CollectAlignmentSummaryMetrics fails on some small inputs due to unsafe Histogram methods

[rickymagner]

Bug Report

Affected tool(s)

CollectAlignmentSummaryMetrics, and maybe others which use the Histogram class from htsjdk.

Affected version(s)

• Latest public release version [3.0.0]

Description

When running this tool on a small bam without good alignments (but properly formatted), the tool throws a NullPointerException. A more useful error message would be appreciated to help users debug what went wrong.

Steps to reproduce

Save the following sam file to null.sam:

@HD	VN:1.6	GO:none	SO:queryname
@RG	ID:test	SM:test_sample	LB:test_lib	PL:ILLUMINA	PU:my_platform	CN:BI	DT:2023-01-01T00:00:00-0500	DS:description
aln1	516	*	0	0	*	*	0	0	AGCTACTG	99999999	RG:Z:test
aln2	516	*	0	0	*	*	0	0	GTCAGTCA	99999999	RG:Z:test
aln3	516	*	0	0	*	*	0	0	CCTTGGAA	99999999	RG:Z:test

Then run

picard CollectAlignmentSummaryMetrics -I null.sam -R <some_ref>.fasta -O test.metrics

to see the error:

Exception in thread "main" java.lang.NullPointerException: Cannot invoke "java.util.Map$Entry.getValue()" because the return value of "java.util.NavigableMap.firstEntry()" is null
	at htsjdk.samtools.util.Histogram.getMin(Histogram.java:438)
	at picard.analysis.AlignmentSummaryMetricsCollector$IndividualAlignmentSummaryMetricsCollector.finish(AlignmentSummaryMetricsCollector.java:271)
	at picard.analysis.AlignmentSummaryMetricsCollector$GroupAlignmentSummaryMetricsPerUnitMetricCollector.finish(AlignmentSummaryMetricsCollector.java:185)
	at picard.metrics.MultiLevelCollector$AllReadsDistributor.finish(MultiLevelCollector.java:208)
	at picard.metrics.MultiLevelCollector.finish(MultiLevelCollector.java:324)
	at picard.analysis.CollectAlignmentSummaryMetrics.finish(CollectAlignmentSummaryMetrics.java:176)
	at picard.analysis.SinglePassSamProgram.makeItSo(SinglePassSamProgram.java:177)
	at picard.analysis.SinglePassSamProgram.doWork(SinglePassSamProgram.java:94)
	at picard.cmdline.CommandLineProgram.instanceMain(CommandLineProgram.java:289)
	at picard.cmdline.PicardCommandLine.instanceMain(PicardCommandLine.java:104)
	at picard.cmdline.PicardCommandLine.main(PicardCommandLine.java:114)

Expected behavior

A more useful error message should be given in this case, or an empty metrics file produced (if appropriate).

Actual behavior

An unhelpful stacktrace is given instead.

More Details

The reason this is happening is related to this issue in htsjdk. One fix is to have this Picard tool be careful about how it calls the htsjdk Histogram methods and essentially ensure it never uses them unless it's safe. Another fix is to have the Histogram methods in htsjdk throw an exception in this sort of case, and make the Picard tools affected have to handle them appropriately, though this feels like a lot of trouble to update every tool depending on it.

[cmnbroad]

@kockan Waiting for response to review comments.