Added some new code for shard support over reads

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@385 348d0f76-0448-11de-a6fe-93d51630548a
This commit is contained in:
aaron 2009-04-13 18:11:43 +00:00
parent d44c30154a
commit dd604799dc
8 changed files with 483 additions and 298 deletions

View File

@ -0,0 +1,80 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import net.sf.samtools.SAMSequenceDictionary;
/**
*
* User: aaron
* Date: Apr 6, 2009
* Time: 8:23:19 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
* This software and its documentation are copyright 2009 by the
* Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
*
* This software is supplied without any warranty or guaranteed support whatsoever. Neither
* the Broad Institute nor MIT can be responsible for its use, misuse, or functionality.
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 6, 2009
* <p/>
* Class LinearShard
* <p/>
* A exponential strategy
*/
public class ExpGrowthLocusShardStrategy extends LocusShardStrategy {
// fixed size
private long baseSize = 100000;
private long currentExp = 0;
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param dic the seq dictionary
*/
ExpGrowthLocusShardStrategy(SAMSequenceDictionary dic, long startSize) {
super(dic);
this.baseSize = startSize;
currentExp = 0;
}
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param strat the shatter to convert from
*/
ExpGrowthLocusShardStrategy(LocusShardStrategy strat) {
super(strat);
this.baseSize = strat.nextShardSize();
currentExp = 0;
}
/**
* set the next shards size
*
* @param size adjust the next size to this
*/
public void adjustNextShardSize(long size) {
baseSize = size;
currentExp = 0;
}
/**
* This is how the various shards strategies implements their approach
*
* @return the next shard size
*/
protected long nextShardSize() {
// we grow the exponentially, we just have to make sure we start at zero
++currentExp;
return (long) Math.floor(Math.pow((double) baseSize, (double) currentExp));
}
}

View File

@ -0,0 +1,75 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import net.sf.samtools.SAMSequenceDictionary;
/**
*
* User: aaron
* Date: Apr 6, 2009
* Time: 7:18:19 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
* This software and its documentation are copyright 2009 by the
* Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
*
* This software is supplied without any warranty or guaranteed support whatsoever. Neither
* the Broad Institute nor MIT can be responsible for its use, misuse, or functionality.
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 6, 2009
* <p/>
* Class AdaptiveShard
* <p/>
* allows you to change the sharding length as you traverse
*/
class LinearLocusShardStrategy extends LocusShardStrategy {
// default the next size to 100,000
private long nextShardSize = 100000;
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param dic the seq dictionary
*/
LinearLocusShardStrategy(SAMSequenceDictionary dic, long startSize) {
super(dic);
this.nextShardSize = startSize;
}
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param strat the shatter to convert from
*/
LinearLocusShardStrategy(LocusShardStrategy strat) {
super(strat);
this.nextShardSize = strat.nextShardSize();
}
/**
* set the next shards size
*
* @param size adjust the next size to this
*/
public void adjustNextShardSize(long size) {
nextShardSize = size;
}
/**
* This is how the various shards strategies implements their approach
*
* @return the next shard size
*/
protected long nextShardSize() {
return nextShardSize;
}
}

View File

@ -0,0 +1,64 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import org.broadinstitute.sting.utils.GenomeLoc;
/**
*
* User: aaron
* Date: Apr 7, 2009
* Time: 1:19:49 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
* This software and its documentation are copyright 2009 by the
* Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
*
* This software is supplied without any warranty or guaranteed support whatsoever. Neither
* the Broad Institute nor MIT can be responsible for its use, misuse, or functionality.
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 7, 2009
* <p/>
* Class Shard
* <p/>
* This is the base class for shards. Right now it does little more then
* wrap GenomeLoc (actually nothing more), but it's good to have the class
* in place so it's easier to change guts later.
*/
public class LocusShard implements Shard {
// currently our location
final GenomeLoc mLoc;
public LocusShard(GenomeLoc loc) {
this.mLoc = loc;
}
/** @return the genome location represented by this shard */
public GenomeLoc getGenomeLoc() {
return mLoc;
}
/**
* what kind of shard do we return
*
* @return ShardType, indicating the type
*/
public ShardType getShardType() {
return ShardType.LOCUS;
}
/**
* return a shard representing the passed in GenomeLoc
*
* @return
*/
public static LocusShard toShard(GenomeLoc loc) {
return new LocusShard(loc);
}
}

