void bigWigFileCreate(
char *inName, /* Input file in ascii wiggle format. */
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. 512 is good. */
boolean clipDontDie, /* If TRUE then clip items off end of chrom rather than dying. */
boolean compress, /* If TRUE then compress data. */
boolean keepAllChromosomes, /* If TRUE then store all chromosomes in chromosomal b-tree. */
boolean fixedSummaries, /* If TRUE then impose fixed summary levels. */
char *outName)
/* Convert ascii format wig file (in fixedStep, variableStep or bedGraph format)
* to binary big wig format. */
{
/* This code needs to agree with code in two other places currently - bigBedFileCreate,
* and bbiFileOpen. I'm thinking of refactoring to share at least between
* bigBedFileCreate and bigWigFileCreate. It'd be great so it could be structured
* so that it could send the input in one chromosome at a time, and send in the zoom
* stuff only after all the chromosomes are done. This'd potentially reduce the memory
* footprint by a factor of 2 or 4. Still, for now it works. -JK */
struct hash *chromSizeHash = bbiChromSizesFromFile(chromSizes);
struct lm *lm = lmInit(0);
struct bwgSection *sectionList = bwgParseWig(inName, clipDontDie, chromSizeHash, itemsPerSlot, lm);
if (sectionList == NULL)
errAbort("%s is empty of data", inName);
bwgCreate(sectionList, chromSizeHash, blockSize, itemsPerSlot, compress, keepAllChromosomes, fixedSummaries, outName);
lmCleanup(&lm);
}
void bedClip(char *inFile, char *chromSizes, char *outFile)
/* bedClip - Remove lines from bed file that refer to off-chromosome places.. */
{
struct hash *chromSizesHash = bbiChromSizesFromFile(chromSizes);
struct lineFile *lf = lineFileOpen(inFile, TRUE);
FILE *f = mustOpen(outFile, "w");
char *line;
while (lineFileNextReal(lf, &line))
{
char *chrom = nextWord(&line);
char *startString = nextWord(&line);
char *endString = nextWord(&line);
if (endString == NULL)
errAbort("Need at least three fields line %d of %s", lf->lineIx, lf->fileName);
if (startString[0] == '-')
{
verbose(2, "Clipping negative line %d of %s\n", lf->lineIx, lf->fileName);
continue; // Clip off negatives
}
if (!isdigit(startString[0]))
errAbort("Expecting number got %s line %d of %s", startString, lf->lineIx, lf->fileName);
if (!isdigit(endString[0]))
errAbort("Expecting number got %s line %d of %s", endString, lf->lineIx, lf->fileName);
int start = sqlUnsigned(startString);
int end = sqlUnsigned(endString);
if (start >= end)
{
verbose(2, "Clipping end <= start line %d of %s\n", lf->lineIx, lf->fileName);
continue;
}
struct hashEl *hel = hashLookup(chromSizesHash, chrom);
if (hel == NULL)
errAbort("Chromosome %s isn't in %s line %d of %s\n", chrom, chromSizes, lf->lineIx, lf->fileName);
int chromSize = ptToInt(hel->val);
if (end > chromSize)
{
verbose(2, "Clipping end > chromSize line %d of %s\n", lf->lineIx, lf->fileName);
continue;
}
fprintf(f, "%s\t%s\t%s", chrom, startString, endString);
line = skipLeadingSpaces(line);
if (line == NULL || line[0] == 0)
fputc('\n', f);
else
fprintf(f, "\t%s\n", line);
}
carefulClose(&f);
}
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);
}