private static int AddAlignmentsFromEdgeState(EdgeState edgeState, Dictionary <PairClassification, List <PairResult> > pairResultLookup,
int numRetrievedFromLastBlock)
{
if (edgeState != null)
{
foreach (var key in edgeState.EdgeAlignments.Keys)
{
if (!pairResultLookup.ContainsKey(key))
{
pairResultLookup.Add(key, new List <PairResult>());
}
var alns = edgeState.EdgeAlignments[key];
pairResultLookup[key].AddRange(alns);
numRetrievedFromLastBlock += alns.Count;
}
}
return(numRetrievedFromLastBlock);
}
private static void AddEdgeHits(EdgeState edgeState, IBinEvidence binEvidence2, int offset, int startInOld)
{
binEvidence2.CombineBinEvidence(edgeState.BinEvidence, offset, startInOld, edgeState.BinEvidence.NumBins);
}
public Task[] GetAndLinkAllClassificationBlocksWithEcFinalization(
ISourceBlock <PairResult> pairClassifierBlock,
int startPosition, int endPosition, ConcurrentDictionary <int, EdgeState> edgeStates,
ConcurrentDictionary <int, Task> edgeToWaitOn, int prevBlockStart,
bool isFinalTask = false)
{
if (_lightDebug)
{
Logger.WriteToLog(
$"Creating tasks for region {_chrom}:{startPosition}-{endPosition}.");
}
var allToWaitFor = new ConcurrentDictionary <Task, int>();
var messySiteWidth = _geminiOptions.MessySiteWidth;
var effectiveMax = 0;
var adjustedStartPosition = startPosition;
adjustedStartPosition = RoundedStartPosition(adjustedStartPosition, messySiteWidth);
var pairResultLookup = new ConcurrentDictionary <PairClassification, List <PairResult> >();
var indelLookup = new ConcurrentDictionary <string, IndelEvidence>();
var regionLength = endPosition - adjustedStartPosition;
var numBins = (regionLength / messySiteWidth) + 1000;
var totalBinCounts = InitializeTotalBinCounts(numBins);
var singleMismatchBinCounts = InitializeSingleMismatchBinCounts(numBins);
var actBlockFactory = _actionBlockFactoryProvider.GetFactory(startPosition, endPosition,
adjustedStartPosition, totalBinCounts, singleMismatchBinCounts, numBins, allToWaitFor);
foreach (var classification in classifications)
{
var toWaitFor = GetAndLinkPerClassificationBlocksWithEcFinalization(pairClassifierBlock, classification, indelLookup);
var doStitch = !_geminiOptions.SkipStitching && TypeClassifier.ClassificationIsStitchable(classification);
var categoryIsRealignable = _categoriesForRealignment.Contains(classification);
// Even if we're not going to realign these reads, they may still be useful for bin evidence, so don't give them the immediate flush
var shouldCollectBinEvidence = TypeClassifier.MessyTypes.Contains(classification) || TypeClassifier._indelTypes.Contains(classification);
var isSingleMismatch = _geminiOptions.AvoidLikelySnvs && (classification == PairClassification.SingleMismatchStitched ||
classification == PairClassification.UnstitchSingleMismatch);
if (!(categoryIsRealignable || doStitch || shouldCollectBinEvidence))
{
var actBlock = actBlockFactory.GetEarlyFlushBlock(classification, isSingleMismatch);
foreach (var transformBlock in toWaitFor)
{
transformBlock.LinkTo(actBlock, new DataflowLinkOptions()
{
PropagateCompletion = true
});
}
var toRemove = allToWaitFor.Keys.Where(x => x.IsCompleted);
foreach (var task in toRemove)
{
allToWaitFor.TryRemove(task, out _);
}
if (!allToWaitFor.TryAdd(actBlock.Completion, 1))
{
throw new Exception("Failed to add task.");
}
}
else
{
var actBlock = actBlockFactory.GetActionablePairsBlock(classification, pairResultLookup);
foreach (var transformBlock in toWaitFor)
{
transformBlock.LinkTo(actBlock, new DataflowLinkOptions()
{
PropagateCompletion = true
});
}
if (!allToWaitFor.TryAdd(actBlock.Completion, 1))
{
throw new Exception("Failed to add task.");
}
}
}
var finalTask = AggregateTask(indelLookup, startPosition, endPosition, isFinalTask, _progressTracker);
var intermediateWriterTask = new TransformBlock <AggregateRegionResults, List <BamAlignment> >(results =>
{
edgeStates.AddOrUpdate(startPosition, results.EdgeState, (s, e) =>
{
Logger.WriteWarningToLog($"Edge state already exists: {s}.");
return(results.EdgeState);
});
return(results.AlignmentsReadyToBeFlushed);
}, new ExecutionDataflowBlockOptions()
{
EnsureOrdered = false
});
var finalWriteTask = actBlockFactory.GetWriterBlock();
finalTask.LinkTo(intermediateWriterTask, new DataflowLinkOptions()
{
PropagateCompletion = true
});
intermediateWriterTask.LinkTo(finalWriteTask, new DataflowLinkOptions()
{
PropagateCompletion = true
});
if (edgeToWaitOn.ContainsKey(prevBlockStart))
{
if (!allToWaitFor.TryAdd(edgeToWaitOn[prevBlockStart], 1))
{
throw new Exception("Failed to add task for previous edge.");
}
}
else
{
Logger.WriteToLog($"At {startPosition}, prev block is {prevBlockStart}, nothing to wait on.");
}
if (!isFinalTask)
{
edgeToWaitOn.AddOrUpdate(startPosition, intermediateWriterTask.Completion, (s, e) =>
{
Logger.WriteWarningToLog($"Edge state task already exists: {s}.");
return(intermediateWriterTask.Completion);
});
}
var allTasks = allToWaitFor.Keys.ToList();
var t = Task.WhenAll(allTasks)
.ContinueWith(_ =>
{
if (_lightDebug)
{
Logger.WriteToLog($"Preparing for aggregation for region {_chrom}:{startPosition}-{endPosition}.");
}
if (allTasks.Any(x => x.Status != TaskStatus.RanToCompletion))
{
Logger.WriteToLog("ERROR: Task did not complete.");
foreach (var task in allTasks)
{
Logger.WriteToLog($"{task.Id}\t{task.Status}\t{task.Exception}");
if (task.Status == TaskStatus.Faulted)
{
// Pass the exception along to the final task so it can be forced to error out.
