src/hg/lib/bamFile.c 1.13

Index: src/hg/lib/bamFile.c
===================================================================
RCS file: /projects/compbio/cvsroot/kent/src/hg/lib/bamFile.c,v
retrieving revision 1.12
retrieving revision 1.13
diff -b -B -U 1000000 -r1.12 -r1.13
--- src/hg/lib/bamFile.c	3 Nov 2009 00:33:45 -0000	1.12
+++ src/hg/lib/bamFile.c	5 Nov 2009 17:50:24 -0000	1.13
@@ -1,465 +1,494 @@
 /* bam -- interface to binary alignment format files using Heng Li's samtools lib. */
 
 #ifdef USE_BAM
 
 #include "common.h"
 #include "htmshell.h"
 #include "hdb.h"
 #include "hgConfig.h"
 #include "udc.h"
 #include "bamFile.h"
 
 static char const rcsid[] = "$Id$";
 
 static boolean ignoreStrand = FALSE;
 
 void bamIgnoreStrand()
 /* Change the behavior of this lib to disregard item strand. 
  * If called, this should be called before any other bam functions. */
 {
 ignoreStrand = TRUE;
 }
 
 char *bamFileNameFromTable(char *db, char *table, char *bamSeqName)
 /* Return file name from table.  If table has a seqName column, then grab the 
  * row associated with bamSeqName (which is not nec. in chromInfo, e.g. 
  * bam file might have '1' not 'chr1'). */
 {
 struct sqlConnection *conn = hAllocConn(db);
 boolean checkSeqName = (sqlFieldIndex(conn, table, "seqName") >= 0);
 if (checkSeqName && bamSeqName == NULL)
     errAbort("bamFileNameFromTable: table %s has seqName column, but NULL seqName passed in",
 	     table);
 char query[512];
 if (checkSeqName)
     safef(query, sizeof(query), "select fileName from %s where seqName = '%s'",
 	  table, bamSeqName);
 else
     safef(query, sizeof(query), "select fileName from %s", table);
 char *fileName = sqlQuickString(conn, query);
 if (fileName == NULL)
     {
     if (checkSeqName)
 	errAbort("Missing fileName for seqName '%s' in %s table", bamSeqName, table);
     else
 	errAbort("Missing fileName in %s table", table);
     }
 hFreeConn(&conn);
 return fileName;
 }
 
