/**
* Reads past the next ambiguous or unambiguous stretch of sequence
* from the given FASTA file and returns its length. Does not do
* anything with the sequence characters themselves; this is purely for
* measuring lengths.
*/
RefRecord fastaRefReadSize(FileBuf& in,
const RefReadInParams& rparms,
bool first,
BitpairOutFileBuf* bpout)
{
int c;
static int lastc = '>'; // last character seen
// RefRecord params
size_t len = 0; // 'len' counts toward total length
// 'off' counts number of ambiguous characters before first
// unambiguous character
size_t off = 0;
// Pick off the first carat and any preceding whitespace
if(first) {
assert(!in.eof());
lastc = '>';
c = in.getPastWhitespace();
if(in.eof()) {
// Got eof right away; emit warning
cerr << "Warning: Empty input file" << endl;
lastc = -1;
return RefRecord(0, 0, true);
}
assert(c == '>');
}
first = true;
// Skip to the end of the id line; if the next line is either
// another id line or a comment line, keep skipping
if(lastc == '>') {
// Skip to the end of the name line
do {
if((c = in.getPastNewline()) == -1) {
// No more input
cerr << "Warning: Encountered empty reference sequence" << endl;
lastc = -1;
return RefRecord(0, 0, true);
}
if(c == '>') {
cerr << "Warning: Encountered empty reference sequence" << endl;
}
// continue until a non-name, non-comment line
} while (c == '>');
} else {
first = false; // not the first in a sequence
off = 1; // The gap has already been consumed, so count it
if((c = in.get()) == -1) {
// Don't emit a warning, since this might legitimately be
// a gap on the end of the final sequence in the file
lastc = -1;
return RefRecord(off, len, first);
}
}
// Now skip to the first DNA character, counting gap characters
// as we go
int lc = -1; // last-DNA char variable for color conversion
while(true) {
int cat = dna4Cat[c];
if(rparms.nsToAs && cat == 2) c = 'A';
if(cat == 1) {
// This is a DNA character
if(rparms.color) {
if(lc != -1) {
// Got two consecutive unambiguous DNAs
break; // to read-in loop
}
// Keep going; we need two consecutive unambiguous DNAs
lc = charToDna5[(int)c];
// The 'if(off > 0)' takes care of the case where
// the reference is entirely unambiguous and we don't
// want to incorrectly increment off.
if(off > 0) off++;
} else {
break; // to read-in loop
}
} else if(cat == 2) {
if(lc != -1 && off == 0) off++;
lc = -1;
off++; // skip over gap character and increment
} else if(c == '>') {
if(off > 0 && lastc == '>') {
cerr << "Warning: Encountered reference sequence with only gaps" << endl;
} else if(lastc == '>') {
cerr << "Warning: Encountered empty reference sequence" << endl;
}
lastc = '>';
return RefRecord(off, 0, first);
}
c = in.get();
if(c == -1) {
// End-of-file
if(off > 0 && lastc == '>') {
cerr << "Warning: Encountered reference sequence with only gaps" << endl;
} else if(lastc == '>') {
cerr << "Warning: Encountered empty reference sequence" << endl;
}
lastc = -1;
return RefRecord(off, 0, first);
}
}
assert(!rparms.color || (lc != -1));
assert_eq(1, dna4Cat[c]); // C must be unambiguous base
if(off > 0 && rparms.color && first) {
// Handle the case where the first record has ambiguous
// characters but we're in color space; one of those counts is
// spurious
off--;
}
// in now points just past the first character of a sequence
// line, and c holds the first character
while(c != -1 && c != '>') {
if(rparms.nsToAs && dna4Cat[c] == 2) c = 'A';
uint8_t cat = dna4Cat[c];
int cc = toupper(c);
