Since a rotation run decides with Friedel's law on, -A changed nothing at all -
not even the table, which stayed the Friedel-merged one. That made the report
unreadable against an anomalous CORRECT.LP: XDS with FRIEDEL'S_LAW=FALSE counts
its uniques, multiplicity, completeness, R_meas and CC1/2 per hand, and a
Friedel-merged table has roughly twice the multiplicity of one.
The merge now also reports itself counted with Friedel's law OFF - an acentric
reflection as two reflections, one per Bijvoet hand, a centric as one - as
ANOM_* and ANOM_REFRES_* keys and a shell table beside the run's own. It costs
no second merge: the per-hand accumulator that SigAno, CCanom and the I(+)/I(-)
export are already formed on carries every number in it, and the completeness
denominator is the same enumeration counted two per acentric. The intensities,
the scaling, the error model and every decision behind both tables are the
Friedel-merged ones, so this is a second way of counting one merge.
-A is therefore presentation only: it selects FRIEDELS_LAW, which says which of
the two tables was asked for. Both are printed either way. On a 6 keV set the
per-hand table reads 15.54 multiplicity and 95.1% completeness over the
reference range against the Friedel-merged 28.65 and 97.5%, where the XDS
reference is 12.42 and 92.4%.
Report-only: the written reflections, the MTZ and the mmCIF are byte-identical
to the run before it, and an -A run is byte-identical to a bare one but for the
FRIEDELS_LAW line.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013nW6FNRP1bBJJ8pfHiByAT
leonarski_fandClaude Opus 5
14b7fce816