static void parseColumnHeaderRow(struct vcfFile *vcff, char *line)
/* Make sure column names are as we expect, and store genotype sample IDs if any are given. */
{
if (line[0] != '#')
{
vcfFileErr(vcff, "Expected to find # followed by column names (\"#CHROM POS ...\"), "
"not \"%s\"", line);
lineFileReuse(vcff->lf);
return;
}
char *words[VCF_MAX_COLUMNS];
int wordCount = chopLine(line+1, words);
if (wordCount >= VCF_MAX_COLUMNS)
vcfFileErr(vcff, "header contains at least %d columns; "
"VCF_MAX_COLUMNS may need to be increased in vcf.c!", VCF_MAX_COLUMNS);
expectColumnName(vcff, "CHROM", words, 0);
expectColumnName(vcff, "POS", words, 1);
expectColumnName(vcff, "ID", words, 2);
expectColumnName(vcff, "REF", words, 3);
expectColumnName(vcff, "ALT", words, 4);
expectColumnName2(vcff, "QUAL", "PROB", words, 5);
expectColumnName(vcff, "FILTER", words, 6);
expectColumnName(vcff, "INFO", words, 7);
if (wordCount > 8)
{
expectColumnName(vcff, "FORMAT", words, 8);
if (wordCount < 10)
vcfFileErr(vcff, "FORMAT column is given, but no sample IDs for genotype columns...?");
vcff->genotypeCount = (wordCount - 9);
vcff->genotypeIds = vcfFileAlloc(vcff, vcff->genotypeCount * sizeof(char *));
int i;
for (i = 9; i < wordCount; i++)
vcff->genotypeIds[i-9] = vcfFileCloneStr(vcff, words[i]);
}
}
struct vcfRecord *vcfRecordFromRow(struct vcfFile *vcff, char **words)
/* Parse words from a VCF data line into a VCF record structure. */
{
struct vcfRecord *record = vcfFileAlloc(vcff, sizeof(struct vcfRecord));
record->file = vcff;
record->chrom = vcfFilePooledStr(vcff, words[0]);
record->chromStart = lineFileNeedNum(vcff->lf, words, 1) - 1;
// chromEnd may be overwritten by parseRefAndAlt and parseInfoColumn.
record->chromEnd = record->chromStart+1;
record->name = vcfFilePooledStr(vcff, words[2]);
parseRefAndAlt(vcff, record, words[3], words[4]);
record->qual = vcfFilePooledStr(vcff, words[5]);
parseFilterColumn(vcff, record, words[6]);
parseInfoColumn(vcff, record, words[7]);
if (vcff->genotypeCount > 0)
{
record->format = vcfFilePooledStr(vcff, words[8]);
record->genotypeUnparsedStrings = vcfFileAlloc(vcff,
vcff->genotypeCount * sizeof(char *));
int i;
// Don't bother actually parsing all these until & unless we need the info:
for (i = 0; i < vcff->genotypeCount; i++)
record->genotypeUnparsedStrings[i] = vcfFileCloneStr(vcff, words[9+i]);
}
return record;
}
static struct vcfFile *vcfFileHeaderFromLineFile(struct lineFile *lf, int maxErr)
/* Parse a VCF file into a vcfFile object. If maxErr not zero, then
* continue to parse until this number of error have been reached. A maxErr
* less than zero does not stop and reports all errors.
* Set maxErr to VCF_IGNORE_ERRS for silence */
{
initVcfSpecInfoDefs();
initVcfSpecGtFormatDefs();
if (lf == NULL)
return NULL;
struct vcfFile *vcff = vcfFileNew();
vcff->lf = lf;
vcff->fileOrUrl = vcfFileCloneStr(vcff, lf->fileName);
vcff->maxErr = (maxErr < 0) ? INT_MAX : maxErr;
struct dyString *dyHeader = dyStringNew(1024);
char *line = NULL;
// First, metadata lines beginning with "##":
while (lineFileNext(lf, &line, NULL) && startsWith("##", line))
{
dyStringAppend(dyHeader, line);
dyStringAppendC(dyHeader, '\n');
parseMetadataLine(vcff, line);
}
slReverse(&(vcff->infoDefs));
slReverse(&(vcff->filterDefs));
slReverse(&(vcff->gtFormatDefs));
// Did we get the bare minimum VCF header with supported version?
if (vcff->majorVersion == 0)
vcfFileErr(vcff, "missing ##fileformat= header line? Assuming 4.1.");
if ((vcff->majorVersion != 4 || (vcff->minorVersion != 0 && vcff->minorVersion != 1)) &&
(vcff->majorVersion != 3))
vcfFileErr(vcff, "VCFv%d.%d not supported -- only v3.*, v4.0 or v4.1",
vcff->majorVersion, vcff->minorVersion);
// Next, one header line beginning with single "#" that names the columns:
if (line == NULL)
// EOF after metadata
return vcff;
dyStringAppend(dyHeader, line);
dyStringAppendC(dyHeader, '\n');
parseColumnHeaderRow(vcff, line);
vcff->headerString = dyStringCannibalize(&dyHeader);
return vcff;
}
static void vcfParseData(struct vcfFile *vcff, int maxRecords)
/* Given a vcfFile into which the header has been parsed, and whose lineFile is positioned
* at the beginning of a data row, parse and store all data rows from lineFile. */
{
if (vcff == NULL)
return;
int recCount = 0, expected = 8;
if (vcff->genotypeCount > 0)
expected = 9 + vcff->genotypeCount;
char *words[VCF_MAX_COLUMNS];
int wordCount;
while ((wordCount = lineFileChop(vcff->lf, words)) > 0)
{
if (maxRecords >= 0 && recCount >= maxRecords)
break;
lineFileExpectWords(vcff->lf, expected, wordCount);
struct vcfRecord *record;
AllocVar(record);
record->file = vcff;
record->chrom = vcfFilePooledStr(vcff, words[0]);
record->chromStart = lineFileNeedNum(vcff->lf, words, 1) - 1;
// chromEnd may be overwritten by parseRefAndAlt and parseInfoColumn.
record->chromEnd = record->chromStart+1;
record->name = vcfFilePooledStr(vcff, words[2]);
parseRefAndAlt(vcff, record, words[3], words[4]);
record->qual = vcfFilePooledStr(vcff, words[5]);
parseFilterColumn(vcff, record, words[6]);
parseInfoColumn(vcff, record, words[7]);
if (vcff->genotypeCount > 0)
{
record->format = vcfFilePooledStr(vcff, words[8]);
record->genotypeUnparsedStrings = vcfFileAlloc(vcff,
vcff->genotypeCount * sizeof(char *));
int i;
// Don't bother actually parsing all these until & unless we need the info:
for (i = 0; i < vcff->genotypeCount; i++)
record->genotypeUnparsedStrings[i] = vcfFileCloneStr(vcff, words[9+i]);
}
slAddHead(&(vcff->records), record);
recCount++;
}
slReverse(&(vcff->records));
lineFileClose(&(vcff->lf));
}
