X-Git-Url: http://woldlab.caltech.edu/gitweb/?p=samtools.git;a=blobdiff_plain;f=bam_import.c;h=9d84328bdbdafb5f4e76a19e7edf176121aca5b4;hp=fccaa022208131b27093a2b44f32e74d13a469c0;hb=cefa18095b2479339b08111936313066ec548657;hpb=b27e00385f41769d03a8cca4dbd71275fc9fa906

diff --git a/bam_import.c b/bam_import.c
index fccaa02..9d84328 100644
--- a/bam_import.c
+++ b/bam_import.c
@@ -5,8 +5,12 @@
 #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"
 
@@ -36,6 +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 {
@@ -93,15 +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) {
@@ -111,6 +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);
@@ -119,6 +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;
@@ -139,115 +168,49 @@ 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,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_header_parse_rg(bam_header_t *h)
-{
-	kstring_t *rgid, *rglib;
-	char *p, *q, *s, *r;
-	int n = 0;
-
-	// free
-	if (h == 0) return 0;
-	bam_strmap_destroy(h->rg2lib); h->rg2lib = 0;
-	if (h->l_text < 3) return 0;
-	// parse @RG lines
-	h->rg2lib = bam_strmap_init();
-	rgid = calloc(1, sizeof(kstring_t));
-	rglib = calloc(1, sizeof(kstring_t));
-	s = h->text;
-	while ((s = strstr(s, "@RG")) != 0) {
-		if (rgid->l && rglib->l) {
-			bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
-			++n;
-		}
-		rgid->l = rglib->l = 0;
-		s += 3;
-		r = s;
-		if ((p = strstr(s, "ID:")) != 0) {
-			q = p + 3;
-			for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
-			kputsn(q, p - q, rgid);
-		} else {
-			fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
-			break;
-		}
-		if (r < p) r = p;
-		if ((p = strstr(s, "LB:")) != 0) {
-			q = p + 3;
-			for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
-			kputsn(q, p - q, rglib);
-		} else {
-			fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
-			break;
-		}
-		if (r < p) r = p;
-		s = r + 3;
-	}
-	if (rgid->l && rglib->l) {
-		bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
-		++n;
-	}
-	free(rgid->s); free(rgid);
-	free(rglib->s); free(rglib);
-	if (n == 0) {
-		bam_strmap_destroy(h->rg2lib);
-		h->rg2lib = 0;
-	}
-	return n;
-}
-
 int sam_header_parse(bam_header_t *h)
 {
+	char **tmp;
 	int i;
-	char *s, *p, *q, *r;
-
-	// free
 	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;
-	// count number of @SQ
-	s = h->text;
-	while ((s = strstr(s, "@SQ")) != 0) {
-		++h->n_targets;
-		s += 3;
-	}
+	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_len = (uint32_t*)calloc(h->n_targets, 4);
-	h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
-	// parse @SQ lines
-	i = 0;
-	s = h->text;
-	while ((s = strstr(s, "@SQ")) != 0) {
-		s += 3;
-		r = s;
-		if ((p = strstr(s, "SN:")) != 0) {
-			q = p + 3;
-			for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
-			h->target_name[i] = (char*)calloc(p - q + 1, 1);
-			strncpy(h->target_name[i], q, p - q);
-		} else goto header_err_ret;
-		if (r < p) r = p;
-		if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
-		else goto header_err_ret;
-		if (r < p) r = p;
-		s = r + 3;
-		++i;
-	}
-	sam_header_parse_rg(h);
+	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;
-
-header_err_ret:
-	fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
-	free(h->target_len); free(h->target_name);
-	h->n_targets = 0; h->target_len = 0; h->target_name = 0;
-	return 0;
 }
 
 bam_header_t *sam_header_read(tamFile fp)
@@ -296,8 +259,19 @@ 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) {
@@ -325,7 +299,7 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 			for (i = 0, s = str->s; i != c->n_cigar; ++i) {
 				x = strtol(s, &t, 10);
 				op = toupper(*t);
-				if (op == 'M') op = BAM_CMATCH;
+				if (op == 'M' || op == '=' || op == 'X') op = BAM_CMATCH;
 				else if (op == 'I') op = BAM_CINS;
 				else if (op == 'D') op = BAM_CDEL;
 				else if (op == 'N') op = BAM_CREF_SKIP;
@@ -339,7 +313,13 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 			if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
 			c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
 			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;
@@ -352,16 +332,18 @@ 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)))
+				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))
@@ -457,9 +439,11 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 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);
 	return fp;
 }