View File

@ -0,0 +1,171 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import net.sf.samtools.SAMSequenceDictionary;
import org.broadinstitute.sting.utils.GenomeLoc;
import java.util.Iterator;
/**
*
* User: aaron
* Date: Apr 6, 2009
* Time: 11:23:17 AM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
* This software and its documentation are copyright 2009 by the
* Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
*
* This software is supplied without any warranty or guaranteed support whatsoever. Neither
* the Broad Institute nor MIT can be responsible for its use, misuse, or functionality.
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 6, 2009
* <p/>
* Interface Shard
* <p/>
* The shard interface, which controls how data is divided
*/
public abstract class LocusShardStrategy implements ShardStrategy {
// this stores the seq dictionary, which is a reference for the
// lengths and names of contigs, which you need to generate an iterative stratagy
protected final SAMSequenceDictionary dic;
// the current genome location
protected GenomeLoc mLoc = null;
// current seq location
protected int seqLoc = 0;
// the actual last size; this can change based on contig endings
protected long lastGenomeLocSize = 0;
// do we have another contig?
private boolean nextContig = false;
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param dic the seq dictionary
*/
LocusShardStrategy(SAMSequenceDictionary dic) {
this.dic = dic;
mLoc = new GenomeLoc(dic.getSequence(0).getSequenceName(), 0, 0);
if (dic.getSequences().size() > 0) {
nextContig = true;
}
}
/**
* the copy constructor,
*
* @param old the old strategy
*/
LocusShardStrategy(LocusShardStrategy old) {
this.dic = old.dic;
this.mLoc = old.mLoc;
this.seqLoc = old.seqLoc;
this.lastGenomeLocSize = old.lastGenomeLocSize;
this.nextContig = old.nextContig;
}
/**
*
* Abstract methods that each strategy has to implement
*
*/
/**
* set the next shards size
*
* @param size adjust the next size to this
*/
public abstract void adjustNextShardSize(long size);
/**
* This is how the various shards strategies implements their approach
*
* @return the next shard size
*/
abstract long nextShardSize();
/**
*
* Concrete methods that each strategy does not have to implement
*
*/
/**
* get the next shard, based on the return size of nextShardSize
*
* @return
*/
public LocusShard next() {
// lets get some background info on the problem
long length = dic.getSequence(seqLoc).getSequenceLength();
long proposedSize = nextShardSize();
long nextStart = mLoc.getStop() + 1;
// can we fit it into the current seq size?
if (nextStart + proposedSize < length) {
lastGenomeLocSize = proposedSize;
mLoc = new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, nextStart + proposedSize);
return LocusShard.toShard(new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, nextStart + proposedSize));
}
// else we can't make it in the current location, we have to stitch one together
else {
lastGenomeLocSize = nextStart + proposedSize - length;
// move to the next contig
jumpContig();
return LocusShard.toShard(new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, lastGenomeLocSize));
}
}
/** jump to the next contig */
private void jumpContig() {
++seqLoc;
if (dic.getSequences().size() <= seqLoc) {
nextContig = false;
return;
}
// the next sequence should start at the begining of the next contig
mLoc = new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), 0, 0);
}
/**
* is there another GenomeLoc to get?
*
* @return
*/
public boolean hasNext() {
return nextContig;
}
/** we don't support remove */
public void remove() {
throw new UnsupportedOperationException("Can not remove records from a shard iterator!");
}
/**
* to be for-each(able), we must implement this method
*
* @return
*/
public Iterator<Shard> iterator() {
return this;
}
}