finalTask = ForceFailFinalTask(intermediateWriterTask, task.Exception);
}
}
}
var numStillToProcess = 0;
foreach (var item in pairResultLookup)
{
effectiveMax = Math.Max(effectiveMax, item.Value.Max(x => x.ReadPair.MaxPosition));
numStillToProcess += item.Value.Count;
}
if (_lightDebug)
{
Logger.WriteToLog($"Preparing edge state info for region {_chrom}:{startPosition}-{endPosition}.");
}
EdgeState edgeState = null;
var extraBins = 0;
if (edgeStates.ContainsKey(prevBlockStart))
{
edgeStates.Remove(prevBlockStart, out edgeState);
if (edgeState.EdgeIndels.Any() || edgeState.EdgeAlignments.Any())
{
var newAdjustedStartPosition = RoundedStartPosition(Math.Min(adjustedStartPosition, edgeState.EffectiveMinPosition), messySiteWidth);
extraBins = (adjustedStartPosition - newAdjustedStartPosition) / messySiteWidth;
adjustedStartPosition = newAdjustedStartPosition;
}
}
allToWaitFor.Clear();
allTasks.Clear();
if (_lightDebug)
{
var totalReadsInRegion = _categoryLookup.Values.Sum();
Console.WriteLine($"STILL TO PROCESS IN REGION ({startPosition}-{endPosition} (eff:{effectiveMax})): {numStillToProcess}");
Console.WriteLine(
$"READS IN REGION ({startPosition}-{endPosition} (eff:{effectiveMax})): {totalReadsInRegion}");
foreach (var kvp in _categoryLookup)
{
Console.WriteLine(
$"CATEGORYCOUNT ({startPosition}-{endPosition} (eff:{effectiveMax})): {kvp.Key}: {kvp.Value} ({Math.Round(kvp.Value * 100 / (float)totalReadsInRegion)}%)");
}
}
var totalNumBins = numBins + extraBins;
var allHits = new uint[totalNumBins];
var singleMismatchHits = new uint[totalNumBins];
for (var i = 0; i < totalNumBins; i++)
{
var newBin = i + extraBins;
if (newBin >= totalNumBins)
{
break;
}
if (totalBinCounts[i] > 0)
{
allHits[newBin] = totalBinCounts[i];
singleMismatchHits[newBin] = singleMismatchBinCounts[i];
}
}
if (_lightDebug)
{
Logger.WriteToLog(
$"Creating bin evidence for region {_chrom}:{startPosition}-{endPosition}.");
}
var binEvidence = _binEvidenceFactory.GetBinEvidence(totalNumBins, adjustedStartPosition);
binEvidence.SetSingleMismatchHits(singleMismatchHits);
binEvidence.AddAllHits(allHits);
if (_lightDebug)
{
Logger.WriteToLog($"Adding edge hits for region {_chrom}:{startPosition}-{endPosition}.");
}
if (edgeState != null)
{
var edgeBinInNew = binEvidence.GetBinId(edgeState.EffectiveMinPosition);
var edgeBinInOld = edgeState.BinEvidence.GetBinId(edgeState.EffectiveMinPosition);
AddEdgeHits(edgeState, binEvidence, edgeBinInOld, edgeBinInOld - edgeBinInNew);
}
if (_lightDebug)
{
Logger.WriteToLog($"Done adding edge hits for region {_chrom}:{startPosition}-{endPosition}.");
}
var finalState = new RegionDataForAggregation()
{
BinEvidence = binEvidence,
PairResultLookup = pairResultLookup,
EdgeState = edgeState,
EffectiveMaxPosition = effectiveMax,
EffectiveMinPosition = adjustedStartPosition
};
finalTask.Post(finalState);
finalTask.Complete();
});
return(new[] { t, finalWriteTask.Completion });
}
private static List <HashableIndel> GetFinalizedIndelsForChrom(string chrom, Dictionary <string, List <HashableIndel> > finalizedIndels, EdgeState edgeState)
{
var finalizedIndelsForChrom = finalizedIndels.ContainsKey(chrom)
? finalizedIndels[chrom]
: new List <HashableIndel>();
if (edgeState != null)
{
finalizedIndelsForChrom.AddRange(edgeState.EdgeIndels);
finalizedIndelsForChrom = finalizedIndelsForChrom.Distinct().ToList();
}
return(finalizedIndelsForChrom);
}