAM_FUNMAP)) {
+				fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines);
+				c->flag |= BAM_FUNMAP;
+			}
+			c->bin = bam_reg2bin(c->pos, c->pos + 1);
+		}
 	}
 	{ // mtid, mpos, isize
 		ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
@@ -226,16 +337,21 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 	}
 	{ // seq and qual
 		int i;
-		uint8_t *p;
+		uint8_t *p = 0;
 		if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
 		z += str->l + 1;
-		c->l_qseq = strlen(str->s);
-		if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
-			parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
-		p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
-		bzero(p, (c->l_qseq+1)/2);
-		for (i = 0; i < c->l_qseq; ++i)
-			p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
+		if (strcmp(str->s, "*")) {
+			c->l_qseq = strlen(str->s);
+			if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b))) {
+				fprintf(stderr, "Line %ld, sequence length %i vs %i from CIGAR\n",
+						(long)fp->n_lines, c->l_qseq, (int32_t)bam_cigar2qlen(c, bam1_cigar(b)));
+				parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
+			}
+			p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
+			memset(p, 0, (c->l_qseq+1)/2);
+			for (i = 0; i < c->l_qseq; ++i)
+				p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
+		} else c->l_qseq = 0;
 		if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
 		z += str->l + 1;
 		if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
@@ -319,23 +435,46 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 				memcpy(s, str->s + 5, str->l - 5);
 				s[str->l - 5] = 0;
 				doff += size;
+			} else if (type == 'B') {
+				int32_t n = 0, Bsize, k = 0, size;
+				char *p;
+				if (str->l < 8) parse_error(fp->n_lines, "too few values in aux type B");
+				Bsize = bam_aux_type2size(str->s[5]); // the size of each element
+				for (p = (char*)str->s + 6; *p; ++p) // count the number of elements in the array
+					if (*p == ',') ++n;
+				p = str->s + 7; // now p points to the first number in the array
+				size = 6 + Bsize * n; // total number of bytes allocated to this tag
+				s = alloc_data(b, doff + 6 * Bsize * n) + doff; // allocate memory
+				*s++ = 'B'; *s++ = str->s[5];
+				memcpy(s, &n, 4); s += 4; // write the number of elements
+				if (str->s[5] == 'c')      while (p < str->s + str->l) ((int8_t*)s)[k++]   = (int8_t)strtol(p, &p, 0),   ++p;
+				else if (str->s[5] == 'C') while (p < str->s + str->l) ((uint8_t*)s)[k++]  = (uint8_t)strtol(p, &p, 0),  ++p;
+				else if (str->s[5] == 's') while (p < str->s + str->l) ((int16_t*)s)[k++]  = (int16_t)strtol(p, &p, 0),  ++p; // FIXME: avoid unaligned memory
+				else if (str->s[5] == 'S') while (p < str->s + str->l) ((uint16_t*)s)[k++] = (uint16_t)strtol(p, &p, 0), ++p;
+				else if (str->s[5] == 'i') while (p < str->s + str->l) ((int32_t*)s)[k++]  = (int32_t)strtol(p, &p, 0),  ++p;
+				else if (str->s[5] == 'I') while (p < str->s + str->l) ((uint32_t*)s)[k++] = (uint32_t)strtol(p, &p, 0), ++p;
+				else if (str->s[5] == 'f') while (p < str->s + str->l) ((float*)s)[k++]    = (float)strtod(p, &p),       ++p;
+				else parse_error(fp->n_lines, "unrecognized array type");
+				s += Bsize * n; doff += size;
 			} else parse_error(fp->n_lines, "unrecognized type");
 			if (dret == '\n' || dret == '\r') break;
 		}
 	}
 	b->l_aux = doff - doff0;
 	b->data_len = doff;
+	if (bam_no_B) bam_remove_B(b);
 	return z;
 }
 
 tamFile sam_open(const char *fn)
 {
 	tamFile fp;
+	gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb");
+	if (gzfp == 0) return 0;
 	fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
 	fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
-	fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
+	fp->fp = gzfp;
 	fp->ks = ks_init(fp->fp);
-	fp->n_lines = 0;
 	return fp;
 }