View File

@ -0,0 +1,57 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import org.broadinstitute.sting.utils.GenomeLoc;
/**
*
* User: aaron
* Date: Apr 10, 2009
* Time: 5:03:13 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
* This software and its documentation are copyright 2009 by the
* Broad Institute/Massachusetts Institute of Technology. All rights are reserved.
*
* This software is supplied without any warranty or guaranteed support whatsoever. Neither
* the Broad Institute nor MIT can be responsible for its use, misuse, or functionality.
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 10, 2009
* <p/>
* Class ReadShard
* <p/>
* A class for sharded reads.
*/
public class ReadShard implements Shard {
// the count of the reads we want to copy off
int size = 0;
/**
* create a read shard, given a read size
* @param size
*/
public ReadShard(int size) {
this.size = size;
}
/** @return the genome location represented by this shard */
public GenomeLoc getGenomeLoc() {
return null; //To change body of implemented methods use File | Settings | File Templates.
}
/**
* what kind of shard do we return
*
* @return ShardType, indicating the type
*/
public ShardType getShardType() {
return ShardType.READ;
}
}

View File

@ -2,11 +2,12 @@ package org.broadinstitute.sting.gatk.dataSources.shards;
import org.broadinstitute.sting.utils.GenomeLoc;
import java.io.Serializable;
/**
*
* User: aaron
* Date: Apr 7, 2009
* Time: 1:19:49 PM
* Date: Apr 10, 2009
* Time: 5:00:27 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
@ -18,38 +19,26 @@ import org.broadinstitute.sting.utils.GenomeLoc;
*
*/
/**
* @author aaron
* @version 1.0
* @date Apr 7, 2009
* @date Apr 10, 2009
* <p/>
* Class Shard
* Interface Shard
* <p/>
* This is the base class for shards. Right now it does little more then
* wrap GenomeLoc (actually nothing more), but it's good to have the class
* in place so it's easier to change guts later.
* The base interface for shards.
*/
public class Shard {
// currently our location
final GenomeLoc mLoc;
public Shard(GenomeLoc loc) {
this.mLoc = loc;
public interface Shard extends Serializable {
enum ShardType {
READ, LOCUS
}
/** @return the genome location represented by this shard */
public GenomeLoc getGenomeLoc() {
return mLoc;
}
public GenomeLoc getGenomeLoc();
/**
* return a shard representing the passed in GenomeLoc
*
* @return
* what kind of shard do we return
* @return ShardType, indicating the type
*/
public static Shard toShard(GenomeLoc loc) {
return new Shard(loc);
}
public ShardType getShardType();
}

