[ADAM-937] Adding check for aligned read predicate or limit projection flags and non-parquet input path

[heuermh]

Fixes #937

[heuermh]

I'm not sure this is comprehensive enough, e.g. sometimes even Parquet files won't have .seqdict files

$ ./bin/adam-submit fasta2adam \
  adam-cli/src/test/resources/contigs.fa contigs.contig.adam

$ ./bin/adam-submit transform -aligned_read_predicate \
  contigs.contig.adam output.adam

Command body threw exception:
java.io.FileNotFoundException: File contigs.contig.adam.seqdict does not exist
  Exception in thread "main" java.io.FileNotFoundException:
    File contigs.contig.adam.seqdict does not exist

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/ADAM-prb/1079/
Test PASSed.

[fnothaft]

Sorry; should be checking that the file is a Parquet file of alignment records.

Alternatively, we could make this a ValidationStringency effect. E.g., if you have ValidationStringency.LENIENT, we give a RecordGroupDictionary.empty/SequenceDictionary.empty if they're missing.

[heuermh]

Sorry; should be checking that the file is a Parquet file of alignment records.

What do you mean by this, flipping the logic of isNonParquet around?

[fnothaft]

What do you mean by this, flipping the logic of isNonParquet around?

Oh, I just meant that it shouldn't check for any Parquet dataset. Rather, the check should be more specific and check that it is a Parquet dataset of AlignmentRecords.

[heuermh]

Can't do that by filename check, so it would have to come later, instead of fail fast in the cli class.

[fnothaft]

Ah, yes, fair. I think we suggest "*.contig.adam" as a convention for files of NucleotideContigFragments, but we don't enforce it anywhere. We can patch that later, if needed.

Do you want to do any of the ValidationStringency stuff I suggested earlier, or shall we skip that for now and possibly come back to it later?

[heuermh]

I think we suggest "*.contig.adam" as a convention for files of NucleotideContigFragments, but we don't enforce it anywhere.

Yep, I cribbed from here
https://github.com/bigdatagenomics/adam/blob/master/adam-core/src/main/scala/org/bdgenomics/adam/rdd/ADAMContext.scala#L858

I can add the check for *.contig.adam if useful. I think validation strategy might be overkill.

[fnothaft]

I can add the check for *.contig.adam if useful. I think validation strategy might be overkill.

This is probably sufficient for now. We can add the two later if need be. I'll merge this shortly.

[massie]

Thanks, Michael.

[massie]

Merged manually.