/**
* Check that the ebwt array is internally consistent up to (and not
* including) the given side index by re-counting the chars and
* comparing against the embedded occ[] arrays.
*/
void Ebwt::sanityCheckUpToSide(int upToSide) const {
assert(isInMemory());
uint32_t occ[] = {0, 0, 0, 0};
ASSERT_ONLY(uint32_t occ_save[] = {0, 0, 0, 0});
uint32_t cur = 0; // byte pointer
const EbwtParams& eh = this->_eh;
bool fw = false;
while(cur < (upToSide * eh._sideSz)) {
assert_leq(cur + eh._sideSz, eh._ebwtTotLen);
for(uint32_t i = 0; i < eh._sideBwtSz; i++) {
uint8_t by = this->ebwt()[cur + (fw ? i : eh._sideBwtSz-i-1)];
for(int j = 0; j < 4; j++) {
// Unpack from lowest to highest bit pair
int twoBit = unpack_2b_from_8b(by, fw ? j : 3-j);
occ[twoBit]++;
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) % 4);
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) % eh._sideBwtLen);
// Finished forward bucket; check saved [A], [C], [G] and [T]
// against the uint32_ts encoded here
ASSERT_ONLY(const uint32_t *u32ebwt = reinterpret_cast<const uint32_t*>(&ebwt()[cur + eh._sideBwtSz]));
ASSERT_ONLY(uint32_t as = u32ebwt[0]);
ASSERT_ONLY(uint32_t cs = u32ebwt[1]);
ASSERT_ONLY(uint32_t gs = u32ebwt[2]);
ASSERT_ONLY(uint32_t ts = u32ebwt[3]);
assert(as == occ_save[0] || as == occ_save[0]-1);
assert_eq(cs, occ_save[1]);
assert_eq(gs, occ_save[2]);
assert_eq(ts, occ_save[3]);
#ifndef NDEBUG
occ_save[0] = occ[0];
occ_save[1] = occ[1];
occ_save[2] = occ[2];
occ_save[3] = occ[3];
#endif
cur += eh._sideSz;
}
}
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;
}
}
}
}
static void driver(
const string& infile,
EList<string>& infiles,
const string& outfile,
bool packed,
int reverse)
{
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(), "rb");
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;
}
if(!reverse) {
#ifdef BOWTIE_64BIT_INDEX
cerr << "Building a LARGE index" << endl;
#else
cerr << "Building a SMALL index" << endl;
#endif
}
// 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) cout << "Reading reference sizes" << endl;
Timer _t(cout, " Time reading reference sizes: ", verbose);
if(!reverse && (writeRef || justRef)) {
filesWritten.push_back(outfile + ".3." + gEbwt_ext);
filesWritten.push_back(outfile + ".4." + gEbwt_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 + ".1." + gEbwt_ext);
filesWritten.push_back(outfile + ".2." + gEbwt_ext);
Ebwt ebwt(
TStr(),
packed,
0,
1, // TODO: maybe not?
lineRate,
offRate, // suffix-array sampling rate
ftabChars, // number of chars in initial arrow-pair calc
nthreads,
outfile, // basename for .?.ebwt 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
doSaFile, // make a file with just the suffix array in it
doBwtFile, // make a file with just the BWT string in it
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(
0,
reverse ? (refparams.reverse == REF_READ_REVERSE) : 0,
true, // load SA sample?
true, // load ftab?
true, // load rstarts?
false,
false);
SString<char> s2;
ebwt.restore(s2);
ebwt.evictFromMemory();
{
SString<char> joinedss = Ebwt::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;
}
}
}
}
