Use tensorestore for downsampling

Author: dstansby

src/stack_to_chunk/_array_helpers.py

@@ -4,7 +4,6 @@
 import numpy as np
 import skimage.measure
 import tensorstore as ts
-import zarr
 from joblib import delayed
 from loguru import logger
 
@@ -32,23 +31,14 @@ def _copy_slab(arr_path: Path, slab: da.Array, zstart: int, zend: int) -> None:
 
     logger.info(f"Writing z={zstart} -> {zend - 1}")
     # Write out data
-    arr_zarr = ts.open(
-        {
-            "driver": "zarr3",
-            "kvstore": {
-                "driver": "file",
-                "path": str(arr_path),
-            },
-            "open": True,
-        }
-    ).result()
+    arr_zarr = _open_with_tensorstore(arr_path)
     arr_zarr[:, :, zstart:zend].write(data).result()
     logger.info(f"Finished copying z={zstart} -> {zend - 1}")
 
 
 @delayed  # type: ignore[misc]
 def _downsample_block(
-    arr_in: zarr.Array, arr_out: zarr.Array, block_idx: tuple[int, int, int]
+    arr_in_path: Path, arr_out_path: Path, block_idx: tuple[int, int, int]
 ) -> None:
     """
     Copy a single block from one array to the next, downsampling by a factor of two.
@@ -59,15 +49,17 @@ def _downsample_block(
 
     Parameters
     ----------
-    arr_in :
-        Input array.
-    arr_out :
-        Output array. Must have the same chunk shape as `arr_in`.
+    arr_in_path :
+        Path to input array.
+    arr_out_path :
+        Path to output array. Must have the same chunk shape as `arr_in`.
     block_idx :
         Index of block to copy. Must be a multiple of the shard shape in `arr_out`.
 
     """
-    shard_shape: tuple[int, int, int] = arr_out.shards
+    arr_in = _open_with_tensorstore(arr_in_path)
+    arr_out = _open_with_tensorstore(arr_out_path)
+    shard_shape: tuple[int, int, int] = arr_out.chunk_layout.write_chunk.shape
     np.testing.assert_equal(
         np.array(block_idx) % np.array(shard_shape),
         np.array([0, 0, 0]),
@@ -85,17 +77,32 @@ def _downsample_block(
             block_idx[2] * 2, min((block_idx[2] + shard_shape[2]) * 2, arr_in.shape[2])
         ),
     )
-    data = arr_in[in_slice]
+    data = arr_in[in_slice].read().result()
 
     # Pad to an even number
     pads = np.array(data.shape) % 2
     pad_width = [(0, p) for p in pads]
     data = np.pad(data, pad_width, mode="edge")
-    data = skimage.measure.block_reduce(data, block_size=2, func=np.mean)
+    data = skimage.measure.block_reduce(data, block_size=2, func=np.mean).astype(
+        data.dtype
+    )
 
     out_slice = (
         slice(block_idx[0], min((block_idx[0] + shard_shape[0]), arr_out.shape[0])),
         slice(block_idx[1], min((block_idx[1] + shard_shape[1]), arr_out.shape[1])),
         slice(block_idx[2], min((block_idx[2] + shard_shape[2]), arr_out.shape[2])),
     )
-    arr_out[out_slice] = data
+    arr_out[out_slice].write(data).result()
+
+
+def _open_with_tensorstore(arr_path: Path) -> ts.TensorStore:
+    return ts.open(
+        {
+            "driver": "zarr3",
+            "kvstore": {
+                "driver": "file",
+                "path": str(arr_path),
+            },
+            "open": True,
+        }
+    ).result()

src/stack_to_chunk/main.py

@@ -387,14 +387,17 @@ def add_downsample_level(self, level: int, *, n_processes: int = 1) -> None:
         assert sink_arr.shards is not None
 
         # Get slice of every shard in the sink array
-        block_indices = [
+        block_indices: list[tuple[int, int, int]] = [
             (x, y, z)
             for x in range(0, sink_arr.shape[0], sink_arr.shards[0])
             for y in range(0, sink_arr.shape[1], sink_arr.shards[1])
             for z in range(0, sink_arr.shape[2], sink_arr.shards[2])
         ]
 
-        all_args = [(source_arr, sink_arr, idxs) for idxs in block_indices]
+        all_args: list[tuple[Path, Path, tuple[int, int, int]]] = [
+            (self._path / str(level_minus_one), self._path / level_str, idxs)
+            for idxs in block_indices
+        ]
 
         logger.info(f"Starting downsampling from level {level_minus_one} > {level}...")
         blosc_use_threads = blosc.use_threads