if(rparms.bisulfite && cc == 'C') c = cc = 'T';
if(cat == 1) {
// It's a DNA character
assert(cc == 'A' || cc == 'C' || cc == 'G' || cc == 'T');
// Consume it
len++;
// Output it
if(bpout != NULL) {
if(rparms.color) {
// output color
bpout->write(dinuc2color[charToDna5[(int)c]][lc]);
} else if(!rparms.color) {
// output nucleotide
bpout->write(charToDna5[c]);
}
}
lc = charToDna5[(int)c];
} else if(cat == 2) {
// It's an N or a gap
lastc = c;
assert(cc != 'A' && cc != 'C' && cc != 'G' && cc != 'T');
return RefRecord(off, len, first);
} else {
// Not DNA and not a gap, ignore it
#ifndef NDEBUG
if(!isspace(c)) {
cerr << "Unexpected character in sequence: ";
if(isprint(c)) {
cerr << ((char)c) << endl;
} else {
cerr << "(" << c << ")" << endl;
}
}
#endif
}
c = in.get();
}
lastc = c;
return RefRecord(off, len, first);
}
static void driver(
const string& infile,
EList<string>& infiles,
const string& snpfile,
const string& htfile,
const string& ssfile,
const string& exonfile,
const string& svfile,
const string& outfile,
bool packed,
int reverse)
{
initializeCntLut();
initializeCntBit();
EList<FileBuf*> is(MISC_CAT);
bool bisulfite = false;
RefReadInParams refparams(false, reverse, nsToAs, bisulfite);
assert_gt(infiles.size(), 0);
if(format == CMDLINE) {
// Adapt sequence strings to stringstreams open for input
stringstream *ss = new stringstream();
for(size_t i = 0; i < infiles.size(); i++) {
(*ss) << ">" << i << endl << infiles[i].c_str() << endl;
}
FileBuf *fb = new FileBuf(ss);
assert(fb != NULL);
assert(!fb->eof());
assert(fb->get() == '>');
ASSERT_ONLY(fb->reset());
assert(!fb->eof());
is.push_back(fb);
} else {
// Adapt sequence files to ifstreams
for(size_t i = 0; i < infiles.size(); i++) {
FILE *f = fopen(infiles[i].c_str(), "r");
if (f == NULL) {
cerr << "Error: could not open "<< infiles[i].c_str() << endl;
throw 1;
}
FileBuf *fb = new FileBuf(f);
assert(fb != NULL);
if(fb->peek() == -1 || fb->eof()) {
cerr << "Warning: Empty fasta file: '" << infile.c_str() << "'" << endl;
continue;
}
assert(!fb->eof());
assert(fb->get() == '>');
ASSERT_ONLY(fb->reset());
assert(!fb->eof());
is.push_back(fb);
}
}
if(is.empty()) {
cerr << "Warning: All fasta inputs were empty" << endl;
throw 1;
}
filesWritten.push_back(outfile + ".1." + gfm_ext);
filesWritten.push_back(outfile + ".2." + gfm_ext);
// Vector for the ordered list of "records" comprising the input
// sequences. A record represents a stretch of unambiguous
// characters in one of the input sequences.
EList<RefRecord> szs(MISC_CAT);
std::pair<size_t, size_t> sztot;
{
if(verbose) cerr << "Reading reference sizes" << endl;
Timer _t(cerr, " Time reading reference sizes: ", verbose);
if(!reverse && (writeRef || justRef)) {
filesWritten.push_back(outfile + ".3." + gfm_ext);
filesWritten.push_back(outfile + ".4." + gfm_ext);
sztot = BitPairReference::szsFromFasta(is, outfile, bigEndian, refparams, szs, sanityCheck);
} else {
sztot = BitPairReference::szsFromFasta(is, string(), bigEndian, refparams, szs, sanityCheck);
}
}
if(justRef) return;
assert_gt(sztot.first, 0);
assert_gt(sztot.second, 0);
assert_gt(szs.size(), 0);
// Construct index from input strings and parameters
filesWritten.push_back(outfile + ".5." + gfm_ext);
filesWritten.push_back(outfile + ".6." + gfm_ext);
filesWritten.push_back(outfile + ".7." + gfm_ext);
filesWritten.push_back(outfile + ".8." + gfm_ext);
TStr s;
HGFM<TIndexOffU> hGFM(
s,
packed,
1, // TODO: maybe not?
