James Bonfield

Well, I started down this road and I think it may well be a rather large exercise in futility! What I have so far (work in progress, be warned) is...

> checking for BZ2_bzBuffToBuffCompress in -lbz2... yes Clearly the configure script found libbz2 on your system, so the config.log file should indicate where (it may take a bit of hunting)....

If anyone knows how to prevent github from uploading its own tarballs we'd be glad to hear! We always add our own, which is the working one, in addition to...

Correct, VCF indices, like BAM, have overlap calculations rather than point sources, so you can find all variants that overlap a region. In bcftools, if you use a combination of...

I've experimented with Zstd and CRAM before. It's a definite win over zlib, at some compression levels both in speed and size. I experimented on this in Staden io_lib for...

Regarding FASTA compression, it's really a special case and not so helpful here. Also note that the higher levels mainly work better by using larger block sizes, harming random access....

The pipe transfer point is an interesting one as if you're only talking about data transfer then you'll be willing to use any format as it's temporary. However if it's...

No I haven't, nor have I looked into how best to create a dictionary. I expect it wouldn't make much difference except for small block sizes (ie when supporting a...

I personally think it would be emminently sensible, although I'd have to check with others if it would be accepted. Certainly for new projects I've tried to push for such...

Agreed. How would you envisage the dictionary is used? There is a Zstd command (and I assume associated function) to compute a dictionary from a file, so we could build...