Update CollectSamErrorMetrics.java

[nh13]

@yfarjoun care to take a look at this? Happy for there to be a test for it, but it ain't my bug.

[nh13]

A few more things I've noticed:

• assumes that the .dict file is present (it may not be)
• assumes base qualities are present (since the default error+stratifiers include them)

[coveralls]

Coverage remained the same at 81.637% when pulling ca9d000 on nh_make_vcf_optional into 0b95405 on master.

[yfarjoun]

hmmm. Collecting error metrics over variants doesn't make much sense...you could give it dbSNP or something like that if you dont' have truth for that sample....

[nh13]

What if my sample has no variants relative the reference? That’s my real use case! No VCF is required. Think sequencing bacteria or synthetic sequences.

[yfarjoun]

OK. could you at least log a warning so that the (less-advanced) user who might be using this against human will be informed that they should be supplying dbSNP or another appropriate source of known variation for that sample?

[nh13]

@yfarjoun I have been trying to think of a way to say that you must intentionally say not to use a VCF. Perhaps we could instead make the option required, but if the value is "<EMPTY>" then it means there are no variants to consider? I'd prefer that to an empty VCF.

[yfarjoun]

Alternatively you could have a hidden boolean variable ALLOW_MISSING_VCF_DESPITE_THIS_NOT_BEING_RECOMMENDED (or something)

…

On Wed, Jun 5, 2019 at 11:09 PM Nils Homer ***@***.***> wrote: @yfarjoun I have been trying to think of a way to say that you must intentionally say not to use a VCF. Perhaps we could instead make the option required, but if the value is "<EMPTY>" then it means there are no variants to consider? I'd prefer that to an empty VCF. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[kockan]

Superseded by #1476