20220120 developer call notes

[tomwhite]

20220120

Pre-notes

• Notes from previous meeting

PRs

• sgkit PR#777 (Tim/Jerome) Add cohort_allele_frequencies method

Issues

• sgkit#787 Update DocSearch
• sgkit#794 Write zarr to HDFS
• sgkit#795 Document best practice of windowing after filtering variants
• sgkit#802 Remove support for Python 3.7
  • sgkit#801 Test on Python 3.10
• sgkit#803 Run on Pyodide
• sgkit#809 Both Upstream and Validation is failing

Discussions

Other topics

Notes

Attendees

1. Jeff Hammerbacher
2. Jerome Kelleher
3. Eric Czech
4. Tom White
5. Rafal Wojdyla

Discussion

• Blog post
  • Nvidia - writing a blog post on GWAS and using GPUs.
  • JH: has already reached out to numfocus
  • JK: happy to help too
  • Link the two blog posts
• sgkit#787 Update DocSearch
  • JH: take a bit more time to look at it
  • JH: OK for API key to be public
• sgkit#794 Write zarr to HDFS
  • JK: may be better to use a cloud object store
• sgkit#802 Remove support for Python 3.7
  • TW:
  • JK: can give you more work. Take path of least resistance. Drop it when things break.
  • JH: funny position - two competing decisions - the code we’ve written runs fine, but latest deps (e.g. Xarray) does work. Inclined to bring people along - more enterprise. 3.7 EOL is next year.
  • JK: debian has 3.8 now
  • TW: will add py 3.10 too
• sgkit#809 Both Upstream and Validation is failing
  • TW: numba has a strict check on numpy version number now. Put numpy < 1.22 in deps?
  • JH: how do we know when to remove it?
  • RW: remove pip resolver? https://github.com/pystatgen/sgkit/blob/main/setup.cfg#L45-L57