void wigCorrelate(int inCount, char **inNames, char *outName)
/* wigCorrelate - Produce a table that correlates all pairs of wigs.. */
{
int i,j;
FILE *f = mustOpen(outName, "w");
verboseTimeInit();
for (i=0; i<inCount-1; ++i)
{
char *iName = inNames[i];
struct metaWig *iMeta = metaWigOpen(iName);
verboseTime(2, "parsed %s into %p", iName, iMeta);
for (j=i+1; j<inCount; ++j)
{
char *jName = inNames[j];
struct metaWig *jMeta = metaWigOpen(jName);
verboseTime(2, "parsed %s into %p", jName, jMeta);
fprintf(f, "%s\t%s\t", iName, jName);
fflush(f);
double r = correlatePair(iMeta, jMeta);
fprintf(f, "%g\n", r);
fflush(f);
verboseTime(2, "correlated %g from %s and %s", r, iName, jName);
metaWigClose(&jMeta);
}
metaWigClose(&iMeta);
}
carefulClose(&f);
}
void splatToEland(char *inName, char *outName)
/* splatToEland - Convert from splat to eland format.. */
{
verboseTimeInit();
struct splatAli *list = splatAliLoadAll(inName);
verboseTime(1, "Loaded %d sequences from %s", slCount(list), inName);
if (needSort(list))
{
slSort(&list, splatAliCmpReadName);
verboseTime(1, "Sorted by read name");
}
FILE *f = mustOpen(outName, "w");
struct splatAli *el, *next;
for (el = list; el != NULL; el = next)
{
int sameCount;
findNextDifferent(el, &next, &sameCount);
int bestScore, bestCount;
splatAliLookForBest(el, next, &bestScore, &bestCount);
int subCounts[3];
countSubsInList(el, next, subCounts);
if (multi)
elandMultiOutput(el, next, bestScore, bestCount, subCounts, f);
else
elandSingleOutput(el, next, bestScore, bestCount, subCounts, f);
}
}
void verboseTime(int verbosity, char *label, ...)
/* Print label and how long it's been since last call. Start time can be
* initialized with verboseTimeInit, otherwise the elapsed time will be
* zero. */
{
assert(label != NULL); // original version allowed this, but breaks some GCCs
if (lastTime < 0)
verboseTimeInit();
long time = clock1000();
va_list args;
va_start(args, label);
verboseVa(verbosity, label, args);
verbose(verbosity, ": %ld millis\n", time - lastTime);
lastTime = time;
va_end(args);
}
void itsaMake(int inCount, char *inputs[], char *output)
/* itsaMake - Make a suffix array file out of input DNA sequences.. */
{
verboseTimeInit();
bits64 maxGenomeSize = 1024LL*1024*1024*4;
itsaBaseToValInit();
/* Load all DNA, make sure names are unique, and alphabetize by name. */
struct dnaSeq *seqList = NULL, *seq;
struct hash *uniqSeqHash = hashNew(0);
bits64 totalDnaSize = 1; /* FOr space between. */
int inputIx;
for (inputIx=0; inputIx<inCount; ++inputIx)
{
char * input = inputs[inputIx];
struct dnaLoad *dl = dnaLoadOpen(input);
while ((seq = dnaLoadNext(dl)) != NULL)
{
verbose(1, "read %s with %d bases\n", seq->name, seq->size);
if (hashLookup(uniqSeqHash, seq->name))
errAbort("Input sequence name %s repeated, all must be unique.", seq->name);
totalDnaSize += seq->size + 1;
if (totalDnaSize > maxGenomeSize)
errAbort("Too much DNA. Can only handle up to %lld bases", maxGenomeSize);
slAddHead(&seqList, seq);
}
dnaLoadClose(&dl);
}
slSort(&seqList, dnaSeqCmpName);
verboseTime(1, "Loaded %lld bases in %d sequences", totalDnaSize, slCount(seqList));
/* Allocate big buffer for all DNA. */
DNA *allDna = globalAllDna = needHugeMem(totalDnaSize);
