Add uncompressed in-memory size to input image summary

prod(shape) * dtype.itemsize, formatted with binary units (B/KiB/MiB/GiB).
The on-disk .nii.gz size hides this; it's the number that matters when the
pipeline loads the array.

Author: nx10

src/rbc/core/nifti.py

@@ -546,14 +546,24 @@ def _space_label(code: int) -> str:
         return str(code)
 
 
+def _human_bytes(n: int) -> str:
+    """Format a byte count with a binary (B/KiB/MiB/GiB) suffix."""
+    size = float(n)
+    for unit in ("B", "KiB", "MiB"):
+        if size < 1024:
+            return f"{size:.0f} {unit}" if unit == "B" else f"{size:.1f} {unit}"
+        size /= 1024
+    return f"{size:.1f} GiB"
+
+
 def log_image_summary(in_file: str | Path, *, label: str = "Raw input") -> None:
     """Log array shape, dtype, and geometry of a raw NIfTI input.
 
     Reads only the NIfTI header (no voxel data is loaded), then emits an
     INFO-level summary so the run log records exactly what entered the
-    pipeline: array shape, on-disk dtype, voxel size, axis orientation,
-    sform/qform coordinate spaces, and (for 4D images) volume count, slice
-    count, and TR.
+    pipeline: array shape, on-disk dtype, uncompressed in-memory size,
+    voxel size, axis orientation, sform/qform coordinate spaces, and (for
+    4D images) volume count, slice count, and TR.
 
     Args:
         in_file: Path to a ``.nii``/``.nii.gz`` file.
@@ -564,20 +574,23 @@ def log_image_summary(in_file: str | Path, *, label: str = "Raw input") -> None:
     img = nib.nifti1.load(path)
     hdr = img.header
     shape = img.shape
+    dtype = hdr.get_data_dtype()
     zooms = hdr.get_zooms()
     spatial_unit = hdr.get_xyzt_units()[0]
 
+    n_bytes = int(np.prod(shape, dtype=np.int64)) * dtype.itemsize
     voxel = " x ".join(f"{z:.3g}" for z in zooms[:3])
     if spatial_unit != "unknown":
         voxel = f"{voxel} {spatial_unit}"
     orientation = "".join(nib.aff2axcodes(img.affine))
 
     _logger.info("%s: %s", label, path)
     _logger.info(
-        "%s: shape=%s, dtype=%s, voxel size=%s",
+        "%s: shape=%s, dtype=%s, uncompressed size=%s, voxel size=%s",
         label,
         shape,
-        hdr.get_data_dtype(),
+        dtype,
+        _human_bytes(n_bytes),
         voxel,
     )
     _logger.info(

tests/unit/test_nifti.py

@@ -634,13 +634,14 @@ class TestLogImageSummary:
     """Tests for log_image_summary()."""
 
     def test_3d_summary(self, nifti_3d: Path, caplog: pytest.LogCaptureFixture) -> None:
-        """3D input logs shape, dtype, voxel size, orientation, and spaces."""
+        """3D input logs shape, dtype, size, voxel size, orientation, spaces."""
         caplog.set_level(logging.INFO, logger="rbc.core.nifti")
         log_image_summary(nifti_3d, label="Anatomical T1w")
         text = "\n".join(caplog.messages)
         assert "Anatomical T1w" in text
         assert "shape=(5, 6, 7)" in text
         assert "dtype=float64" in text
+        assert "uncompressed size=1.6 KiB" in text  # 5*6*7 * 8 = 1680 bytes
         assert "voxel size=1 x 1 x 1 mm" in text
         assert "orientation=RAS" in text
         assert "sform=MNI" in text
@@ -663,18 +664,30 @@ def test_4d_summary_includes_volumes_and_tr(
         log_image_summary(nifti_4d, label="Functional BOLD")
         text = "\n".join(caplog.messages)
         assert "shape=(5, 6, 7, 10)" in text
+        assert "uncompressed size=16.4 KiB" in text  # 5*6*7*10 * 8 bytes
         assert "volumes=10" in text
         assert "slices=7" in text
         assert "header TR=2 s" in text
 
-    def test_dtype_reflects_on_disk_type(
+    def test_dtype_and_size_reflect_on_disk_type(
         self, tmp_path: Path, caplog: pytest.LogCaptureFixture
     ) -> None:
-        """Logged dtype is the on-disk dtype, not float from get_fdata()."""
+        """Logged dtype/size use the on-disk dtype, not float64 get_fdata()."""
         path = _make_nifti(tmp_path, "int16.nii.gz", (4, 5, 6), dtype=np.int16)
         caplog.set_level(logging.INFO, logger="rbc.core.nifti")
         log_image_summary(path)
-        assert any("dtype=int16" in m for m in caplog.messages)
+        text = "\n".join(caplog.messages)
+        assert "dtype=int16" in text
+        assert "uncompressed size=240 B" in text  # 4*5*6 * 2 bytes
+
+    def test_size_uses_binary_units(
+        self, tmp_path: Path, caplog: pytest.LogCaptureFixture
+    ) -> None:
+        """Uncompressed size scales to binary units."""
+        path = _make_nifti(tmp_path, "big.nii.gz", (64, 64, 64), dtype=np.int16)
+        caplog.set_level(logging.INFO, logger="rbc.core.nifti")
+        log_image_summary(path)
+        assert any("uncompressed size=512.0 KiB" in m for m in caplog.messages)
 
     def test_emitted_at_info_level(
         self, nifti_3d: Path, caplog: pytest.LogCaptureFixture