New class handles (vastly more cleanly) the db of tribble codecs, features, and names for use throughout the GATK.
Added SelfScopingFeatureCodec interface that allows a FeatureCodec to examine a file and determine if the file can be parsed. This is the first step towards allowing the GATK to dynamically determine the type of a RodBinding.
-- CombineVariants now uses the new RodBinding syntax, -V / --variants. Passed all integration tests on first run
-- Exposed gapping bug in the List<RodBinding<T>> system now fixed. ParserEngine now has a addRodBinding() that is called by RodBindingArgumentTypeDescriptor when it encounters each RodBinding. This allows the system to work with collection types that are recursively parsed by the system.
-- Cleaning up old interface to RMDT, docs and contracts added
-- Proper type checking for RodBinding for cases where the Tribble type isn't found or is the wrong type
- Floating point column widths are measured correctly
- Using fixed width columns instead of white space separated which allows spaces embedded in cell values
- Legacy support for parsing white space separated v0.1 tables where the columns may not be fixed width
- Enforcing that table descriptions do not contain newlines so that tables can be parsed correctly
Replaced GATKReportTableParser with existing functionality in GATKReport
-- support for explicit naming of bindings (-X:name,type x)
-- support for automatic naming of bindings in lists (-X:vcf foo.vcf -X:vcf bar.vcf will generate internal names X and X2)
-- ParserEngineUnitTest expanded to cover all of the Rodbinding cases
-- RodBindingUnitTest tests all of the low-level accessors
-- Parsing engine throws UserExceptions when bad bindings are provided on the command line
initial size for a Java hash table (HashMap, HashSet, etc.) given an expected
maximum number of elements. The optimum size is the smallest size that's
guaranteed not to result in any rehash / table-resize operations.
Example Usage:
Map<String, Object> hash = new HashMap<String, Object>(Utils.optimumHashSize(expectedMaxElements));
I think we're paying way too heavy a price in unnecessary rehash operations across
the GATK. If you don't specify an initial size, you get a table of size 16 that gets
completely rehashed and doubles in size every time it becomes 75% full. This means you
do at least twice as much work as you need to in order to populate your table:
(n + n/2 + n/4 + ... 16 ~= (1 + 1/2 + 1/4...) * n ~= 2 * n
This was no longer working properly given that DBSNP indels needed to be moved around. The adaptor system is being refactored and you will need to convert files from X -> VCF for many tools to work.
RodBinding no longer duplicates the get() methods in RMDT. This is just an object now that connects the command line system to the RMDT.
Updated programs to use new style
Added UnitTests for the RodBinding accessors.
You know have to provide an explicit list of RODRecordLists upfront to the constructor. RefMetaDataTracker is now immutable. Changes in engine to incorporate these differences
Extensive UnitTests for RefMetaDataTracker now.
On the path towards converging getVariantContext() and getValues() in tracker so that we can have a single approach to get values from RODs with the new RodBinding() types
Problem is that Novelty sees multiple records at a site (SNP, INDEL) to calculate whether a site is novel, but VariantEvalWalker makes an arbitrary decision which to use for analysis and CompOverlap may not see a comp record of the same type as eval. So you get lines where the stratification is known but there are 10 novel sites!
It's no longer possible to provided allowed types, as this was a very rarely used feature in the engine. These get methods have been removed and local uses replaced with tests directly in their code. This simplified the RefMetaDataTracker significantly
VariantContextRodBinding now forwards along all of the RefMetaDataTracker methods, so it is possible to create a full equivalent VariantContextRodBinding now as a walker field variable.
All walkers updated to the new RefMetaDataTracker function call style
Renamed many functions to more clearly state what they are actually doing
Removed unnecessary / unused functionality, reducing interface complexity
Updated all uses of this code in GATK
Added generic, type-safe accessors to RefMetaDataTracker such as public <T> List<T> getValues(final String name, Class<T> clazz)
Added standard refMetaDataTracker accessors to RodBinding, so you can do everything you can for generic rods with the tracker directly with with the RodBinding
*New style.css, new template for the walker auto-generated html. Short description is no longer repeated in the long description of the walker.
*Updated DiffObjectsWalker and ContigStatsWalker as "reference" documented walkers.
overlaps exactly on disk with the first shard in contig n+1, the shards
would be merged together to avoid duplicate extraction. Unfortunately,
the interval overlap filter couldn't handle shards spanning contigs, and
was choosing to filter out reads from contig n+1 which should have been
included.
I'm not completely sure why the BAM indexing code would ever specify that the
end of one chromosome had the same on-disk location as the start of the next
one. I suspect that this is a indexer performance bug.
Initial version of RodBinding class.
Flow from walker Rodbinding @Arguments -> RMDTriplet (old system) -> GATK engine (standard). Will need refactoring.
Removing unused files RODRecordIterator, ReferenceOrderedData, QueryableTrack, RMDTrackCreationException, GATKFeatureIterator, ReferenceOrderedDataUnitTest
Refactored dbSNP and refseq utilities to be closer to the other files implementing these features
David Roazen
Mark DePristo
Eric Banks
Ryan Poplin
Khalid Shakir
Guillermo del Angel
Mauricio Carneiro
Kiran V Garimella
Matt Hanna