DIALS core meeting 2021-04-22

Previous Actions

• MG: organise a typing intro lecture, likely to happen in Q2.
• ND to set up a cbflib_adaptbx pull request for MAR detector fixes. Verify with labelit_regression.
• dx2 proposal
  • MG to put in a PR to create a proposal in the proposal-space
  • MG+ND: Come up with a proposal to move away from all code being in header files and consolidate into a single library
• Move to hdf5plugin
  • MG to come up with a set of spells to use hdf5plugin
  • GW to run a performance comparison
    • Provided an initial version of performance comparison code: https://gist.github.com/ndevenish/bf03b390adddb33089a8c6a70e8f4936 - we can run this with different plugins to assess whether there is any difference in run-time just paging through the data.
  • Need to test that the conda-forge package isn’t over-optimized by running on ws133 – MG: conda-forge package has AVX2 disabled.

Agenda

Discuss changes merged into cctbx modify_cosym

• Hooking into cosym to do functionality, which was inadvertantly broken by dials/dials#1647
  • Several pull requests in flight to fix this
  • Main consequence is that reindexed datasets are not coloured any more
• (Discussion about differences between Brehm/Diederichs and what Cosym is doing)
• Limiting B/D to indexing ambiguity avoids problems with trying to match full sets of symmetry operators together

Adding XFEL-regression to dials Azure tests

• Draft branch to add to Azure here
• linux_uncached_xfel pipeline runs and passes
• Wants to make a PR to start having it better tested on dials repository
• AB: Observing timeout problems with some pipelines - would be good to check on main PR
• Discussions on what happens when downstream xfel_regression tests break - concerns that DIALS team would have no way to make these fixes go away
• MG: Perhaps the build time requirements would benefit from building on top of the conda-forge cctbx package instead of rebuilding every time
• Discussed running on PRs only instead of every push
• Conclusion: Aaron to make pull request for wider study and discussion

cbflib conda-forge package

• ASB to organise a side-meeting next to the HDRMX workshop
• What happes with cbflib maintenance in the future
• Discussed outcome from HDRMX meeting and next steps
  • Changing so that tests pass on Debian downstream - this is an essential use case and new releases cannot break this
  • HB estimates graduate student project with approximately 3 months of effort required
• Try current release - regenerate pycbf but leave old sources; and see if it works Action: Nick to do testing this
  • we could just release 0.9.6 on conda-forge with some light patching over the top if this works

Deferred to next meeting

cctbx/dxtbx#288 “Construct Experiments directly” post-merge outstanding issues

• cctbx/dxtbx#336 - DataBlockFactory broken for single images
• XFAIL test for this now put in: cctbx/dxtbx#337

Next meeting

Thursday, May 6th, 4pm UK (BST), 8am PDT.