+static boolean isRegularFile(char *filename)
+/* File not only exists, but is also a file not a directory. */
+{
+struct stat mystat;
+if (stat(filename, &mystat) != 0)
+    return FALSE;
+return S_ISREG(mystat.st_mode);
+}
+
 static char *samtoolsFileName(char *fileOrUrl)
 /* If udcFuse is configured, and we have a URL, convert it into a filename in
  * the udcFuse filesystem for use by samtools.  Thus samtools will think it's
  * working on a local file, but udcFuse will pass access requests to udc and
  * we'll get the benefits of sparse-file local caching and https support.
  * udcFuse needs us to open udc files before invoking udcFuse paths, so open
  * both the .bam and .bai (index) URLs with udc here.  
  * If udcFuse is not configured, or fileOrUrl is not an URL, just pass through fileOrUrl. */
 {
 char *udcFuseRoot = cfgOption("udcFuse.mountPoint");
 char *protocol = NULL, *afterProtocol = NULL, *colon = NULL;
 udcParseUrl(fileOrUrl, &protocol, &afterProtocol, &colon);
 if (udcFuseRoot != NULL && afterProtocol != NULL)
     {
     struct dyString *dy = dyStringNew(0);
     dyStringPrintf(dy, "%s/%s/%s", udcFuseRoot, protocol, afterProtocol);
     char *bamFileName = dyStringCannibalize(&dy);
-    if (!fileExists(bamFileName))
+    if (!isRegularFile(bamFileName))
 	{
+	verbose(2, "going to call udcFileMayOpen(%s).\n", fileOrUrl);
 	struct udcFile *udcf = udcFileMayOpen(fileOrUrl, NULL);
 	if (udcf != NULL)
 	    {
 	    udcFileClose(&udcf);
-	    if (!fileExists(bamFileName))
+	    verbose(2, "closed udcf. testing existence of %s.\n", bamFileName);
+	    if (!isRegularFile(bamFileName))
 		{
 		warn("Cannot find %s -- remount udcFuse?", bamFileName);
 		freeMem(bamFileName);
 		return cloneString(fileOrUrl);
 		}
-	    // Look for index file: xxx.bam.bai or xxx.bai.
+	    }
+	else
+	    errAbort("Failed to open BAM URL \"%s\" with udc", fileOrUrl);
+	}
+    // Look for index file: xxx.bam.bai or xxx.bai.  Look for both in udcFuse,
+    // and only open the URL with udc if neither udcFuse file exists.
 	    int urlLen = strlen(fileOrUrl), fLen = strlen(bamFileName);
-	    char *indexUrl = needMem(urlLen+5), *indexFileName = needMem(fLen+5);
-	    safef(indexUrl, urlLen+5, "%s.bai", fileOrUrl);
+    char *indexFileName = needMem(fLen+5);
 	    safef(indexFileName, fLen+5, "%s.bai", bamFileName);
-	    udcf = udcFileMayOpen(indexUrl, NULL);
-	    if (udcf == NULL)
+    if (!isRegularFile(indexFileName))
 		{
+	verbose(2, "%s does not already exist\n", indexFileName);
+	char *altIndexFileName = NULL;
 		if (endsWith(fileOrUrl, ".bam"))
 		    {
-		    strcpy(indexUrl+urlLen-1, "i");
-		    strcpy(indexFileName+fLen-1, "i");
-		    udcf = udcFileMayOpen(indexUrl, NULL);
+	    altIndexFileName = cloneString(indexFileName);
+	    strcpy(altIndexFileName+fLen-1, "i");
+	    }
+	if (!(altIndexFileName && isRegularFile(altIndexFileName)))
+	    {
+	    char *indexUrl = needMem(urlLen+5);
+	    safef(indexUrl, urlLen+5, "%s.bai", fileOrUrl);
+	    verbose(2, "going to call udcFileMayOpen(%s).\n", indexUrl);
+	    struct udcFile *udcf = udcFileMayOpen(indexUrl, NULL);
+	    if (udcf != NULL)
+		udcFileClose(&udcf);
+	    else if (altIndexFileName != NULL)
+		{
+		char *altIndexUrl = cloneString(indexUrl);
+		strcpy(altIndexUrl+urlLen-1, "i");
+		verbose(2, "going to call udcFileMayOpen(%s).\n", altIndexUrl);
+		udcf = udcFileMayOpen(altIndexUrl, NULL);
 		    if (udcf == NULL)
-			errAbort("Cannot find BAM index file (%s.bai or %s)", fileOrUrl, indexUrl);
+		    errAbort("Cannot find BAM index file (%s or %s)", indexUrl, altIndexUrl);
 		    udcFileClose(&udcf);
+		freeMem(altIndexUrl);
 		    }
 		else
 		    errAbort("Cannot find BAM index file for \"%s\"", fileOrUrl);
-		}
 	    freeMem(indexUrl);
-	    freeMem(indexFileName);
 	    }
-	else
-	    errAbort("Failed to read BAM file \"%s\"", fileOrUrl);
+	freeMem(altIndexFileName);
 	}
+    freeMem(indexFileName);
     return bamFileName;
     }
 return cloneString(fileOrUrl);
 }
 