allDna[0] = 0;
bits64 chromOffset = 1; /* Have zeroes between each chrom, and before and after. */
/* Copy DNA to a single big buffer, and create chromInfo on each sequence. */
struct chromInfo *chrom, *chromList = NULL;
for (seq = seqList; seq != NULL; seq = seq->next)
{
AllocVar(chrom);
chrom->name = cloneString(seq->name);
chrom->size = seq->size;
chrom->offset = chromOffset;
slAddHead(&chromList, chrom);
toUpperN(seq->dna, seq->size);
memcpy(allDna + chromOffset, seq->dna, seq->size + 1);
chromOffset += seq->size + 1;
}
slReverse(&chromList);
/* Free up separate dna sequences because we're going to need a lot of RAM soon. */
/* Allocate index array, and offset and list arrays. */
dnaSeqFreeList(&seqList);
bits32 *index13;
AllocArray(index13, itsaSlotCount);
bits32 *offsetArray = needHugeMem(totalDnaSize * sizeof(bits32));
bits32 *listArray = needHugeZeroedMem(totalDnaSize * sizeof(bits32));
verboseTime(1, "Allocated buffers %lld bytes total",
(long long)(9LL*totalDnaSize + itsaSlotCount*sizeof(bits32)));
/* Where normally we'd keep some sort of structure with a next element to form a list
* of matching positions in each slot of our index, to conserve memory we'll do this
* with two parallel arrays. Because we're such cheapskates in terms of memory we'll
* (and still using 9*genomeSize bytes of RAM) we'll use these arrays for two different
* purposes.
* In the first phase they will together be used to form linked lists of
* offsets, and the 13mer index will point to the first item in each list. In this
* phase the offsetArray contains offsets into the allDna structure, and the listArray
* contains the next pointers for the list. After the first phase we write out the
* suffix array to disk.
* In the second phase we read the suffix array back into the offsetArray, and
* use the listArray for the traverseArray. We write out the traverse array to finish
* things up. */
/* Load up all DNA buffer. */
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
verbose(2, " About to do first pass index\n");
indexChromPass1(chrom, allDna, offsetArray, listArray, index13);
verbose(2, " Done first pass index\n");
}
verboseTime(1, "Done big bucket sort");
slReverse(&chromList);
itsaWriteMerged(chromList, allDna, offsetArray, listArray, index13, output);
}
void bbFileCreate(
char *inName, /* Input file in a tabular bed format <chrom><start><end> + whatever. */
char *chromSizes, /* Two column tab-separated file: <chromosome> <size>. */
int blockSize, /* Number of items to bundle in r-tree. 1024 is good. */
int itemsPerSlot, /* Number of items in lowest level of tree. 64 is good. */
char *asText, /* Field definitions in a string */
struct asObject *as, /* Field definitions parsed out */
boolean doCompress, /* If TRUE then compress data. */
struct slName *extraIndexList, /* List of extra indexes to add */
char *outName) /* BigBed output file name. */
/* Convert tab-separated bed file to binary indexed, zoomed bigBed version. */
{
/* Set up timing measures. */
verboseTimeInit();
struct lineFile *lf = lineFileOpen(inName, TRUE);
bits16 fieldCount = slCount(as->columnList);
bits16 extraIndexCount = slCount(extraIndexList);
struct bbExIndexMaker *eim = NULL;
