Add a flag to preserve the call.name field

If not using the sample optimized tables, but only using the variant optimized tables, it would be nice for vcf_to_bq to produce a table with the name field in the call record instead of the sample_id.

One can translate the sample_id to the name, and the table with the sample_id is going to be smaller than one with the name (typically), but the SQL complexity and overhead of doing the translation does not seem worthwhile for all users.

Adding a flag, such as --preserve-call-name (boolean) would be one option. That said, it might be nice to allow users to specify what fields they want in the call record. For example:

--call-fields="name,genotype,DP,AD"

The tricky case would be around the sample_id and name and making the most common cases straight-forward. It would be nice not to have to specify a full list of fields just to pick up the call name instead of the sample_id, so I propose the following:

• All fields, replace sample_id with name:

    --call-fields="name,*"
  

• All fields, replace name with sample_id (the current default behavior):

    --call-fields="sample_id,*"
  

• All fields, include both name and sample_id:

    --call-fields="sample_id,name,*"
  

I’m not really sure when someone would use the last option, so I’m not opposed to it being unavailable.

You could make the options SQL-like with the SELECT * EXCEPT syntax, but that feels a bit like overkill.

[tneymanov]

This changed was pushed to master in PR #677 and will be included in the next release.

[tneymanov]

Yes, most probably call.name, to keep consistency with v1 schema.