If a read was hard clipped for being low quality and no does not overlap the interval anymore, this read will now be discarded instead of treated as an error by the GATK traversal engine.
Clips both tails of a read if the tails are below a given quality threshold (default Q2).
*Added special treatment for reads that get completely clipped.
When a read is clipped once, and then in the second operation, because of indels, it doesn't reach the coordinate initially set for hard clipping, the indices were wrong. This should fix it.
When a read had been previously soft clipped, the UnclippedEnd could not be used directly as Reference Coordinate for clipping , because the read does not go that far.
The clipping op extra information being kept by this walker was specific to the walker, not to the read clipper. Created a wrapper ReadClipperWithData class that keeps the extra information and leaves the ReadClipper slim.
(this is a quick commit to unbreak the build, performing integration tests and will make further commits if necessary)
* Added reference coordinate based hard clipping functions. This allows you to set a hard cut on where you need the read to be trimmed despite indels.
* soft clipping was messing up cigar string if there was already a hard clip at the beginning of the read. Fixed.
* hard clipping now works with previously hard clipped reads.
* Hard clipping was wrongfully hard clipping unmapped reads while soft clipping then hard clipping mapped reads. Now we throw exception if we try to hard/soft clip unmapped reads and use the soft->hard clip procedure fore every mapped read.
* Interval containment needed a <= and >= to make sure it caught the borders right.
reads are clipped in map() and now we cover almost all cases. Left behind the case where the read stretches through two intervals. This will need special treatment later.
-To correctly handle indels and MNPs, only consider features that start at the current locus,
rather than features that span the current locus, when selecting the most significant effect.
-Throw a UserException when a SnpEff rodbinding is not provided instead of simply not adding
any annotations and silently returning.
-- Verified now to be correct at runtime
-- UnitTest covers this
-- createTypeDefault now takes a Type, not a Class, so that parameterized classes can have their parameter fetched in the defaults.
Mauricio Carneiro
David Roazen
Guillermo del Angel
Menachem Fromer
Eric Banks
Ryan Poplin
Mark DePristo