ebanks 438d21842a The new recalibrator had been mimicking the behavior of the old one in that if there was no dinuc available (following a no-call base or at either end of a read), it didn't try to recalibrate. Now that Ryan has modularized the system, we no longer need to skip the base completely (we just need to skip the dinuc value)... which is good because the Picard people complained after realizing that cycle #1 never got recalibrated. ... The major effects of this commit are as follows: 1. We no longer skip any good bases (of course, this change alone breaks every single integration test). 2. The dinuc covariate returns a "no dinuc" value for the first base of a read (but not for the last base anymore, since there is a valid dinuc) or if the previous base is a bad base (e.g. 'N'). I've done a bunch of testing on real data and everything looks right; however, let's wait until the recalibrator guru gets back from vacation next week and can double-check everything before shipping this out in another early access release. git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@2443 348d0f76-0448-11de-a6fe-93d51630548a		2009-12-24 20:41:29 +00:00
..
datasources	Updated, optimized REadBackedPileup. Updated test that was breaking the build -- it created a pileup from reads without bases...	2009-11-25 23:30:39 +00:00
executive	Added IterableLocusContextQueue and test. Cleaned up tests, adding BaseTest where it didn't exist. Enhanced test runner to run only classes ending in ...Test.java, so that utility classes can sit alongside the tests but won't be run by JUnit.	2009-05-13 21:32:05 +00:00
io	Fix for problem where err stream remapped to output stream in certain cases, (hopefully) completing Matt's hat trick of fail. Thanks, unit tests.	2009-09-16 08:33:56 +00:00
iterators	a fix for a bug Andrey discovered: in read-based interval traversals we're dupplicating reads in rare cases. The problem was that to accomidate a bug in SAM JDK indexing, we were forced to add one to the stop of our QueryOverlapping() calls to ensure we always got all of the overlapping reads.	2009-11-05 05:26:33 +00:00
refdata	Changed the GLF record to store it's contig name and position in each record instead of in the Reader. Integration tests all stay the same.	2009-12-18 22:54:56 +00:00
traversals	Fix	2009-12-17 17:55:00 +00:00
walkers	The new recalibrator had been mimicking the behavior of the old one in that if there was no dinuc available (following a no-call base or at either end of a read), it didn't try to recalibrate. Now that Ryan has modularized the system, we no longer need to skip the base completely (we just need to skip the dinuc value)... which is good because the Picard people complained after realizing that cycle #1 never got recalibrated.	2009-12-24 20:41:29 +00:00
GATKArgumentCollectionTest.java	Better integrate the @ArgumentCollection into the command-line argument parser. Walkers can now specify their own @ArgumentCollections. Also cleaned up a bit of the CommandLineProgram template method pattern to minimize duplicate code.	2009-09-29 22:23:19 +00:00