Change locuspredicate to unique read mapped predicate #47
Conversation
|
Thanks, @arahuja . Is it possible to make this respect the mapped and nonDuplicate arguments to the function (i.e. only filter out unmapped reads if mapped=True, and only filter out duplicate reads of nonDuplicate=True)? If it's not trivial, we can merge this as is so we fix the build and address this in another commit. |
| @@ -89,7 +89,7 @@ object Common extends Logging { | |||
| * @return | |||
| */ | |||
| def loadReads(args: Arguments.Reads, sc: SparkContext, mapped: Boolean = true, nonDuplicate: Boolean = true): RDD[ADAMRecord] = { | |||
| var reads: RDD[ADAMRecord] = sc.adamLoad(args.reads, Some(classOf[LocusPredicate])) | |||
| var reads: RDD[ADAMRecord] = sc.adamLoad(args.reads, Some(classOf[UniqueMappedReadPredicate])) | |||
| progress("Loaded %d reads.".format(reads.count)) | |||
| if (mapped) reads = reads.filter(read => read.readMapped && read.contig.contigName != null && read.contig.contigLength > 0) | |||
There was a problem hiding this comment.
we should be able to get rid of lines 94 and 95, right?
There was a problem hiding this comment.
Yes, those should be redundant with the filter - or we just remove the filter for now if you want the functions arguments to be respected.
|
@timodonnell Yea, we still can't have runtime parameters to the Parquet predicates - they need to be predefined classes. I was playing with a new ParquetInputFormat, but not sure if it'll work, hopefully coming soon The way around this for now would be to have predefined classes for the 4 combinations and then load based on the argument parameters. |
Change locuspredicate to unique read mapped predicate
This was changed in: bigdatagenomics/adam#234