 boolean bamFileExists(char *fileOrUrl)
 /* Return TRUE if we can successfully open the bam file and its index file. */
 {
 char *bamFileName = samtoolsFileName(fileOrUrl);
 samfile_t *fh = samopen(bamFileName, "rb", NULL);
 if (fh != NULL)
     {
     // When file is an URL, this caches the index file in addition to validating:
     bam_index_t *idx = bam_index_load(bamFileName);
     samclose(fh);
     if (idx == NULL)
 	{
 	warn("bamFileExists: failed to read index corresponding to %s", bamFileName);
 	return FALSE;
 	}
     freeMem(idx);
     return TRUE;
     }
 return FALSE;
 }
 
 void bamFetch(char *fileOrUrl, char *position, bam_fetch_f callbackFunc, void *callbackData)
 /* Open the .bam file, fetch items in the seq:start-end position range,
  * and call callbackFunc on each bam item retrieved from the file plus callbackData. 
  * Note: if sequences in .bam file don't begin with "chr" but cart position does, pass in 
  * cart position + strlen("chr") to match the .bam file sequence names. */
 {
 char *bamFileName = samtoolsFileName(fileOrUrl);
 samfile_t *fh = samopen(bamFileName, "rb", NULL);
 if (fh == NULL)
     errAbort("samopen(%s, \"rb\") returned NULL", bamFileName);
 
 int chromId, start, end;
 int ret = bam_parse_region(fh->header, position, &chromId, &start, &end);
 if (ret != 0)
     // If the bam file does not cover the current chromosome, OK
     return;
 
 bam_index_t *idx = bam_index_load(bamFileName);
 if (idx == NULL)
     errAbort("bam_index_load(%s) failed.", bamFileName);
 ret = bam_fetch(fh->x.bam, idx, chromId, start, end, callbackData, callbackFunc);
 if (ret != 0)
     errAbort("bam_fetch(%s, %s (chromId=%d) failed (%d)", bamFileName, position, chromId, ret);
 samclose(fh);
 }
 
 boolean bamIsRc(const bam1_t *bam)
 /* Return TRUE if alignment is on - strand.  If bamIgnoreStrand has been called,
  * then this always returns FALSE. */
 {
 const bam1_core_t *core = &bam->core;
 return (core->flag & BAM_FREVERSE) && !ignoreStrand;
 }
 
 char *bamGetQuerySequence(const bam1_t *bam)
 /* Return the nucleotide sequence encoded in bam.  The BAM format 
  * reverse-complements query sequence when the alignment is on the - strand,
  * so here we rev-comp it back to restore the original query sequence. */
 {
 const bam1_core_t *core = &bam->core;
 char *qSeq = needMem(core->l_qseq + 1);
 uint8_t *s = bam1_seq(bam);
 int i;
 for (i = 0; i < core->l_qseq; i++)
     qSeq[i] = bam_nt16_rev_table[bam1_seqi(s, i)];
 if (bamIsRc(bam))
     reverseComplement(qSeq, core->l_qseq);
 return qSeq;
 }
 
 UBYTE *bamGetQueryQuals(const bam1_t *bam)
 /* Return the base quality scores encoded in bam as an array of ubytes. */
 {
 const bam1_core_t *core = &bam->core;
 int qLen = core->l_qseq;
 UBYTE *arr = needMem(qLen);
 boolean isRc = bamIsRc(bam);
 UBYTE *qualStr = bam1_qual(bam);
 int i;
 for (i = 0;  i < qLen;  i++)
     {
     int offset = isRc ? (qLen - 1 - i) : i;
     arr[offset] = (qualStr[0] == 255) ? 255 : qualStr[i];
     }
 return arr;
 }
 
 char *bamGetCigar(const bam1_t *bam)
 /* Return a BAM-enhanced CIGAR string, decoded from the packed encoding in bam. */
 {
 unsigned int *cigarPacked = bam1_cigar(bam);
 const bam1_core_t *core = &bam->core;
 struct dyString *dyCigar = dyStringNew(min(8, core->n_cigar*4));
 int i;
 for (i = 0;  i < core->n_cigar;  i++)
     {
     char op;
     int n = bamUnpackCigarElement(cigarPacked[i], &op);
     dyStringPrintf(dyCigar, "%d", n);
     dyStringAppendC(dyCigar, op);
     }
 return dyStringCannibalize(&dyCigar);
 }
 
