The current VCFReader implementation depends on cyvcf2 to parse the underlying file, and then converts each row into a data frame array. However, this is fairly slow. For vcfio to be more useful in VCF annotation flows, the parser needs to be ~5x faster than it is currently.