evidence: restore HDF5-to-tile smoke runner + model.py axis-order fix

dsa110_continuum/calibration/model.py

@@ -68,11 +68,10 @@
 import time
 from pathlib import Path
 
-
 import astropy.units as u  # noqa: E402
-from dsa110_continuum.adapters import casa_tables as tb  # noqa: E402
 import numpy as np  # noqa: E402
 from astropy.coordinates import SkyCoord  # noqa: E402
+from dsa110_continuum.adapters import casa_tables as tb  # noqa: E402
 
 # Set up logger
 logger = logging.getLogger(__name__)
@@ -444,20 +443,53 @@ def _calculate_manual_model_data(
         print(f"  Processing {nselected:,} selected rows...")
         sys.stdout.flush()
 
-        # Read DATA shape to create MODEL_DATA with matching shape
+        # Read DATA shape to create MODEL_DATA with matching shape.
+        # CASA tables can store DATA as either (nchan, npol) or (npol, nchan)
+        # depending on writer convention; identify the channel axis by matching
+        # against the SPW channel count rather than assuming an order.
         print("    Reading DATA shape...")
         sys.stdout.flush()
         data_sample = main_tb.getcell("DATA", 0)
-        data_shape = data_sample.shape  # In CASA: (nchan, npol)
-        nchan, npol = data_shape[0], data_shape[1]
-        logger.debug(f"Data shape: {nchan} channels, {npol} polarizations")
-        print(f"    DATA shape: {nchan} channels × {npol} pols")
+        data_shape = data_sample.shape
+        if len(data_shape) != 2:
+            raise ValueError(
+                f"Unsupported DATA cell shape for MODEL_DATA calculation: {data_shape}"
+            )
+
+        expected_nchan = int(chan_freq.shape[1])
+        if data_shape[0] == expected_nchan:
+            chan_axis = 0
+            corr_axis = 1
+        elif data_shape[1] == expected_nchan:
+            corr_axis = 0
+            chan_axis = 1
+        else:
+            raise ValueError(
+                "Cannot identify channel axis for MODEL_DATA calculation: "
+                f"DATA cell shape={data_shape}, expected channel count={expected_nchan}"
+            )
+
+        nchan = data_shape[chan_axis]
+        npol = data_shape[corr_axis]
+        logger.debug(
+            "Data shape: %s, channel axis=%d (%d channels), correlation axis=%d (%d correlations)",
+            data_shape,
+            chan_axis,
+            nchan,
+            corr_axis,
+            npol,
+        )
+        print(
+            f"    DATA shape: {data_shape} (channel axis={chan_axis}, correlation axis={corr_axis})"
+        )
         sys.stdout.flush()
 
         # Initialize MODEL_DATA array with correct shape (nrows, nchan, npol)
         print(f"    Allocating MODEL_DATA array ({nrows * nchan * npol * 8 / 1e9:.2f} GB)...")
         sys.stdout.flush()
-        model_data = np.zeros((nrows, nchan, npol), dtype=np.complex64)
+        # Allocate MODEL_DATA with the per-row shape that matches DATA exactly,
+        # so axis order matches whatever the writer used.
+        model_data = np.zeros((nrows, *data_shape), dtype=np.complex64)
         logger.debug(f"Allocated MODEL_DATA array: {model_data.nbytes / 1e9:.2f} GB")
         print("    MODEL_DATA allocated")
         sys.stdout.flush()
@@ -565,8 +597,12 @@ def _calculate_manual_model_data(
             # Create complex model: amplitude * exp(i*phase)
             model_complex = amplitude * (np.cos(phase) + 1j * np.sin(phase))
 
-            # Broadcast to all polarizations and store in output array
-            model_data[chunk_indices, :, :] = model_complex[:, :, np.newaxis]
+            # Broadcast the channel-dependent point-source model to every
+            # correlation while preserving the MS DATA column's axis order.
+            if chan_axis == 0:
+                model_data[chunk_indices, :, :] = model_complex[:, :, np.newaxis]
+            else:
+                model_data[chunk_indices, :, :] = model_complex[:, np.newaxis, :]
 
             # Progress output
             if (chunk_idx + 1) % 3 == 0 or chunk_idx == n_chunks - 1:

dsa110_continuum/evidence/__init__.py

@@ -0,0 +1,2 @@
+"""Evidence workflows for pipeline validation artifacts."""
+