fixed bug in tandem repeat annotation

[davidbenjamin]

@fleharty Here is a quick M2 and HC edge case bug for you.

In the edge case where an alt haplotype starts with an indel, and hence the variant context start is one base before the assembly region due to padding a leading matching base, there was a problem. The variant start was before the reference context window (since in this part of the code the reference context window is pegged to the assembly region, which caused problems in the lines

final int startIndex = vc.getStart() - ref.getWindow().getStart();
        final byte[] refBasesStartingAtVariantLocus = new String(refBases).substring(startIndex).getBytes();

The fix is for refBasesStartingAtVariantLocus to exclude the padding base, which as you can see was done previously in the line

final String refBasesStartingAtVCWithoutPad = new String(refBasesStartingAtVCWithPad).substring(1);

That is, instead of taking a subarray of the reference bytes twice, once from the variant context start with padding and once more to remove the padding, we take a single subarray without the padding.

While I was at it, I replaced some slow conversions from byte[] to String to String back to byte[] with subarray operations.

[droazen]

@fleharty Is this a fix you would want to see in the next release?

[fleharty]

@droazen yes, if that is possible, sorry this slipped through the crack. I'll review this this evening.

[fleharty]

@davidbenjamin Looks great!