Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/bdg-formats-prb/101/
Test PASSed.

see my comment bigdatagenomics/adam#1039 (comment)
about moving sampleDescription and processingDescription into this Sample

@heuermh - if you want to keep this PR focused on schema you can assign an issue to me to update the ETL code to load and save using the new Sample and with these fields in Genotype moved.

How are those fields populated? What do they mean? Do they map to SRA metadata fields, which per earlier conversation and the doc comments all go into attributes?

How are those fields populated?

VCF spec, in context of "5.4.10 Sample mixtures" has

##SAMPLE=<ID=Blood,Genomes=Germline,Mixture=1.,Description="Patient germline genome">

but I'm not so sure that is going into one of these fields.
I'd suspect they don't get populated now.
However, these two fields sampleDescription and processingDescription seems reasonable enough in how they are currently described in comments in bdg-formats genotype and not expensive if they are moved to Sample - and can be null

The cardinality for Genomes, Mixture, and Description is 0..*, so they would need to go in array fields. Or alternatively a single array of Genome records, which would have name, mixture, and description fields.

However, the doc already specifically recommends those go in attributes, where cardinality doesn't matter (repeated keys are ok).

the doc already specifically recommends those go in attributes

putting in attributes is fine with me

Although now that I've said that repeated keys in attributes are ok, they are not, since we're using an avro map<string>. In which case last in wins.

I'm not so convinced that these two fields are coming from this sample mixture related tag anyhow, or that they have any source in VCF currently, so maybe wait to worry about modeling Genomes, Mixture, Description cardinality for now until we know more.

I don't think these fields get populated anyhow - I am more concerned for the principle that in case sample level metadata like this appears in future and somehow are shoe-horned into these fields that it not appear in our inner-inner loop inside of Genotype

Do we want to get this into the next bdg-formats release? CC @heuermh @jpdna

Yes, and #83. I'll rebase to resolve conflicts in a sec

Attributes with duplicate keys are going to get clobbered. That wasn't a major issue in the Feature record, where this also came up, but will be here. The cardinality of the VCF sample attributes and nearly all of the SRA attributes are 0..*.

Is this a reasonable workaround?

  // ...
  array<Attribute> attributes = [];
}

record Attribute {
  string key;
  string value;
}

or do we want to reopen discussion about a bunch of array fields?

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/bdg-formats-prb/102/
Test PASSed.

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/bdg-formats-prb/104/
Test PASSed.

Thanks!