Guided discussion (remainder of scheduled meeting time)
This list to be updated prior to the meeting to reflect the questions/discussion points proposed in this forum thread
To help us make the best use of everyone’s time please post any discussion topics to this thread before the developer meeting. This will help us identify and prepare for any questions/topics ahead of the call. Topics or questions of discussion are open to anyone, developers or users, but we ask that these discussion points be limited to thematic QIIME 2 questions (e.g. feature proposal, plugin idea, etc.) — we will not be able to provide technical support during these calls.
The developer meeting will be recorded and posted to YouTube, we will update this post with the streaming video link shortly before the meeting.
Sorry we were a little late getting this out this week. In the future we will try to post immediately following the dev meetings, that way there is a full week for discussion.
We are cancelling today’s dev meeting because there haven’t been any agenda items / discussion points added. We will individually follow up here with updates on our end.
There will be a new forum post requesting agenda items for next week’s meeting available shortly.
We have been setting up a Concourse-based continuous integration / continuous deployment system (CI/CD) for QIIME 2. We are planning to have this system in place before the next scheduled release (end of March). This system will automate as many tasks as we can, including integration testing across multiple QIIME 2 repos, building and publishing conda packages, git branching to support concurrent/LTS releases, and cutting release candidates and final releases.
We’re pushing out a beta release for gneiss on Wednesday and everything seems to be working. However, there are some things that would like to bring up at some point that could ease development in the future.
I’m currently working on a visualization for q2-demux. It is an interactive quality plot that will subsample the dataset and create essentially a “line graph” of box plots for the quality scores at each base. The boxplots are set up and good to go, now it’s just the zooming and interactivity that need to get worked on, and scaling the sampling to work on real datasets (currently developing on Moving Pictures tutorial data).
I’m working on a few new methods and visualizers that will go into 2017.3, and also working on increasing test coverage across the plugin repositories. This has included working with @Nicholas_Bokulich on vsearch and BLAST+ consensus taxonomy assigners (basically reviewing his PR), and making sure all of the plugin repos are reporting test coverage to coveralls. Next week I’ll be adding Mantel tests to q2-diversity (included the “sample metadata to distance matrix” functionality that we have in QIIME 1), and writing new unit tests for a lot of the plugins.
I’m working on vsearch and BLAST+ methods for q2-feature-classifier, as well as parameters sweeps on these and other methods to test and compare taxonomy classification methods in q2-feature-classifier.
Bit late on this, I’m working on adding in tabular stats visualizations into q2-deblur. It’s in PR right now. I’m also working on threading and refined SIMD support for q2-state-unifrac which, as of last week, was able to process the EMP 90nt set (~25k samples) in under 34h with 4 threads and peak resident at 12GB. On the memory, it balloons at completion due to the cast to a dense distance matrix – it may be nice to add into q2-types support for a condensed on disk binary format for a DistanceMatrix.