Unicycler 0.4.7 - b1 - fixes

[apeltzer]

• I have read the guidelines for bioconda recipes.
• This PR adds a new recipe.
• AFAIK, this recipe is directly relevant to the biological sciences (otherwise, please submit to the more general purpose conda-forge channel).
• This PR updates an existing recipe.
• This PR does something else (explain below).

This is an attempt to fix some of the weird errors seen in Unicycler 0.4.6 and 0.4.7 when using bioconda, e.g. rrwick/Unicycler#140

I tried tracking this down so far but wasn't successful until now - I think the change I removed here in the build recipe could potentially get the issue fixed as reverting to 0.4.4 is also not really an option for some people (e.g. when having/requiring multiple tools in the same container installed).

[biocondabot]

Package(s) built on CircleCI are ready for inspection:

Arch	Package	Repodata
linux-64	unicycler-0.4.7-py37hdbcaa40_1.tar.bz2	repodata.json
osx-64	unicycler-0.4.7-py37hd9629dc_1.tar.bz2	repodata.json

---

You may alsu use conda to install these:

• For packages in linux-64:

  conda install -c https://59432-42372094-gh.circle-artifacts.com/0/tmp/artifacts/packages <package name>
  

• For packages in osx-64:

  conda install -c https://59433-42372094-gh.circle-artifacts.com/0/tmp/artifacts/packages <package name>
  

---

Docker image(s) built:

Package	Tag	Install with docker
unicycler	0.4.7--py37hdbcaa40_1	show curl "https://59432-42372094-gh.circle-artifacts.com/0/tmp/artifacts/images/unicycler%3A0.4.7--py37hdbcaa40_1.tar.gz" | gzip -dc | docker load

[PertuyF]

@apeltzer please see rrwick/Unicycler#140 (comment).
This change was made for a reason, and attempts were made to try fix this issue, be sure you get is all before trying this.

[PertuyF]

Looks like unicycler now builds fine with the genuine seqan 2.3 included in the original repo. Which would suggest the fix introduced in PR #10510 is no longer required.
Am I missing something from the build log or does it look right @corburn @bgruening ?

Could you test this package and confirm you do not experience the stalling issue before we consider merging @apeltzer ? any help appreciated in testing this.

[apeltzer]

Nice one! Thanks a bunch, I'm rebuilding my docker container for nf-core/bacass and will then start a new run to see if this changes things :-) !

[apeltzer]

Couple of issues: The recipe uploaded was built using python3.7:

https://anaconda.org/bioconda/unicycler/files

Although explicitly marked to be not built using this version?

https://github.com/bioconda/bioconda-recipes/blob/master/recipes/unicycler/meta.yaml

[PertuyF]

It could be due to python pinning at the Bioconda scale. Older versions are dropped progressively, so if python 3.6 is not supported anymore it may have forced 3.7.
@bioconda/core am I right? And is there a way to have a lint warning for this, like "Recipe pinning does not match bioconda's pinning for package xxx"?

Is 3.7 an issue for you @apeltzer?

[apeltzer]

Unfortunately yes :-(

I'm trying to build a pipeline and many of the tools required are build with older versions of Python3.X, meaning I can't really install these in the same environment anymore due to Python3.7 being used for the new build of unicycler 0.4.7 - the old (incorrect recipe) gets installed then automatically, which breaks my pipeline due to the stalling issue mentioned above.

[apeltzer]

One could bump all other recipes to Python3.7 however, but not sure that all of these will be built correctly then :-(

[PertuyF]

In most cases rebuilding recipes for a new python 3.x should be painless.
But if required there may be a way to force rebuild unicycler for 3.6. Although I'm not certain.

[dpryan79]

For C/C++ packages, you need to put a skip: True # [something] in the build section for actually skip building. Otherwise I think conda ignores it. Regarding rebuilding dependencies for python 3.7, that should generally work fine (at least if they already have a python 3.6 build). I should probably start that in the bulk branch so you don't have to deal with individual packages.

[apeltzer]

Would be great - its just a bunch of recipes though:

https://github.com/nf-core/bacass/blob/092a082fb08076aed5affffe7d092b2d3d7b75a1/environment.yml

[apeltzer]

• MultiQC
• Quast
• Nanopolish
• Racon
• Porechop
• Nanoplot
• Pycoqc

(I think the rest is already in Python3.7 on conda-forge or bioconda)

[dpryan79]

Are those the ones needing to be rebuilt? At least a few of them are noarch: python (e.g., multiqc), so they should be fine already.

[apeltzer]

Fair point - I probably should definitely update the quast recipe, but the others could be fine.

Yes, that's the complete list of Python packages/python dependent packages.

[dpryan79]

Yeah, the only one of those that needs rebuilding is quast. Note that that had compilation errors the last time I tried to bump everything for new pinnings, so it'll need some significant TLC.

[epruesse]

Quast also has a serious data package issue. Or had, last time I checked. It downloads SILVA somewhere in the middle, unpacks it, then renames the sequences and builds a BLAST index. So the package is incomplete, uses a different dataset than plain SILVA, modified in a trivial and technically uncessary manner, and breaks with clobbered data if you run it in parallel on a cluster. (end rant...)

[apeltzer]

I am quite tempted to drop that tool just because of this. Are there alternatives that can do similar things-did a quick google search but couldn't find much unfortunately... ?

[epruesse]

Not that I am aware of. Double check current versions. It's not like this would be hard to fix.