DIALS core meeting 2021-02-25

Previous Actions

• MG+ND: write a proposal for ‘overall architecture discussion’ and ‘new installer’
• MG: organise a typing intro lecture, likely to happen in Q2.
• cbflib conda-forge package
  • ND to prepare an agenda for the HDRMX meeting 16-18. March 2021
  • ASB to organise a side-meeting next to the HDRMX workshop
  • ASB to explore what should happen to cbflib maintenance in the future
• ND to set up a cbflib_adaptbx pull request for MAR detector fixes. Verify with labelit_regression.
• ASB to run tests on dxtbx#288 as per comment
• PR dials#1567
  • GW to run tests and merge if passing
• 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

Agenda

Construct Experiments directly rather than via DataBlocks #288

Can this be unblocked/merged?

• AP: Dan to try and generate a non-psana reproducer for the issue found by ASB

DC2: DIALS installer

ND and MG prepared a proposal (DC2) on how the installer should work in the future.

Outcome of discussion: This is accepted as useful and we will do that. WP1 and WP2 will happen first, then we will wait a bit before going to WP3/4.

Change HDF5 plugin library

We currently use the plugin collection hdf5-external-filter-plugins, which is built from https://github.com/nexusformat/HDF5-External-Filter-Plugins. There is no activity really in that repository. Although I did manage to build the plugin collection on Windows I never tested that, and it looks like until last week nobody else did either.

hdf5plugin is an alternative collection. It contains a proper superset of the plugins we already have, is more actively supported, has more users, and a Windows build.

When we do move over:

• there may be a need to manually uninstall the hdf5-e-f-p package on existing installations
• the hdf5plugin package needs an import hdf5plugin before the plugins become available, which brings us back to where we were once before.
• AP: MG to come up with a set of spells to use hdf5plugin
• AP: GW to run a performance comparison
• AP: ws133 does not support AVX2, need to test that the conda-forge package isn’t over-optimized
• DP would like to run tests before we change over

Renaming master → main

xia2 has now renamed its main branch.

dials#1606 is ready to go in, so is dxtbx#317.

AP: MG to make this happen

End DIALS/dxtbx for Python 2

Should we upgrade the DIALS/dxtbx UserWarning

if sys.version_info.major == 2:
    warnings.warn(
        "Python 2 is no longer supported. "
        "If you need Python 2.7 support please use the DIALS 2.2 release branch.",
        UserWarning,
    )

to a xia2 hard stop?

if sys.version_info.major == 2:
    sys.exit("Python 2 is no longer supported")

AP: MG to make this happen

dxtbx comparator pseudo-classes

dxtbx#302 found dxtbx comparator classes in dxtbx.datablock. These classes (SequenceDiff, BeamDiff, ScanDiff, GoniometerDiff, DetectorDiff) should really be functions. Given they are in datablock they could move to dxtbx.model.experiment_list, but possibly something like dxtbx.model.compare makes more sense?

(This is independent of dxtbx#288 above)

AP: GW to add comment to that PR, rediscuss when ASB is present.

Next meeting

Thursday, March 11th, 4pm UK time, 8am PDT.