lineRate,
offRate, // suffix-array sampling rate
ftabChars, // number of chars in initial arrow-pair calc
localOffRate,
localFtabChars,
nthreads,
snpfile,
htfile,
ssfile,
exonfile,
svfile,
outfile, // basename for .?.ht2 files
reverse == 0, // fw
!entireSA, // useBlockwise
bmax, // block size for blockwise SA builder
bmaxMultSqrt, // block size as multiplier of sqrt(len)
bmaxDivN, // block size as divisor of len
noDc? 0 : dcv,// difference-cover period
is, // list of input streams
szs, // list of reference sizes
(TIndexOffU)sztot.first, // total size of all unambiguous ref chars
refparams, // reference read-in parameters
seed, // pseudo-random number generator seed
-1, // override offRate
verbose, // be talkative
autoMem, // pass exceptions up to the toplevel so that we can adjust memory settings automatically
sanityCheck); // verify results and internal consistency
// Note that the Ebwt is *not* resident in memory at this time. To
// load it into memory, call ebwt.loadIntoMemory()
if(verbose) {
// Print Ebwt's vital stats
hGFM.gh().print(cerr);
}
if(sanityCheck) {
// Try restoring the original string (if there were
// multiple texts, what we'll get back is the joined,
// padded string, not a list)
hGFM.loadIntoMemory(
reverse ? (refparams.reverse == REF_READ_REVERSE) : 0,
true, // load SA sample?
true, // load ftab?
true, // load rstarts?
false,
false);
SString<char> s2;
hGFM.restore(s2);
hGFM.evictFromMemory();
{
SString<char> joinedss = GFM<>::join<SString<char> >(
is, // list of input streams
szs, // list of reference sizes
(TIndexOffU)sztot.first, // total size of all unambiguous ref chars
refparams, // reference read-in parameters
seed); // pseudo-random number generator seed
if(refparams.reverse == REF_READ_REVERSE) {
joinedss.reverse();
}
assert_eq(joinedss.length(), s2.length());
assert(sstr_eq(joinedss, s2));
}
if(verbose) {
if(s2.length() < 1000) {
cout << "Passed restore check: " << s2.toZBuf() << endl;
} else {
cout << "Passed restore check: (" << s2.length() << " chars)" << endl;
}
}
}
}
/**
* Parse a single quality string from fb and store qualities in r.
* Assume the next character obtained via fb.get() is the first
* character of the quality string. When returning, the next
* character returned by fb.peek() or fb.get() should be the first
* character of the following line.
*/
int parseQuals(
Read& r,
FileBuf& fb,
int firstc,
int readLen,
int trim3,
int trim5,
bool intQuals,
bool phred64,
bool solexa64)
{
int c = firstc;
assert(c != '\n' && c != '\r');
r.qual.clear();
if (intQuals) {
while (c != '\r' && c != '\n' && c != -1) {
bool neg = false;
int num = 0;
while(!isspace(c) && !fb.eof()) {
if(c == '-') {
neg = true;
assert_eq(num, 0);
} else {
if(!isdigit(c)) {
char buf[2048];
cerr << "Warning: could not parse quality line:" << endl;
fb.getPastNewline();
cerr << fb.copyLastN(buf);
buf[2047] = '\0';
cerr << buf;
throw 1;
}
assert(isdigit(c));
num *= 10;
num += (c - '0');
}
c = fb.get();
}
if(neg) num = 0;
// Phred-33 ASCII encode it and add it to the back of the
// quality string
r.qual.append('!' + num);
// Skip over next stretch of whitespace
while(c != '\r' && c != '\n' && isspace(c) && !fb.eof()) {
c = fb.get();
}
}
} else {
while (c != '\r' && c != '\n' && c != -1) {
r.qual.append(charToPhred33(c, solexa64, phred64));
c = fb.get();
while(c != '\r' && c != '\n' && isspace(c) && !fb.eof()) {
c = fb.get();
}
}
}
if ((int)r.qual.length() < readLen-1 ||
((int)r.qual.length() < readLen && !r.color))
{
tooFewQualities(r.name);
}
r.qual.trimEnd(trim3);
if(r.qual.length()-trim5 < r.patFw.length()) {
assert(gColor && r.primer != -1);
