Move flex_table to dxtbx #2113
Move flex_table to dxtbx #2113
Conversation
|
During the DIALS catch up at UCLA we noticed there is some duplicated code in |
|
Discussed in DIALS core. Probably when this merge is fixed, we'd want to have the header files remain, but just import the dxtbx versions and print a deprecation warning when the old headers are included. This way, downstream dependencies will warning but not break. |
… c++ level to reflection_table_suite.h.
toastisme
force-pushed
the
move_flex_table_to_dxtbx
branch
from
bcf609d
to
19bf305
Compare
|
Deprecation warnings have been added. I've kept There's currently a build failure for Windows, but this looks to be due to the Windows Azure build using an older version of dxtbx. |
src/dials/algorithms/integration/boost_python/integration_integrator_ext.cc
Outdated
Show resolved
Hide resolved
src/dials/algorithms/integration/boost_python/integration_integrator_ext.cc
Outdated
Show resolved
Hide resolved
…tbx. Removed some unused local variables.
…de of flex_table_suite to integration_integrator_ext.cc.
…de of flex_table_suite to integration_integrator_ext.cc.
…#1903) Adds damage_series.dose_group_size= option to dials.damage_analysis, to generate some merging statistics plots for subsets of the data. The corresponding datafiles can also be output by turning on the options output.damage_series= and output.accumulation_series=. This generates additional plots after the pychef analysis which are output in the html (cc_half, r_pim, i_over_sigma and completeness). Output files are optionally generated for the damage group series: damage_series_0_100.{expt,refl} damage_series_100_200.{expt,refl} etc. and also optionally for the damage accumulation series: damage_series_0_100.{expt,refl} damage_series_0_200.{expt,refl} etc.
And hide conda_base/ -based errors for the entire CI build on Azure.
towncrier --name=DIALS --version='3.14.0'
Python 3.10 remains the default, for now. Also, update CI to either disable or suppress PYTHONDEVMODE warnings from this matplotlib bug: matplotlib/matplotlib#25244
…ls#2397) The existing code meant that the nth ring was at the d_min of the image meaning that only n-1 rings were every actually visible on the image
towncrier --name "DIALS" --version 3.14.1
These are now separate packages, and conda-forge packages them separately.
Edit: I am getting confused between this PR and the same branch name on dxtbx. But it's this-issue-related. |
|
Okay, yes, the specific issue is that dxtbx now imports the scitbx/.../ref_pickle_double_buffered.h directly, because the header-inclusion-guard version wasn't moved across. Thus, even though dials only uses the inclusion-guard version, everything inside still gets declared twice. This needs a DXTBX change to fix. |
|
cctbx/dxtbx#636 will fix this, but I will also update the DIALS references |
This PR along with #521 moves
flex_tableto dxtbx, with the fewreflection_tablespecific functions (i.e those involvingexperiment_identifiers) moved toreflection_table_suite.h. The motivation for this is to useflex_tablefor other purposes in dxtbx (specifically a properties table inScan), as it's essentially a generic dictionary with some nice functionality thatreflection_tableextends.I'm not really happy with the
flex_table_suite.handreflection_table_suite.hsetup. At the c++ level you could accidentally use methods from the former and might not notice the lack of experiment identifiers updating. As nearly allflex_tablefunctions are not actually part of theflex_tableclass, clean inheritance isn't really possible..So I have just copied the convention.Tests will not pass without #521 being merged first. No additional tests added as all functionality should be the same, and should be covered by
tests/algorithms/array_family/test_identifiers_handling.pyandtests/algorithms/array_family/test_reflection_table.py