reading and then writing doesn't produce byte identical output?

[eroller]

I think this should be a requirement. I noticed the following differences from just reading in and writing out a vcf file:

1. header lines are in different order
2. INFO fields are in different order
3. 0 padding is added to fields (e.g. 2 becomes 2.0)
4. truncate sample genotype columns are replaced with missing "." for each field (e.g. "." becomes ".:.:.:.:.:." depending on the number of fields)

Neither of these breaks the vcf specification, but I hope this library could be updated to follow this requirement. As modifications are made to the vcf using this library I want to see only the changes that were made, which means all these other trivial modifications are just noise.

[martijnvermaat]

Hi @eroller, thanks for the report!

While I'm not particularly concerned about this (especially the first two seem purely cosmetic), I can see that this would be nice to have. I'd review and merge PRs for this if they are not too invasive otherwise.

[jamescasbon]

This is, I believe, a very hard requirement. I think it needs a source map plus change tracking.

VCF writing was really not proper, so this would possibly need a lot of changes to the parser as well.