assert_gt(trim5, 0);
trim5--;
}
r.qual.trimBegin(trim5);
if(r.qual.length() <= 0) return 0;
assert_eq(r.qual.length(), r.patFw.length());
while(fb.peek() == '\n' || fb.peek() == '\r') fb.get();
return (int)r.qual.length();
}
static void driver(const string& infile,
vector<string>& infiles,
const string& outfile,
bool reverse = false)
{
vector<FileBuf*> is;
bool bisulfite = false;
RefReadInParams refparams(color, reverse ? reverseType : REF_READ_FORWARD, nsToAs, bisulfite);
assert_gt(infiles.size(), 0);
if(format == CMDLINE) {
// Adapt sequence strings to stringstreams open for input
stringstream *ss = new stringstream();
for(size_t i = 0; i < infiles.size(); i++) {
(*ss) << ">" << i << endl << infiles[i] << endl;
}
FileBuf *fb = new FileBuf(ss);
assert(fb != NULL);
assert(!fb->eof());
assert(fb->get() == '>');
ASSERT_ONLY(fb->reset());
assert(!fb->eof());
is.push_back(fb);
} else {
// Adapt sequence files to ifstreams
for(size_t i = 0; i < infiles.size(); i++) {
FILE *f = fopen(infiles[i].c_str(), "rb");
if (f == NULL) {
cerr << "Error: could not open "<< infiles[i] << endl;
throw 1;
}
FileBuf *fb = new FileBuf(f);
assert(fb != NULL);
assert(!fb->eof());
assert(fb->get() == '>');
ASSERT_ONLY(fb->reset());
assert(!fb->eof());
is.push_back(fb);
}
}
// Vector for the ordered list of "records" comprising the input
// sequences. A record represents a stretch of unambiguous
// characters in one of the input sequences.
vector<RefRecord> szs;
vector<uint32_t> plens;
std::pair<size_t, size_t> sztot;
{
if(verbose) cout << "Reading reference sizes" << endl;
Timer _t(cout, " Time reading reference sizes: ", verbose);
if(!reverse && (writeRef || justRef)) {
// For forward reference, dump it to .3.ebwt and .4.ebwt
// files
string file3 = outfile + ".3." + gEbwt_ext;
string file4 = outfile + ".4." + gEbwt_ext;
// Open output stream for the '.3.ebwt' file which will
// hold the size records.
ofstream fout3(file3.c_str(), ios::binary);
if(!fout3.good()) {
cerr << "Could not open index file for writing: \"" << file3 << "\"" << endl
<< "Please make sure the directory exists and that permissions allow writing by" << endl
<< "Bowtie." << endl;
throw 1;
}
BitpairOutFileBuf bpout(file4.c_str());
// Read in the sizes of all the unambiguous stretches of
// the genome into a vector of RefRecords. The input
// streams are reset once it's done.
writeU<int32_t>(fout3, 1, bigEndian); // endianness sentinel
if(color) {
refparams.color = false;
// Make sure the .3.ebwt and .4.ebwt files contain
// nucleotides; not colors
TIndexOff numSeqs = 0;
fastaRefReadSizes(is, szs, plens, refparams, &bpout, numSeqs);
refparams.color = true;
writeU<TIndexOffU>(fout3, (TIndexOffU)szs.size(), bigEndian); // write # records
for(size_t i = 0; i < szs.size(); i++) {
szs[i].write(fout3, bigEndian);
}
szs.clear();
plens.clear();
// Now read in the colorspace size records; these are
// the ones that were indexed
TIndexOff numSeqs2 = 0;
sztot = fastaRefReadSizes(is, szs, plens, refparams, NULL, numSeqs2);
assert_geq(numSeqs, numSeqs2);
} else {
TIndexOff numSeqs = 0;
sztot = fastaRefReadSizes(is, szs, plens, refparams, &bpout, numSeqs);
writeU<TIndexOffU>(fout3, (TIndexOffU)szs.size(), bigEndian); // write # records
for(size_t i = 0; i < szs.size(); i++) szs[i].write(fout3, bigEndian);
}
if(sztot.first == 0) {
cerr << "Error: No unambiguous stretches of characters in the input. Aborting..." << endl;
throw 1;
}
assert_gt(sztot.first, 0);
assert_gt(sztot.second, 0);
bpout.close();
fout3.close();
#ifndef NDEBUG
if(sanityCheck) {
BitPairReference bpr(
outfile, // ebwt basename
color, // expect color?