 void bamShowCigarEnglish(const bam1_t *bam)
 /* Print out cigar in English e.g. "20 (mis)Match, 1 Deletion, 3 (mis)Match" */
 {
 unsigned int *cigarPacked = bam1_cigar(bam);
 const bam1_core_t *core = &bam->core;
 int i;
 for (i = 0;  i < core->n_cigar;  i++)
     {
     char op;
     int n = bamUnpackCigarElement(cigarPacked[i], &op);
     if (i > 0)
 	printf(", ");
     switch (op)
 	{
 	case 'M': // match or mismatch (gapless aligned block)
 	    printf("%d (mis)Match", n);
 	    break;
 	case 'I': // inserted in query
 	    printf("%d Insertion", n);
 	    break;
 	case 'S': // skipped query bases at beginning or end ("soft clipping")
 	    printf("%d Skipped", n);
 	    break;
 	case 'D': // deleted from query
 	    printf("%d Deletion", n);
 	    break;
 	case 'N': // long deletion from query (intron as opposed to small del)
 	    printf("%d deletioN", n);
 	    break;
 	case 'H': // skipped query bases not stored in record's query sequence ("hard clipping")
 	    printf("%d Hard clipped query", n);
 	    break;
 	case 'P': // P="silent deletion from padded reference sequence"
 	    printf("%d Padded / silent deletion", n);
 	    break;
 	default:
 	    errAbort("bamShowCigarEnglish: unrecognized CIGAR op %c -- update me", op);
 	}
     }
 }
 
 static void descFlag(unsigned flag, unsigned bitMask, char *desc, boolean makeRed,
 	      boolean *retFirst)
 /* Describe a flag bit (or multi-bit mask) if it is set in flag. */
 {
 if ((flag & bitMask) == bitMask) // *all* bits in bitMask are set in flag
     {
     if (!*retFirst)
 	printf(" | ");
     printf("<span%s>(<TT>0x%02x</TT>) %s</span>",
 	   (makeRed ? " style='color: red'" : ""), bitMask, desc);
     *retFirst = FALSE;
     }
 }
 
 void bamShowFlagsEnglish(const bam1_t *bam)
 /* Print out flags in English, e.g. "Mate is on '-' strand; Properly paired". */
 {
 const bam1_core_t *core = &bam->core;
 unsigned flag = core->flag;
 boolean first = TRUE;
 descFlag(flag, BAM_FDUP, "Optical or PCR duplicate", TRUE, &first);
 descFlag(flag, BAM_FQCFAIL, "QC failure", TRUE, &first);
 descFlag(flag, BAM_FSECONDARY, "Not primary alignment", TRUE, &first);
 descFlag(flag, BAM_FREAD2, "Read 2 of pair", FALSE, &first);
 descFlag(flag, BAM_FREAD1, "Read 1 of pair", FALSE, &first);
 descFlag(flag, BAM_FMREVERSE, "Mate is on '-' strand", FALSE, &first);
 descFlag(flag, BAM_FREVERSE, "Read is on '-' strand", FALSE, &first);
 descFlag(flag, BAM_FMUNMAP, "Mate is unmapped", TRUE, &first);
 if (flag & BAM_FUNMAP)
     errAbort("Read is unmapped (what is it doing here?!?)");
 descFlag(flag, (BAM_FPROPER_PAIR | BAM_FPAIRED), "Properly paired", FALSE, &first);
 if ((flag & BAM_FPAIRED) && !(flag & BAM_FPROPER_PAIR))
     descFlag(flag, BAM_FPAIRED, "Not properly paired", TRUE, &first);
 }
 
