read compressed alignment files?

[cmccoy]

A low priority request/question: can pplacer and friends be made to read compressed alignment files (eg, .gz or .bz2) natively? This could potentially save a lot of room on disk for work in progress.

[matsen]

Would Zip suit? I know it's not hip, even a little bit, but we are already using it.

https://github.com/matsen/pplacer/blob/dev/pplacer_src/refpkg_parse.ml

[cmccoy]

camlzip supports zip and gzip; we could use that without adding
dependencies.

On Mon, Nov 18, 2013 at 11:12 AM, Erick Matsen notifications@github.comwrote:

Would Zip suit? I know it's not hip, even a little bit, but we are already
using it.

https://github.com/matsen/pplacer/blob/dev/pplacer_src/refpkg_parse.ml

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/323#issuecomment-28727910
.

[matsen]

Ah, nice. I'll have a go on this some afternoon.

On Mon, Nov 18, 2013 at 11:16 AM, Connor McCoy notifications@github.comwrote:

camlzip supports zip and gzip; we could use that without adding
dependencies.

On Mon, Nov 18, 2013 at 11:12 AM, Erick Matsen notifications@github.comwrote:

Would Zip suit? I know it's not hip, even a little bit, but we are
already
using it.

https://github.com/matsen/pplacer/blob/dev/pplacer_src/refpkg_parse.ml

—
Reply to this email directly or view it on GitHub<
https://github.com/matsen/pplacer/issues/323#issuecomment-28727910>
.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/323#issuecomment-28728274
.

Frederick "Erick" Matsen, Assistant Member
Fred Hutchinson Cancer Research Center
http://matsen.fhcrc.org/

[nhoffman]

gzip would probably be preferable for single files if available

[matsen]

Camlzip knows how to read bytes, characters, and sets thereof. When we read in fasta (e.g.) files, they get read in line by line and tokenized (see ppatteries for the definition of gen_parsers). It's important for the tokenize functions that things arrive a line at a time. We could read chunks of our compressed file at a time (say, 80 chars), look for newlines in them and spit out an Enum of lines as the newlines appear. Does that seem reasonable? Will that be sufficiently efficient?

[cmccoy]

Maybe we could hook into the Batteries I/O interface? IO.create_in (for camlzip in_channel) combined with IO.lines_of would give an Enum of strings.

[matsen]

Excellent!

On Mon, Nov 18, 2013 at 5:19 PM, Connor McCoy notifications@github.comwrote:

Maybe we could hook into the Batteries I/O interface? IO.create_inhttp://ocaml-batteries-team.github.io/batteries-included/hdoc2/BatIO.html#VALcreate_in(for camlzip
in_channel) combined with IO.lines_ofhttp://ocaml-batteries-team.github.io/batteries-included/hdoc2/BatIO.html#VALlines_ofwould give an
Enum of strings.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/323#issuecomment-28757285
.

Frederick "Erick" Matsen, Assistant Member
Fred Hutchinson Cancer Research Center
http://matsen.fhcrc.org/

[cmccoy]

This is working for me on the microbiome demo - sequence files ending in .gz get decompressed with gzip.

I added .jplace.gz compression support while I was there. Doing so required changing the JSON parser from acting on raw input channels to a wrapped Batteries IO.input (see b4e10c4) or IO.output (see d4701a0).
I'm hoping that doesn't incur any serious performance overhead. Happy to test more or drop those parts.

[matsen]

Whiplash. Nice work, and glad to see tests.