Files
Jungfraujoch/image_analysis/IndexAndRefine.h
T
leonarski_fandClaude Opus 5 fb07263025 predict: integrate every centring node, so a fixed space group costs no data
With -S the prediction rejected the fixed group's centring absences, so those reflections were never
integrated. Two things followed, and only the second was known.

The P1 cross-check was withheld on such a run, because a P1 merge missing whole centring classes is
misleading rather than merely small - 50% of the nodes on an I lattice, 75% on F, 67% on R. That was
the documented reason and it was right.

The unknown one is that it cost intensity accuracy. Every predicted reflection marks its signal region
so a neighbour's background ring can exclude it (BraggIntegrationEngineCPU, the reflection mask); an
unpredicted node is an unclaimed patch of detector, and the neighbouring reflections sweep those pixels
into their background and over-subtract - worst at high angle, where the background dominates. On a
fixed F-centred group that is three quarters of the nodes: measured against the de-novo run of the same
data, <I/sigma> 16.07 against 17.21, CC1/2 0.9862 against 0.9895, ISa 10.93 against 11.92.

Predicting them costs nothing downstream, because both merges already decide absence against the group
they are merging in: the run's own merge drops them again, and the P1 cross-check keeps them because P1
has none. One integration, two correct merges. The -S output becomes byte-identical to the de-novo run
on the five crystals measured, which is the point - pinning a group should not change the answer - and
such a run can never be slower than de novo, since it predicts the same reflections and additionally
skips the space-group search.

The de-novo path is untouched by construction: it already predicted in P.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EFEJG6WBQv8th4UJFNe53N
2026-09-02 09:18:13 +02:00