 int bamGetTargetLength(const bam1_t *bam)
 /* Tally up the alignment's length on the reference sequence from
  * bam's packed-int CIGAR representation. */
 {
 unsigned int *cigarPacked = bam1_cigar(bam);
 const bam1_core_t *core = &bam->core;
 int tLength=0;
 int i;
 for (i = 0;  i < core->n_cigar;  i++)
     {
     char op;
     int n = bamUnpackCigarElement(cigarPacked[i], &op);
     switch (op)
 	{
 	case 'M': // match or mismatch (gapless aligned block)
 	    tLength += n;
 	    break;
 	case 'I': // inserted in query
 	case 'S': // skipped query bases at beginning or end ("soft clipping")
 	    break;
 	case 'D': // deleted from query
 	case 'N': // long deletion from query (intron as opposed to small del)
 	    tLength += n;
 	    break;
 	case 'H': // skipped query bases not stored in record's query sequence ("hard clipping")
 	case 'P': // P="silent deletion from padded reference sequence" -- ignore these.
 	    break;
 	default:
 	    errAbort("bamGetTargetLength: unrecognized CIGAR op %c -- update me", op);
 	}
     }
 return tLength;
 }
 
 struct ffAli *bamToFfAli(const bam1_t *bam, struct dnaSeq *target, int targetOffset)
 /* Convert from bam to ffAli format.  (Adapted from psl.c's pslToFfAli.) */
 {
 struct ffAli *ffList = NULL, *ff;
 const bam1_core_t *core = &bam->core;
 boolean isRc = bamIsRc(bam);
 DNA *needle = (DNA *)bamGetQuerySequence(bam);
 if (isRc)
     reverseComplement(target->dna, target->size);
 DNA *haystack = target->dna;
 unsigned int *cigarPacked = bam1_cigar(bam);
 int tStart = targetOffset, qStart = 0, i;
 // If isRc, need to go through the CIGAR ops backwards, but sequence offsets still count up.
 int iStart = isRc ? (core->n_cigar - 1) : 0;
 int iIncr = isRc ? -1 : 1;
 for (i = iStart;  isRc ? (i >= 0) : (i < core->n_cigar);  i += iIncr)
     {
     char op;
     int size = bamUnpackCigarElement(cigarPacked[i], &op);
     switch (op)
 	{
 	case 'M': // match or mismatch (gapless aligned block)
 	    AllocVar(ff);
 	    ff->left = ffList;
 	    ffList = ff;
 	    ff->nStart = needle + qStart;
 	    ff->nEnd = ff->nStart + size;
 	    ff->hStart = haystack + tStart - targetOffset;
 	    ff->hEnd = ff->hStart + size;
 	    tStart += size;
 	    qStart += size;
 	    break;
 	case 'I': // inserted in query
 	case 'S': // skipped query bases at beginning or end ("soft clipping")
 	    qStart += size;
 	    break;
 	case 'D': // deleted from query
 	case 'N': // long deletion from query (intron as opposed to small del)
 	    tStart += size;
 	    break;
 	case 'H': // skipped query bases not stored in record's query sequence ("hard clipping")
 	case 'P': // P="silent deletion from padded reference sequence" -- ignore these.
 	    break;
 	default:
 	    errAbort("bamToFfAli: unrecognized CIGAR op %c -- update me", op);
 	}
     }
 ffList = ffMakeRightLinks(ffList);
 ffCountGoodEnds(ffList);
 return ffList;
 }
 
 bam1_t *bamClone(const bam1_t *bam)
 /* Return a newly allocated copy of bam. */
 {
 // Using typecasts to get around compiler complaints about bam being const:
 bam1_t *newBam = cloneMem((void *)bam, sizeof(*bam));
 newBam->data = cloneMem((void *)bam->data, bam->data_len*sizeof(bam->data[0]));
 return newBam;
 }
 
