DIALS core meeting 2021-09-23

Previous Actions

• “Contiguous Nexus” cctbx/dxtbx#356
  • Non-draft state is pending checks against issues @dwpaley raised, and checking for non-contiguous cases [prev]
• dxtbx src/ layout [prev]
  • Nick: Ask cctbxbb if anyone has knowledge or experience of re-export of dispatchers on windows
  • This is a higher priority now - David feeling pain of missing
  • Also, seeing related errors “in the wild” e.g. cctbx/dxtbx#434
• Removal of DataBlock [prev, prev]
  • Nick: More visible deprecationwarning
  • Impact assessment to see what/who is affected
    • Other repositories - stuff like, iota, prime, ls49, labelit, labelit-regression, xfel-regression, lunus, uc_metrics gitlab/cctbx*, phenix and phenix_regression (phenix.image_viewer)
• DXTBX/pycbf [prev]
  • Outstanding: dxtbx#368 for optional cbflib/_adaptbx .
  • ND: conda-forge pycbf: Make new release to use dials-data directly for tests
  • Last status: one labelit/labelit_regression test still doesn’t pass
  • All except one labelit tests are now working. On NKS todo list to fix.
  • Progress on CBFlib 0.9.7?
• Nonconventional P1 indexing [prev]
  • RG to look and see how hard it is to return P1 in user-specified order

Agenda

Nonconventional P1 indexing

• Problem with P1 indexing not being returned in user-specified setting/order
  • Not doing so has XFEL gui repercussions
• Issue at dials/dials#1878
• Richard has written dials/dials#1880
  • Outstanding test fail on mac/windows - possible numerical instability?

Deterministic Experiment Identifiers

• (Previous dicussion)
• Previously: It would be useful to describe the problem that is trying to be solved
• LBL had technical discussions:
  • cctbx.xfel.merge - modular - can specify which list of workers you want to do in a row. Use case: 380k images - each with 20k reflections - millions into memory. First thing it does is throw away image sets - uses too much memory - even when distributed
  • Discussion about ImageSequence and why we moved that
  • Sounds positive that this might be a good way to solve
  • Action: Aaron Change dials.stills_process to default to composite mode and not write experiment files
  • Action: Aaron/GW: Organise another meeting to go over “load before heat death” PR and reasons why, solutions etc
• Discussed whether general metadata field would still be useful
  • After some spirited discussion, the conclusion was that this would be ripe for abuse - if there is a convincing case then we are happy to reconsider, but shelving for now

Generating Collection Strategy

• Art: Hit a wall with labelit, wanted to use DIALS
• Have two images, each taken from 90° apart, and wanted to get an indexing solution from this
• Want to use this to help users develop a collection strategy
• Glued various DIALS parts together, integration had issues - reference reflections did not match indexed reflections
• AB: Strategy is a hole in our infrastructure
  • Discussion on DIALS East work on xia2.strategy and how we were never quite satisfied with it
  • Discussed dials.export_best and our reasoning about strategy generation from still vs small wedge data
  • [Nick] Probably fair to conclude - Strategy is a hole in our publicliy advertised infrastructure

Next meeting

Thursday, October 7th, 4pm UK (BST), 8am PDT.