if (extraIndexList != NULL)
eim = bbExIndexMakerNew(extraIndexList, as);
/* Load in chromosome sizes. */
struct hash *chromSizesHash = NULL;
if (sizesIs2Bit)
chromSizesHash = twoBitChromHash(chromSizes);
else
chromSizesHash = bbiChromSizesFromFile(chromSizes);
verbose(2, "Read %d chromosomes and sizes from %s\n", chromSizesHash->elCount, chromSizes);
/* Do first pass, mostly just scanning file and counting hits per chromosome. */
int minDiff = 0;
double aveSize = 0;
bits64 bedCount = 0;
bits32 uncompressBufSize = 0;
struct bbiChromUsage *usageList = bbiChromUsageFromBedFile(lf, chromSizesHash, eim,
&minDiff, &aveSize, &bedCount, tabSep);
verboseTime(1, "pass1 - making usageList (%d chroms)", slCount(usageList));
verbose(2, "%d chroms in %s. Average span of beds %f\n", slCount(usageList), inName, aveSize);
/* Open output file and write dummy header. */
FILE *f = mustOpen(outName, "wb");
bbiWriteDummyHeader(f);
bbiWriteDummyZooms(f);
/* Write out autoSql string */
bits64 asOffset = ftell(f);
mustWrite(f, asText, strlen(asText) + 1);
verbose(2, "as definition has %d columns\n", fieldCount);
/* Write out dummy total summary. */
struct bbiSummaryElement totalSum;
ZeroVar(&totalSum);
bits64 totalSummaryOffset = ftell(f);
bbiSummaryElementWrite(f, &totalSum);
/* Write out dummy header extension */
bits64 extHeaderOffset = ftell(f);
bits16 extHeaderSize = 64;
repeatCharOut(f, 0, extHeaderSize);
/* Write out extra index stuff if need be. */
bits64 extraIndexListOffset = 0;
bits64 extraIndexListEndOffset = 0;
if (extraIndexList != NULL)
{
extraIndexListOffset = ftell(f);
int extraIndexSize = 16 + 4*1; // Fixed record size 16, plus 1 times field size of 4
repeatCharOut(f, 0, extraIndexSize*extraIndexCount);
extraIndexListEndOffset = ftell(f);
}
/* Write out chromosome/size database. */
bits64 chromTreeOffset = ftell(f);
bbiWriteChromInfo(usageList, blockSize, f);
/* Set up to keep track of possible initial reduction levels. */
int resScales[bbiMaxZoomLevels], resSizes[bbiMaxZoomLevels];
int resTryCount = bbiCalcResScalesAndSizes(aveSize, resScales, resSizes);
/* Write out primary full resolution data in sections, collect stats to use for reductions. */
bits64 dataOffset = ftell(f);
bits32 blockCount = 0;
bits32 maxBlockSize = 0;
struct bbiBoundsArray *boundsArray = NULL;
writeOne(f, bedCount);
if (bedCount > 0)
{
blockCount = bbiCountSectionsNeeded(usageList, itemsPerSlot);
AllocArray(boundsArray, blockCount);
lineFileRewind(lf);
if (eim)
bbExIndexMakerAllocChunkArrays(eim, bedCount);
writeBlocks(usageList, lf, as, itemsPerSlot, boundsArray, blockCount, doCompress,
f, resTryCount, resScales, resSizes, eim, bedCount, fieldCount, &maxBlockSize);
}
verboseTime(1, "pass2 - checking and writing primary data (%lld records, %d fields)",
(long long)bedCount, fieldCount);
/* Write out primary data index. */
bits64 indexOffset = ftell(f);
cirTreeFileBulkIndexToOpenFile(boundsArray, sizeof(boundsArray[0]), blockCount,
blockSize, 1, NULL, bbiBoundsArrayFetchKey, bbiBoundsArrayFetchOffset,
indexOffset, f);
freez(&boundsArray);
verboseTime(2, "index write");
/* Declare arrays and vars that track the zoom levels we actually output. */
bits32 zoomAmounts[bbiMaxZoomLevels];
