Merge pull request #71 from scipp/multi-detector-tonxeventdata

Add returning data from all detectors as nxevent_data

src/tof/result.py

@@ -390,29 +390,45 @@ def __repr__(self) -> str:
     def __str__(self) -> str:
         return self.__repr__()
 
-    def to_nxevent_data(self, key: str) -> sc.DataArray:
+    def to_nxevent_data(self, key: Optional[str] = None) -> sc.DataArray:
         """
-        Convert a component reading to event data that resembles event data found in a
+        Convert a detector reading to event data that resembles event data found in a
         NeXus file.
 
         Parameters
         ----------
         key:
-            Name of the component.
+            Name of the detector. If ``None``, all detectors are included.
         """
         start = sc.datetime("2024-01-01T12:00:00.000000")
         period = sc.reciprocal(self.source.frequency)
-        raw_data = self[key].data.flatten(to='event')
-        # Select only the neutrons that make it to the detector
-        event_data = raw_data[~raw_data.masks['blocked_by_others']].copy()
+
+        keys = list(self._detectors.keys()) if key is None else [key]
+
+        event_data = []
+        for name in keys:
+            raw_data = self._detectors[name].data.flatten(to='event')
+            events = (
+                raw_data[~raw_data.masks['blocked_by_others']]
+                .copy()
+                .drop_masks('blocked_by_others')
+            )
+            events.coords['distance'] = sc.broadcast(
+                events.coords['distance'], sizes=events.sizes
+            ).copy()
+            event_data.append(events)
+
+        event_data = sc.concat(event_data, dim=event_data[0].dim)
         dt = period.to(unit=event_data.coords['toa'].unit)
         event_time_zero = (dt * (event_data.coords['toa'] // dt)).to(dtype=int) + start
         event_data.coords['event_time_zero'] = event_time_zero
         event_data.coords['event_time_offset'] = event_data.coords.pop(
             'toa'
         ) % period.to(unit=dt.unit)
-        return (
+        out = (
             event_data.drop_coords(['tof', 'speed', 'time', 'wavelength'])
-            .group('event_time_zero')
+            .group('distance', 'event_time_zero')
             .rename_dims(event_time_zero='pulse')
-        )
+        ).rename_dims(distance='detector_number')
+        out.coords['Ltotal'] = out.coords.pop('distance')
+        return out

tests/model_test.py

@@ -428,3 +428,7 @@ def test_to_nxevent_data():
         assert nxevent_data.bins.concat().value.coords[
             'event_time_offset'
         ].max() <= sc.reciprocal(source.frequency).to(unit='us')
+
+    # Test when we include all detectors at once
+    nxevent_data = res.to_nxevent_data()
+    assert nxevent_data.sizes == {'detector_number': 2, 'pulse': 2}

tests/source_test.py

@@ -104,24 +104,9 @@ def test_creation_from_distribution():
     assert sc.isclose(mid / right, sc.scalar(10.0), rtol=rtol)
 
     # Make sure distribution is monotonically increasing
-    locs = np.linspace(1.0, 4.0, 20)
-    step = 0.5 * (locs[1] - locs[0])
-    for i in range(len(locs) - 2):
-        a = da.hist(
-            wavelength=sc.array(
-                dims=['wavelength'],
-                values=[locs[i] - step, locs[i] + step],
-                unit='angstrom',
-            )
-        ).data.sum()
-        b = da.hist(
-            wavelength=sc.array(
-                dims=['wavelength'],
-                values=[locs[i + 1] - step, locs[i + 1] + step],
-                unit='angstrom',
-            )
-        ).data.sum()
-        assert b > a
+    h = da.hist(wavelength=10)
+    diff = h.data[1:] - h.data[:-1]
+    assert sc.all(diff > sc.scalar(0.0, unit='counts'))
 
 
 def test_non_integer_sampling():