155 lines
8.4 KiB
C++

// SPDX-FileCopyrightText: 2025 Filip Leonarski, Paul Scherrer Institute <filip.leonarski@psi.ch>
// SPDX-License-Identifier: GPL-3.0-only
#pragma once
#include <atomic>
#include <vector>
#include <mutex>
#include <functional>
#include "../common/DiffractionSpot.h"
#include "../common/DiffractionExperiment.h"
#include "../common/AzimuthalIntegrationMapping.h"
#include "../common/AzimuthalIntegrationProfile.h"
#include "../common/Reflection.h"
#include "bragg_prediction/BraggPrediction.h"
#include "indexing/IndexerThreadPool.h"
#include "lattice_search/LatticeSearch.h"
#include "rotation_indexer/RotationIndexer.h"
#include "rotation_indexer/RotationIndexerCounter.h"
#include "scale_merge/ReindexAmbiguity.h"
#include "scale_merge/ScaleOnTheFly.h"
#include "scale_merge/ScalingResult.h"
#include "IntegrationOutcome.h"
// Integrates the predicted reflections off whatever image the caller holds: the preprocessed GPU/CPU
// buffer on the WithoutFPGA path (GPU when available), or the assembled detector image read straight,
// on the CPU, on the forced-CPU FPGA path. Keeps IndexAndRefine independent of the image representation.
using BraggIntegrateFn = std::function<std::vector<Reflection>(
const std::vector<Reflection> &predicted, size_t npredicted, int64_t image_number)>;
class IndexAndRefine {
// When false, the current image's result is still returned via the outgoing message, but the
// whole-run integration_outcome vector is not retained (viewer live/interactive use, which never
// scales the accumulated run). rugnux/receiver keep it true so ScaleAllImages/merge have the data.
const bool retain_outcomes_;
const bool real_time; // see the constructor
const DiffractionExperiment& experiment;
const DiffractionGeometry geom_;
std::optional<CrystalLattice> indexed_lattice;
std::optional<GoniometerAxis> axis_;
IndexerThreadPool *indexer_;
std::unique_ptr<RotationIndexer> rotation_indexer;
RotationIndexerCounter rotation_indexer_counter;
struct IndexingOutcome {
std::optional<CrystalLattice> lattice_candidate;
std::vector<CrystalLattice> extra_lattice_candidates;
std::vector<Coord> extra_lattice_rotations;
DiffractionExperiment experiment;
LatticeMessage symmetry{
.centering = 'P',
.niggli_class = 0,
.crystal_system = gemmi::CrystalSystem::Triclinic
};
bool beam_center_updated = false;
explicit IndexingOutcome(const DiffractionExperiment& experiment_ref)
: experiment(experiment_ref) {}
};
mutable std::mutex reflections_mutex;
std::vector<IntegrationOutcome> integration_outcome;
std::vector<float> mosaicity;
// Optional per-frame mosaicity used for Bragg prediction, indexed by image number. When set (the
// second pass of the rotation two-pass), it overrides the per-image spot-shape estimate so prediction
// uses the frame-order-SMOOTHED mosaicity that RotationScaleMerge already fitted in the first pass,
// rather than re-deriving it from scratch.
std::vector<float> prediction_mosaicity_override_;
// Predict every node of the lattice, ignoring the centring absences of a fixed space group. Set
// for the rotation two-pass GEOMETRY pre-pass, whose job is to measure the detector geometry from
// spot positions and whose intensities are thrown away: rejecting the absences there costs it half
// its events and buys nothing. Measured with an I-centred group fixed - the pre-pass fitted a
// different error model (ISa 7.8 -> 3.6), post-refined the distance 119 um away, and the second
// pass re-indexed 49 of 60 frames instead of 60.
bool predict_all_centring_nodes_ = false;
// The lattice centring the last prediction actually ran in: 'P' unless a user-fixed space group
// made it reject that group's centring absences. What a caller needs to know whether the
// integration covers every node of the lattice - the P1 cross-check merge does.
std::atomic<char> prediction_centring_ = 'P';
// Whether the outgoing message carries its own copy of the integrated reflections. The per-image
// file writer and the online stream are the only readers of it - the whole-run scaling/merge reads
// the retained outcome instead - and it is a copy of every reflection of every image, so a caller
// that writes no per-image file switches it off.
bool keep_reflections_in_message_ = true;
std::vector<float> scale_cc;
std::vector<std::optional<UnitCell> > unit_cells;
IndexingOutcome DetermineLatticeAndSymmetryRotation(DataMessage &msg);
IndexingOutcome DetermineLatticeAndSymmetry(DataMessage &msg);
// Shared indexing path: determine the lattice/symmetry, refine geometry, and run AnalyzeIndexing.
// Returns the outcome (ready for integration) when the frame indexes, nullopt otherwise. Both the
// real per-image ProcessImage and the first-pass scheme validation go through this, so they cannot
// diverge.
std::optional<IndexingOutcome> DetermineRefineAnalyze(DataMessage &msg,
const SpotFindingSettings &spot_finding_settings);
void RefineGeometryIfNeeded(DataMessage &msg, IndexingOutcome &outcome);
void QuickPredictAndIntegrate(DataMessage &msg,
const SpotFindingSettings &spot_finding_settings,
BraggPrediction &prediction,
const BraggIntegrateFn &integrate,
const IndexingOutcome &outcome);
std::unique_ptr<ReindexAmbiguityResolver> reindex_resolver;
void ScaleImage(DataMessage &msg, IntegrationOutcome& outcome);
std::optional<float> RotationAngle(int64_t image) const; // mid-exposure angle for the indexer
public:
// real_time: bound the geometry refinements - the per-image one here and the candidate-cell ones
// in the rotation indexer - by WALL CLOCK, as online acquisition must, it having a real budget.
// Offline (rugnux, the viewer) passes false and they are bounded by iteration count instead, so the
// same file reprocesses to the same answer regardless of what else the machine was doing.
IndexAndRefine(const DiffractionExperiment &x, IndexerThreadPool *indexer, bool retain_outcomes = true,
bool real_time = false);
void AddImageToRotationIndexer(DataMessage &msg);
void ForceRotationIndexerLattice(const CrystalLattice& lattice);
void ForceRotationIndexerResult(const RotationIndexerResult& result);
// Supply a per-frame (by image number) mosaicity for prediction, overriding the per-image estimate.
void SetPredictionMosaicityOverride(std::vector<float> mosaicity_per_frame) {
prediction_mosaicity_override_ = std::move(mosaicity_per_frame);
}
// Predict the centring-absent reflections too, even with a fixed space group - see the member.
void PredictAllCentringNodes(bool on) { predict_all_centring_nodes_ = on; }
// The centring prediction ran in - see the member.
char GetPredictionCentring() const { return prediction_centring_; }
// Whether the outgoing message keeps its own copy of the reflections - see the member.
void KeepReflectionsInMessage(bool on) { keep_reflections_in_message_ = on; }
// Returns whether the frame indexed (a lattice was found and refined). Integration, when it runs,
// is a further step gated on quick_integration.
void ProcessImage(DataMessage &msg, const SpotFindingSettings &settings,
BraggPrediction &prediction, const BraggIntegrateFn &integrate);
// Index a single frame (no integration) with the current forced rotation lattice; used to score
// first-pass sampling schemes on the real per-image path. Returns whether the frame indexed.
bool IndexFrameOnly(DataMessage &msg, const SpotFindingSettings &settings);
IndexAndRefine& ReferenceIntensities(std::vector<MergedReflection> &reference);
ScalingResult ScaleAllImages(const std::vector<MergedReflection> &reference, size_t nthreads = 0);
std::optional<RotationIndexerResult> FinalizeRotationIndexing();
std::optional<UnitCell> GetConsensusUnitCell() const;
// Not thread safe, need to be run after processing is all done
const std::vector<float> &GetImageCC() const;
const std::vector<std::optional<UnitCell> > &GetUnitCells() const;
std::vector<IntegrationOutcome> &GetIntegrationOutcome();
const std::vector<IntegrationOutcome> &GetIntegrationOutcome() const;
};