bits64 zoomDataOffsets[bbiMaxZoomLevels];
bits64 zoomIndexOffsets[bbiMaxZoomLevels];
/* Call monster zoom maker library function that bedGraphToBigWig also uses. */
int zoomLevels = 0;
if (bedCount > 0)
{
zoomLevels = bbiWriteZoomLevels(lf, f, blockSize, itemsPerSlot,
bedWriteReducedOnceReturnReducedTwice, fieldCount,
doCompress, indexOffset - dataOffset,
usageList, resTryCount, resScales, resSizes,
zoomAmounts, zoomDataOffsets, zoomIndexOffsets, &totalSum);
}
/* Write out extra indexes if need be. */
if (eim)
{
int i;
for (i=0; i < eim->indexCount; ++i)
{
eim->fileOffsets[i] = ftell(f);
maxBedNameSize = eim->maxFieldSize[i];
qsort(eim->chunkArrayArray[i], bedCount,
sizeof(struct bbNamedFileChunk), bbNamedFileChunkCmpByName);
assert(sizeof(struct bbNamedFileChunk) == sizeof(eim->chunkArrayArray[i][0]));
bptFileBulkIndexToOpenFile(eim->chunkArrayArray[i], sizeof(eim->chunkArrayArray[i][0]),
bedCount, blockSize, bbNamedFileChunkKey, maxBedNameSize, bbNamedFileChunkVal,
sizeof(bits64) + sizeof(bits64), f);
verboseTime(1, "Sorting and writing extra index %d", i);
}
}
/* Figure out buffer size needed for uncompression if need be. */
if (doCompress)
{
int maxZoomUncompSize = itemsPerSlot * sizeof(struct bbiSummaryOnDisk);
uncompressBufSize = max(maxBlockSize, maxZoomUncompSize);
}
/* Go back and rewrite header. */
rewind(f);
bits32 sig = bigBedSig;
bits16 version = bbiCurrentVersion;
bits16 summaryCount = zoomLevels;
bits32 reserved32 = 0;
bits64 reserved64 = 0;
bits16 definedFieldCount = bedN;
/* Write fixed header */
writeOne(f, sig);
writeOne(f, version);
writeOne(f, summaryCount);
writeOne(f, chromTreeOffset);
writeOne(f, dataOffset);
writeOne(f, indexOffset);
writeOne(f, fieldCount);
writeOne(f, definedFieldCount);
writeOne(f, asOffset);
writeOne(f, totalSummaryOffset);
writeOne(f, uncompressBufSize);
writeOne(f, extHeaderOffset);
assert(ftell(f) == 64);
/* Write summary headers with data. */
int i;
verbose(2, "Writing %d levels of zoom\n", zoomLevels);
for (i=0; i<zoomLevels; ++i)
{
verbose(3, "zoomAmounts[%d] = %d\n", i, (int)zoomAmounts[i]);
writeOne(f, zoomAmounts[i]);
writeOne(f, reserved32);
writeOne(f, zoomDataOffsets[i]);
writeOne(f, zoomIndexOffsets[i]);
}
/* Write rest of summary headers with no data. */
for (i=zoomLevels; i<bbiMaxZoomLevels; ++i)
{
writeOne(f, reserved32);
writeOne(f, reserved32);
writeOne(f, reserved64);
writeOne(f, reserved64);
}
/* Write total summary. */
fseek(f, totalSummaryOffset, SEEK_SET);
bbiSummaryElementWrite(f, &totalSum);
/* Write extended header */
fseek(f, extHeaderOffset, SEEK_SET);
writeOne(f, extHeaderSize);
writeOne(f, extraIndexCount);
writeOne(f, extraIndexListOffset);
repeatCharOut(f, 0, 52); // reserved
assert(ftell(f) - extHeaderOffset == extHeaderSize);
/* Write extra index offsets if need be. */
if (extraIndexCount != 0)
{
fseek(f, extraIndexListOffset, SEEK_SET);
int i;
for (i=0; i<extraIndexCount; ++i)
{
// Write out fixed part of index info
bits16 type = 0; // bPlusTree type
bits16 indexFieldCount = 1;
writeOne(f, type);
writeOne(f, indexFieldCount);
writeOne(f, eim->fileOffsets[i]);
repeatCharOut(f, 0, 4); // reserved
// Write out field list - easy this time because for now always only one field.
