I have created a conda package for pyrad that pulls in ALL deps (muscle and vsearch too)

[xguse]

Because I generally work on multiple clusters, I have to replicate my scientific software stack not infrequently and I generally hate my life for a few days each time I have to do this. I used to use environment modules etc in part to try to manage this. Since making the switch to anaconda python and the conda build/package manager, all I have to do is install anaconda, clone the basic install to a fresh env, and run a few conda install xxx commands and I am done in less than an hour -- EVEN MY NON-PYTHON programs.

This has allowed me to write a pyrad package recipe that installs all of pyrad's python deps AND pre-compiled versions of vsearch and muscle at the same time. I have built a quick github pages website to host the repository of my personal pre-built packages which includes one for pyrad. The landing page explains how to add the repo to your .condarc file and lets you browse the few packages I have so far built and hosted.

I am listing it here both because I think that conda is the way of the future and simply because I believe that some folks may benefit from the easy pyrad complete-install package it hosts.

Thanks again to Deren for the tool. Hopefully EVEN more folks will use it as it becomes even easier to deploy!

http://xguse.github.io/conda-package-repo/

Gus

[dereneaton]

Hey Gus, this is a great idea. I totally agree about conda, and have been toying with this exact idea.

One important note, I believe that when done this way we need to include the vsearch LICENSE along with the binary.

I will add your method as an install option on the README for now, and try to transition towards conda being the primary install method in the future.

Because I'm mostly developing on the ipyrad package now, the next release of pyrad (3.1.0) should hopefully remain stable for a while.

[xguse]

That is a good point about the license. The vsearch recipe's meta.yaml file lists its license but I think its not a big deal to have a copy of the doc tag along with the install. At least in the linux version. I have ZERO experience with windows batch files. But the build.sh file can just do something like install -D path/to/src/LICENSE path/to/pkg/loc/LICENSE I believe.

FWIW: here is my fork of the conda-recipes repo that I add my personal build recipes to: https://github.com/xguse/conda-recipes

I dont usually merge even with my fork's master though. Rather I use a personal feature branch: f/personal_repo. Not really sure why I dont use my fork's master to tell the truth. I do for other forks. But you should know that if you are looking for recipes not already in upstream's master.

[xguse]

ok upon researching an appropriate place to store the licence upon installation I have settled upon:

$PREFIX/share/doc/pkg_name/LICENSE

Does this fit with your understanding? Its where libxml and libxlst store their copyright documents. If you have a better place let me know and I will change it. I am updating the build-recipe and pushing a new build version to my repo. I want to do thins while it is on my mind or I will forget.

Gus

EDIT: completed.