Updating PacBio FALCON - dealing with package version naming problem

[peterjc]

Currently the recipe https://github.com/bioconda/bioconda-recipes/blob/master/recipes/pacbio_falcon/meta.yaml for https://anaconda.org/bioconda/pacbio_falcon uses:

{% set name = "pacbio_falcon" %}
{% set version = "052016" %}

package:
  name: {{ name }}
  version: {{ version }}

about:
  home: 'https://github.com/PacificBiosciences/FALCON'
  license: "Standard PacBio Open Source License"
  summary: "A set of tools for fast aligning long reads for consensus and assembly"

source:
  fn: funzip_052016.tar.gz
  url: https://github.com/PacificBiosciences/FALCON/archive/funzip_052016.tar.gz
  md5: be4a225262ee4bd9729b6b7e81a83615

...

The https://github.com/PacificBiosciences/FALCON/ repository only currently has tags for v0.1, v0.2 and funzip_052016 which seems to be the latest tag:

https://github.com/PacificBiosciences/FALCON/tree/funzip_052016

That was presumably how things stood when the recipe was first written (@bebatut Jan 2017):

https://github.com/bioconda/bioconda-recipes/commit/fd53551df5042f2d554504c752216fcb0cf390a1#diff-9cd1477e255f1bcb7b1f395abbdf2d30

The README is outdated https://github.com/PacificBiosciences/FALCON/issues/637 but if I understand correctly releases since have been via https://github.com/PacificBiosciences/FALCON-integrate which includes tags for 0.2.2, 0.3.0, ..., 2.1.1, 2.1.2, 2.1.4 which appears to be the latest.

It would appear that the tag funzip_052016 (i.e. May 2016, sadly not year month order for sorting) was never intended to be regarded as a version number, and instead the major.minor.revision pattern is being used.

Unfortunately, we now have a BioConda package with version 052016 - how will this behave if we try to package FALCON v2.1.4 using 2.1.4 as the version?

Can we delete or tag the current package in some way as not to be used? Thus far https://anaconda.org/bioconda/pacbio_falcon reports only 416 downloads.

[peterjc]

As of #9080, all the recent versions of pypeFLOW, one of the PacBio dependencies, will be one BioConda.

[rvalieris]

Unfortunately, we now have a BioConda package with version 052016 - how will this behave if we try to package FALCON v2.1.4 using 2.1.4 as the version?

>> from conda.models.version import VersionOrder
>> VersionOrder('052016') > VersionOrder('2.1.4')
True
>> VersionOrder('052016') == VersionOrder('52016.0.0')
True

we will have to remove/mark as broken the old release otherwise conda will still pick it as the latest version.

[peterjc]

Thanks @rvalieris, that's what I thought. Once I have 2.1.4 working I think we should do that then (mark the old one as broken).

[peterjc]

We could probably do this for the recent FALCON releases which were on GitHub, but after that, things are likely to change...

It's worse than you think. We no longer even push the latest code into GitHub, so tagging would be pointless. Sorry.

There is a plan to release into Bio-conda, using pre-built wheels or something. Someday.

https://github.com/PacificBiosciences/FALCON/issues/637#issuecomment-399282800

[peterjc]

This seems to have been solved with a new package instead, https://github.com/bioconda/bioconda-recipes/tree/master/recipes/pb-falcon CC @pb-cdunn

Should the old pacbio_falcon package be formally retired?

[pb-cdunn]

Yes, if retirement is possible. I don't know how to do that.

[peterjc]

I can't find anything written down about this but am perhaps struggling to recall the terminology used - @bgruening @daler?

At very least we could update the human readable description of pacbio_falcon to recommend pb-falcon instead?

[bgruening]

There are 3 options. I can mark the old packages as broken. I can move the repository into bioconda-legacy or I can simply remove the repo. Whatever fits your needs :)

[peterjc]

I think moving the pacbio_falcon recipe into bioconda-legacy makes sense, and looking at #6844 it might be worth marking the publish package as broken too?

[bgruening]

Just let me know what I should do :)

[peterjc]

@bgruening are there any policy documents on the role of https://anaconda.org/bioconda-legacy/ to guide us here?

[bgruening]

No police. Sorry, just let me know what you prefer.

[peterjc]

@bebatut as the original author of this recipe fd53551df5042f2d554504c752216fcb0cf390a1 what would you like to happen to it?

[bgruening]

ping @bebatut

[bebatut]

Moving the recipe to bioconda-legacy sounds legitimate