bits16 fieldId = eim->indexFields[i];
writeOne(f, fieldId);
repeatCharOut(f, 0, 2); // reserved
}
assert(ftell(f) == extraIndexListEndOffset);
}
/* Write end signature. */
fseek(f, 0L, SEEK_END);
writeOne(f, sig);
/* Clean up. */
lineFileClose(&lf);
carefulClose(&f);
freeHash(&chromSizesHash);
bbiChromUsageFreeList(&usageList);
asObjectFreeList(&as);
}
void bedGraphToBigWig(char *inName, char *chromSizes, char *outName)
/* bedGraphToBigWig - Convert a bedGraph program to bigWig.. */
{
verboseTimeInit();
struct lineFile *lf = lineFileOpen(inName, TRUE);
struct hash *chromSizesHash = bbiChromSizesFromFile(chromSizes);
verbose(2, "%d chroms in %s\n", chromSizesHash->elCount, chromSizes);
int minDiff = 0, i;
double aveSize = 0;
bits64 bedCount = 0;
bits32 uncompressBufSize = 0;
struct bbiChromUsage *usageList = bbiChromUsageFromBedFile(lf, chromSizesHash, NULL,
&minDiff, &aveSize, &bedCount);
verboseTime(2, "pass1");
verbose(2, "%d chroms in %s, minDiff=%d, aveSize=%g, bedCount=%lld\n",
slCount(usageList), inName, minDiff, aveSize, bedCount);
/* Write out dummy header, zoom offsets. */
FILE *f = mustOpen(outName, "wb");
bbiWriteDummyHeader(f);
bbiWriteDummyZooms(f);
/* Write out dummy total summary. */
struct bbiSummaryElement totalSum;
ZeroVar(&totalSum);
bits64 totalSummaryOffset = ftell(f);
bbiSummaryElementWrite(f, &totalSum);
/* Write out chromosome/size database. */
bits64 chromTreeOffset = ftell(f);
bbiWriteChromInfo(usageList, blockSize, f);
/* Set up to keep track of possible initial reduction levels. */
int resScales[bbiMaxZoomLevels], resSizes[bbiMaxZoomLevels];
int resTryCount = bbiCalcResScalesAndSizes(aveSize, resScales, resSizes);
/* Write out primary full resolution data in sections, collect stats to use for reductions. */
bits64 dataOffset = ftell(f);
bits64 sectionCount = bbiCountSectionsNeeded(usageList, itemsPerSlot);
writeOne(f, sectionCount);
struct bbiBoundsArray *boundsArray;
AllocArray(boundsArray, sectionCount);
lineFileRewind(lf);
bits32 maxSectionSize = 0;
writeSections(usageList, lf, itemsPerSlot, boundsArray, sectionCount, f,
resTryCount, resScales, resSizes, doCompress, &maxSectionSize);
verboseTime(2, "pass2");
/* Write out primary data index. */
bits64 indexOffset = ftell(f);
cirTreeFileBulkIndexToOpenFile(boundsArray, sizeof(boundsArray[0]), sectionCount,
blockSize, 1, NULL, bbiBoundsArrayFetchKey, bbiBoundsArrayFetchOffset,
indexOffset, f);
verboseTime(2, "index write");
/* Declare arrays and vars that track the zoom levels we actually output. */
bits32 zoomAmounts[bbiMaxZoomLevels];
bits64 zoomDataOffsets[bbiMaxZoomLevels];
bits64 zoomIndexOffsets[bbiMaxZoomLevels];
/* Call monster zoom maker library function that bedToBigBed also uses. */
int zoomLevels = bbiWriteZoomLevels(lf, f, blockSize, itemsPerSlot,
bedGraphWriteReducedOnceReturnReducedTwice, 4,
doCompress, indexOffset - dataOffset,
usageList, resTryCount, resScales, resSizes,
zoomAmounts, zoomDataOffsets, zoomIndexOffsets, &totalSum);
/* Figure out buffer size needed for uncompression if need be. */
if (doCompress)
{
int maxZoomUncompSize = itemsPerSlot * sizeof(struct bbiSummaryOnDisk);
uncompressBufSize = max(maxSectionSize, maxZoomUncompSize);
}
/* Go back and rewrite header. */
rewind(f);
bits32 sig = bigWigSig;
bits16 version = bbiCurrentVersion;
bits16 summaryCount = zoomLevels;
bits16 reserved16 = 0;
bits32 reserved32 = 0;
bits64 reserved64 = 0;
/* Write fixed header */
writeOne(f, sig);
writeOne(f, version);