 void bamShowTags(const bam1_t *bam)
 /* Print out tags in HTML: bold key, no type indicator for brevity. */
 {
 // adapted from part of bam.c bam_format1:
 uint8_t *s = bam1_aux(bam);
 while (s < bam->data + bam->data_len)
     {
     uint8_t type, key[2];
     key[0] = s[0]; key[1] = s[1];
     s += 2; type = *s; ++s;
     printf(" <B>%c%c</B>:", key[0], key[1]);
     if (type == 'A') { printf("%c", *s); ++s; }
     else if (type == 'C') { printf("%u", *s); ++s; }
     else if (type == 'c') { printf("%d", *s); ++s; }
     else if (type == 'S') { printf("%u", *(uint16_t*)s); s += 2; }
     else if (type == 's') { printf("%d", *(int16_t*)s); s += 2; }
     else if (type == 'I') { printf("%u", *(uint32_t*)s); s += 4; }
     else if (type == 'i') { printf("%d", *(int32_t*)s); s += 4; }
     else if (type == 'f') { printf("%g", *(float*)s); s += 4; }
     else if (type == 'd') { printf("%lg", *(double*)s); s += 8; }
     else if (type == 'Z' || type == 'H')
 	{
 	while (*s) putc(*s++, stdout);
 	++s;
 	}
     }
 putc('\n', stdout);
 }
 
 char *bamGetTagString(const bam1_t *bam, char *tag, char *buf, size_t bufSize)
 /* If bam's tags include the given 2-character tag, place the value into 
  * buf (zero-terminated, trunc'd if nec) and return a pointer to buf,
  * or NULL if tag is not present. */
 {
 if (tag == NULL)
     errAbort("NULL tag passed to bamGetTagString");
 if (! (isalpha(tag[0]) && isalnum(tag[1]) && tag[2] == '\0'))
     errAbort("bamGetTagString: invalid tag '%s'", htmlEncode(tag));
 char *val = NULL;
 // adapted from part of bam.c bam_format1:
 uint8_t *s = bam1_aux(bam);
 while (s < bam->data + bam->data_len)
     {
     uint8_t type, key[2];
     key[0] = s[0]; key[1] = s[1];
     s += 2; type = *s; ++s;
     if (key[0] == tag[0] && key[1] == tag[1])
 	{
 	if (type == 'A') { snprintf(buf, bufSize, "%c", *s);}
 	else if (type == 'C') { snprintf(buf, bufSize, "%u", *s); }
 	else if (type == 'c') { snprintf(buf, bufSize, "%d", *s); }
 	else if (type == 'S') { snprintf(buf, bufSize, "%u", *(uint16_t*)s); }
 	else if (type == 's') { snprintf(buf, bufSize, "%d", *(int16_t*)s); }
 	else if (type == 'I') { snprintf(buf, bufSize, "%u", *(uint32_t*)s); }
 	else if (type == 'i') { snprintf(buf, bufSize, "%d", *(int32_t*)s); }
 	else if (type == 'f') { snprintf(buf, bufSize, "%g", *(float*)s); }
 	else if (type == 'd') { snprintf(buf, bufSize, "%lg", *(double*)s); }
 	else if (type == 'Z' || type == 'H') strncpy(buf, (char *)s, bufSize);
 	else buf[0] = '\0';
 	buf[bufSize-1] = '\0'; // TODO: is this nec?? see man pages
 	val = buf;
 	break;
 	}
     else
 	{
 	if (type == 'A' || type == 'C' || type == 'c') { ++s; }
 	else if (type == 'S' || type == 's') { s += 2; }
 	else if (type == 'I' || type == 'i' || type == 'f') { s += 4; }
 	else if (type == 'd') { s += 8; }
 	else if (type == 'Z' || type == 'H')
 	    {
 	    while (*s++);
 	    }
 	}
     }
 return val;
 }
 
 #endif//def USE_BAM