View File

@ -1,17 +1,11 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import net.sf.samtools.SAMSequenceDictionary;
import org.apache.log4j.Logger;
import org.broadinstitute.sting.utils.GenomeLoc;
import java.util.Iterator;
import java.util.List;
/**
*
* User: aaron
* Date: Apr 6, 2009
* Time: 11:23:17 AM
* Date: Apr 10, 2009
* Time: 4:55:37 PM
*
* The Broad Institute
* SOFTWARE COPYRIGHT NOTICE AGREEMENT
@ -26,239 +20,12 @@ import java.util.List;
/**
* @author aaron
* @version 1.0
* @date Apr 6, 2009
* @date Apr 10, 2009
* <p/>
* Interface Shard
* Interface ShardStrategy
* <p/>
* The shard interface, which controls how data is divided
* The base interface for the sharding strategy; before we had a base abstract
* class, but not this will be an interface to accomidate read based sharding
*/
public abstract class ShardStrategy implements Iterator<Shard>, Iterable<Shard> {
// this stores the seq dictionary, which is a reference for the
// lengths and names of contigs, which you need to generate an iterative stratagy
protected final SAMSequenceDictionary dic;
// the current genome location
protected GenomeLoc mLoc = null;
// current seq location
protected int seqLoc = 0;
// the actual last size; this can change based on contig endings
protected long lastGenomeLocSize = 0;
// do we have another contig?
private boolean nextContig = false;
/** our log, which we want to capture anything from this class */
private static Logger logger = Logger.getLogger(ShardStrategy.class);
/** our interal list * */
private List<GenomeLoc> intervals = null;
/** our interal list * */
private int currentInterval = -1;
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param dic the seq dictionary
*/
ShardStrategy(SAMSequenceDictionary dic) {
this.dic = dic;
mLoc = new GenomeLoc(dic.getSequence(0).getSequenceName(), 0, 0);
if (dic.getSequences().size() > 0) {
nextContig = true;
}
}
/**
* the copy constructor,
*
* @param old the old strategy
*/
ShardStrategy(ShardStrategy old) {
this.dic = old.dic;
this.mLoc = old.mLoc;
this.seqLoc = old.seqLoc;
this.lastGenomeLocSize = old.lastGenomeLocSize;
this.nextContig = old.nextContig;
}
/**
* the constructor, taking a seq dictionary to parse out contigs
*
* @param dic the seq dictionary
* @param intervals file
*/
ShardStrategy(SAMSequenceDictionary dic, List<GenomeLoc> intervals) {
this.dic = dic;
this.intervals = intervals;
this.currentInterval = 0;
mLoc = new GenomeLoc(intervals.get(0).getContig(), intervals.get(0).getStart() - 1, intervals.get(0).getStart() - 1);
if (dic.getSequences().size() > 0) {
nextContig = true;
}
}
/**
*
* Abstract methods that each strategy has to implement
*
*/
/**
* set the next shards size
*
* @param size adjust the next size to this
*/
public abstract void adjustNextShardSize(long size);
/**
* This is how the various shards strategies implements their approach
*
* @return the next shard size
*/
abstract long nextShardSize();
/**
*
* Concrete methods that each strategy does not have to implement
*
*/
/**
* get the next shard, based on the return size of nextShardSize
*
* @return the next shard
*/
public Shard next() {
// lets get some background info on the problem
long length = dic.getSequence(seqLoc).getSequenceLength();
long proposedSize = nextShardSize();
long nextStart = mLoc.getStop() + 1;
// if we don't have an interval file, use the non interval based approach. Simple, eh?
if (this.intervals == null) {
return nonIntervaledNext(length, proposedSize, nextStart);
} else {
return intervaledNext(length, proposedSize, nextStart);
}
}
private Shard intervaledNext(long length, long proposedSize, long nextStart) {
// get the current genome location
GenomeLoc loc = intervals.get(currentInterval);
if (nextStart + proposedSize > loc.getStop()) {
// we need to move the next interval
proposedSize = loc.getStop() - nextStart;
lastGenomeLocSize = proposedSize;
// the next sequence should start at the begining of the next contig
Shard ret = Shard.toShard(new GenomeLoc(intervals.get(currentInterval).getContig(), nextStart, nextStart + proposedSize - 1));
++currentInterval;
if (intervals.size() > currentInterval) {
mLoc = new GenomeLoc(intervals.get(currentInterval).getContig(), intervals.get(currentInterval).getStart() - 1, intervals.get(currentInterval).getStart() - 1);
}
return ret;// return
} else {
// we need to move the next interval
lastGenomeLocSize = proposedSize;
// the next sequence should start at the begining of the next contig
Shard ret = Shard.toShard(new GenomeLoc(intervals.get(currentInterval).getContig(), nextStart, nextStart + proposedSize - 1));
mLoc = new GenomeLoc(intervals.get(currentInterval).getContig(), nextStart, nextStart + proposedSize - 1);
return ret;// return
}
}
/**
* Get the next shard, if we don't have intervals to traverse over
*
* @param length the length of the contig
* @param proposedSize the proposed size
* @param nextStart the next start location
* @return the shard to return to the user
*/
private Shard nonIntervaledNext(long length, long proposedSize, long nextStart) {
// can we fit it into the current seq size?
if (nextStart + proposedSize - 1 < length) {
lastGenomeLocSize = proposedSize;
mLoc = new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, nextStart + proposedSize - 1);
return Shard.toShard(new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, nextStart + proposedSize - 1));
}
// else we can't make it in the current location, we have to stitch one together
else {
// lets find out the remaining size of the current contig
long overflow = nextStart + proposedSize - 1 - length;
logger.debug("Overflow = " + overflow + " length: " + length);
// set our last size counter to the remaining size
lastGenomeLocSize = proposedSize - overflow;
// move to the next contig
// the next sequence should start at the begining of the next contig
Shard ret = Shard.toShard(new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), nextStart, nextStart + lastGenomeLocSize));
// now jump ahead to the next contig
jumpContig();
// return the shard
return ret;
}
}
/** jump to the next contig */
private void jumpContig() {
++seqLoc;
if (!(seqLoc < dic.getSequences().size())) {
nextContig = false;
return;
}
logger.debug("Next contig, name = " + dic.getSequence(seqLoc).getSequenceName());
mLoc = new GenomeLoc(dic.getSequence(seqLoc).getSequenceName(), 0, 0);
}
/**
* is there another GenomeLoc to get?
*
* @return
*/
public boolean hasNext() {
// if we don't have an interval file, use the non interval based approach. Simple, eh?
if (this.intervals == null) {
return nextContig;
} else {
return (this.currentInterval < this.intervals.size());
}
}
/** we don't support remove */
public void remove() {
throw new UnsupportedOperationException("Can not remove records from a shard iterator!");
}
/**
* to be for-each(able), we must implement this method
*
* @return
*/
public Iterator<Shard> iterator() {
return this;
}
public interface ShardStrategy extends Iterator<Shard>, Iterable<Shard> {
}