true, // sanity check?
&infiles,// files to check against
NULL, // sequences to check against
format == CMDLINE, // whether infiles contains strings
true, // load sequence?
false, // use memory-mapped files
false, // use shared memory
false, // sweep through memory-mapped memory
false, // be talkative
false); // be talkative
}
#endif
} else {
// Read in the sizes of all the unambiguous stretches of the
// genome into a vector of RefRecords
TIndexOff numSeqs = 0;
sztot = fastaRefReadSizes(is, szs, plens, refparams, NULL, numSeqs);
#ifndef NDEBUG
if(refparams.color) {
refparams.color = false;
vector<RefRecord> szs2;
vector<uint32_t> plens2;
TIndexOff numSeqs2 = 0;
fastaRefReadSizes(is, szs2, plens2, refparams, NULL, numSeqs2);
assert_leq(numSeqs, numSeqs2);
// One less color than base
refparams.color = true;
}
#endif
}
}
if(justRef) return;
assert_gt(sztot.first, 0);
assert_gt(sztot.second, 0);
assert_gt(szs.size(), 0);
// Construct Ebwt from input strings and parameters
Ebwt<TStr> ebwt(refparams.color ? 1 : 0,
lineRate,
linesPerSide,
offRate, // suffix-array sampling rate
-1, // ISA sampling rate
ftabChars, // number of chars in initial arrow-pair calc
nthreads,
outfile, // basename for .?.ebwt files
!reverse, // fw
!entireSA, // useBlockwise
bmax, // block size for blockwise SA builder
bmaxMultSqrt, // block size as multiplier of sqrt(len)
bmaxDivN, // block size as divisor of len
noDc? 0 : dcv,// difference-cover period
is, // list of input streams
szs, // list of reference sizes
plens, // list of not-all-gap reference sequence lengths
(TIndexOffU)sztot.first, // total size of all unambiguous ref chars
refparams, // reference read-in parameters
seed, // pseudo-random number generator seed
-1, // override offRate
-1, // override isaRate
verbose, // be talkative
autoMem, // pass exceptions up to the toplevel so that we can adjust memory settings automatically
sanityCheck); // verify results and internal consistency
// Note that the Ebwt is *not* resident in memory at this time. To
// load it into memory, call ebwt.loadIntoMemory()
if(verbose) {
// Print Ebwt's vital stats
ebwt.eh().print(cout);
}
if(sanityCheck) {
// Try restoring the original string (if there were
// multiple texts, what we'll get back is the joined,
// padded string, not a list)
ebwt.loadIntoMemory(
refparams.color ? 1 : 0,
-1,
false,
false);
TStr s2; ebwt.restore(s2);
ebwt.evictFromMemory();
{
TStr joinedss = Ebwt<TStr>::join(
is, // list of input streams
szs, // list of reference sizes
(TIndexOffU)sztot.first, // total size of all unambiguous ref chars
refparams, // reference read-in parameters
seed); // pseudo-random number generator seed
if(refparams.reverse == REF_READ_REVERSE) {
reverseInPlace(joinedss);
}
assert_eq(length(joinedss), length(s2));
assert_eq(joinedss, s2);
}
if(verbose) {
if(length(s2) < 1000) {
cout << "Passed restore check: " << s2 << endl;
} else {
cout << "Passed restore check: (" << length(s2) << " chars)" << endl;
}
}
}
}
