Merge pull request #4 from data-exp-lab/dtype_handling

handle different dtypes

pyproject.toml

@@ -131,6 +131,7 @@ ignore = [
   "PLR09",    # Too many <...>
   "PLR2004",  # Magic value used in comparison
   "ISC001",   # Conflicts with formatter
+  "SIM108",   # Use ternary operator
 ]
 isort.required-imports = ["from __future__ import annotations"]
 # Uncomment if using a _compat.typing backport

src/pyramid_sampler/sampler.py

@@ -132,8 +132,10 @@ def _downsample_by_one_level(
         level = coarse_level
         fine_level = level - 1
         lev_shape = self._get_level_shape(level)
+
         field1 = zarr.open(self.zarr_store_path)[zarr_field]
-        field1.empty(level, shape=lev_shape, chunks=self.chunks)
+        dtype = field1[fine_level].dtype
+        field1.empty(level, shape=lev_shape, chunks=self.chunks, dtype=dtype)
 
         numchunks = field1[str(level)].nchunks
 
@@ -197,7 +199,8 @@ def _write_chunk_values(
     )
 
     coarse_zarr = zarr.open(zarr_file)[zarr_field][str(level)]
-    coarse_zarr[si[0] : ei[0], si[1] : ei[1] :, si[2] : ei[2]] = outvals
+    dtype = coarse_zarr.dtype
+    coarse_zarr[si[0] : ei[0], si[1] : ei[1] :, si[2] : ei[2]] = outvals.astype(dtype)
 
     return 1
 
@@ -208,15 +211,27 @@ def initialize_test_image(
     base_resolution: tuple[int, int, int],
     chunks: int | tuple[int, int, int] | None = None,
     overwrite_field: bool = True,
+    dtype: str | type | None = None,
 ) -> None:
+    if dtype is None:
+        dtype = np.float64
     field1 = zarr_store.create_group(zarr_field, overwrite=overwrite_field)
 
     if chunks is None:
         chunks = (64, 64, 64)
-    lev0 = da.random.random(base_resolution, chunks=chunks)
+    fac: int | float
+    if np.issubdtype(dtype, np.integer):
+        fac = 100
+    elif np.issubdtype(dtype, np.floating):
+        fac = 1.0
+    else:
+        msg = f"Unexpected dtype of {dtype}"
+        raise RuntimeError(msg)
+    lev0 = fac * da.random.random(base_resolution, chunks=chunks)
+    lev0 = lev0.astype(dtype)
     halfway = np.asarray(base_resolution) // 2
     lev0[0 : halfway[0], 0 : halfway[1], 0 : halfway[2]] = (
-        lev0[0 : halfway[0], 0 : halfway[1], 0 : halfway[2]] + 0.5
+        lev0[0 : halfway[0], 0 : halfway[1], 0 : halfway[2]] + 0.5 * fac
     )
-    field1.empty(0, shape=base_resolution, chunks=chunks)
+    field1.empty(0, shape=base_resolution, chunks=chunks, dtype=dtype)
     da.to_zarr(lev0, field1["0"])

tests/test_sampler.py

@@ -27,20 +27,24 @@ def test_initialize_test_image(tmp_path):
     assert zarr_store[fieldname][0].shape == res
 
 
-def test_downsampler(tmp_path):
+@pytest.mark.parametrize("dtype", ["float32", np.float64, "int", np.int32, np.int16])
+def test_downsampler(tmp_path, dtype):
     tmp_zrr = str(tmp_path / "myzarr.zarr")
     zarr_store = zarr.open(tmp_zrr)
     res = (32, 32, 32)
     chunks = (8, 8, 8)
     fieldname = "test_field"
-    initialize_test_image(zarr_store, fieldname, res, chunks, overwrite_field=False)
+    initialize_test_image(
+        zarr_store, fieldname, res, chunks, overwrite_field=False, dtype=dtype
+    )
 
     dsr = Downsampler(tmp_zrr, (2, 2, 2), res, chunks)
 
     dsr.downsample(10, fieldname)
     expected_max_lev = 2
     for lev in range(expected_max_lev + 1):
         assert lev in zarr_store[fieldname]
+        assert zarr_store[fieldname][lev].dtype == np.dtype(dtype)
 
     with pytest.raises(ValueError, match="max_level must exceed 0"):
         dsr.downsample(0, fieldname)