View File

@ -1,10 +1,6 @@
package org.broadinstitute.sting.gatk.dataSources.shards;
import net.sf.samtools.SAMSequenceDictionary;
import org.apache.log4j.Logger;
import org.broadinstitute.sting.utils.GenomeLoc;
import java.util.List;
/**
*
@ -35,13 +31,9 @@ import java.util.List;
*/
public class ShardStrategyFactory {
public enum SHATTER_STRATEGY {
LINEAR, EXPONENTIAL
LINEAR, EXPONENTIAL, READS
}
/** our log, which we want to capture anything from this class */
private static Logger logger = Logger.getLogger(ShardStrategyFactory.class);
/**
* get a new shatter strategy
*
@ -53,29 +45,9 @@ public class ShardStrategyFactory {
static public ShardStrategy shatter(SHATTER_STRATEGY strat, SAMSequenceDictionary dic, long startingSize) {
switch (strat) {
case LINEAR:
return new LinearShardStrategy(dic, startingSize);
return new LinearLocusShardStrategy(dic, startingSize);
case EXPONENTIAL:
return new ExpGrowthShardStrategy(dic, startingSize);
default:
throw new RuntimeException("Strategy: " + strat + " isn't implemented");
}
}
/**
* get a new shatter strategy
*
* @param strat what's our strategy - SHATTER_STRATEGY type
* @param dic the seq dictionary
* @param startingSize the starting size
* @return
*/
static public ShardStrategy shatter(SHATTER_STRATEGY strat, SAMSequenceDictionary dic, long startingSize, List<GenomeLoc> lst) {
switch (strat) {
case LINEAR:
return new LinearShardStrategy(dic, startingSize, lst);
case EXPONENTIAL:
return new ExpGrowthShardStrategy(dic, startingSize, lst);
return new ExpGrowthLocusShardStrategy(dic, startingSize);
default:
throw new RuntimeException("Strategy: " + strat + " isn't implemented");
}
@ -89,16 +61,26 @@ public class ShardStrategyFactory {
* @param convertFrom convert from this strategy
* @return
*/
static public ShardStrategy transitionToShardStrategy(SHATTER_STRATEGY strat, ShardStrategy convertFrom) {
static public ShardStrategy transitionToShardStrategy(SHATTER_STRATEGY strat, LocusShardStrategy convertFrom) {
switch (strat) {
case LINEAR:
return new LinearShardStrategy(convertFrom);
return new LinearLocusShardStrategy(convertFrom);
case EXPONENTIAL:
return new ExpGrowthShardStrategy(convertFrom);
return new ExpGrowthLocusShardStrategy(convertFrom);
default:
throw new RuntimeException("Strategy: " + strat + " isn't implemented");
}
}
/**
* convert between types
*
* @param readCount the number of reads to include in each shard
* @return
*/
static public ShardStrategy shatterByReadCount(long readCount) {
return null;
}
}