writeOne(f, summaryCount);
writeOne(f, chromTreeOffset);
writeOne(f, dataOffset);
writeOne(f, indexOffset);
writeOne(f, reserved16); // fieldCount
writeOne(f, reserved16); // definedFieldCount
writeOne(f, reserved64); // autoSqlOffset
writeOne(f, totalSummaryOffset);
writeOne(f, uncompressBufSize);
writeOne(f, reserved64); // nameIndexOffset
assert(ftell(f) == 64);
/* Write summary headers with data. */
verbose(2, "Writing %d levels of zoom\n", zoomLevels);
for (i=0; i<zoomLevels; ++i)
{
verbose(3, "zoomAmounts[%d] = %d\n", i, (int)zoomAmounts[i]);
writeOne(f, zoomAmounts[i]);
writeOne(f, reserved32);
writeOne(f, zoomDataOffsets[i]);
writeOne(f, zoomIndexOffsets[i]);
}
/* Write rest of summary headers with no data. */
for (i=zoomLevels; i<bbiMaxZoomLevels; ++i)
{
writeOne(f, reserved32);
writeOne(f, reserved32);
writeOne(f, reserved64);
writeOne(f, reserved64);
}
/* Write total summary. */
fseek(f, totalSummaryOffset, SEEK_SET);
bbiSummaryElementWrite(f, &totalSum);
/* Write end signature. */
fseek(f, 0L, SEEK_END);
writeOne(f, sig);
lineFileClose(&lf);
carefulClose(&f);
}
void encode2MakeEncode3(char *sourceDir, char *sourceManifest, char *destDir, char *outMake,
char *outManifest, char *outRename)
/* encode2MakeEncode3 - Copy files in encode2 manifest and in case of tar'd files rezip them
* independently. */
{
struct encode2Manifest *fileList = encode2ManifestShortLoadAll(sourceManifest);
verbose(2, "Loaded information on %d files from %s\n", slCount(fileList), sourceManifest);
verboseTimeInit();
FILE *f = mustOpen(outMake, "w");
FILE *manF = mustOpen(outManifest, "w");
FILE *renameF = mustOpen(outRename, "w");
struct encode2Manifest *mi;
int destDirPrefixSize = strlen(destDir)+1; /* +1 for / */
struct hash *destDirHash = hashNew(0);
makeDirOnlyOnce(destDir, destDirHash);
/* Print first dependency in makefile - the one that causes all files to be made. */
fprintf(f, "startHere: all\n\techo all done\n\n");
/* Write out each file target, and save also list of all targets. */
struct slName *targetList = NULL;
int itemNo = 0;
for (mi = fileList; mi != NULL; mi = mi->next)
{
char *origFileName = cloneString(mi->fileName);
/* Keep a list of what all it goes to. */
struct slName *oneTargetList = NULL;
/* Make path to source file. */
char sourcePath[PATH_LEN];
safef(sourcePath, sizeof(sourcePath), "%s/%s", sourceDir, mi->fileName);
/* Make destination dir */
char localDir[PATH_LEN];
splitPath(mi->fileName, localDir, NULL, NULL);
char destSubDir[PATH_LEN];
safef(destSubDir, sizeof(destSubDir), "%s/%s", destDir, localDir);
makeDirOnlyOnce(destSubDir, destDirHash);
char destPath[PATH_LEN];
safef(destPath, sizeof(destPath), "%s/%s", destDir, mi->fileName);
processManifestItem(++itemNo, mi, sourceDir, destDir, &oneTargetList, f, manF);
/* Write out rename info. */
if (oneTargetList != NULL)
{
if (oneTargetList->next != NULL || !sameString(origFileName, oneTargetList->name + destDirPrefixSize))
{
fprintf(renameF, "%s", origFileName);
struct slName *target;
for (target = oneTargetList; target != NULL; target = target->next)
{
fprintf(renameF, " %s", target->name + destDirPrefixSize);
}
fprintf(renameF, "\n");
}
}
/* Save local target list to big one. */
targetList = slCat(oneTargetList, targetList);
}
slReverse(&targetList);
fprintf(f, "all:");
struct slName *target;
for (target = targetList; target != NULL; target = target->next)
fprintf(f, " %s", target->name);
fprintf(f, "\n");
carefulClose(&renameF);
carefulClose(&manF);
carefulClose(&f);
}