Merge branch 'main' into n5

Author: yuriyzubov

src/zarrify/formats/zarr2.py

@@ -87,12 +87,14 @@ def write_to_zarr(
         )
 
         array_paths = self._find_arrays()
+        logger.info(f"Found {len(array_paths)} zarr v2 arrays: {array_paths}")
 
         dst_store = zarr.storage.LocalStore(dest)
         dst_root = zarr.open_group(store=dst_store, mode='a')
         self._copy_group_attrs(dst_root)
 
         for rel_path in array_paths:
+            logger.info(f"Processing array: {rel_path}")
             zarray_path = os.path.join(self.src_path, rel_path, '.zarray')
             with open(zarray_path) as f:
                 zarray_meta = json.load(f)
@@ -126,8 +128,10 @@ def write_to_zarr(
             dest_arr = open_ts(dst_spec)
             dest_chunks = dest_arr.chunk_layout.write_chunk.shape
 
+            logger.info(f"{rel_path}: shape={shape}, dtype={dtype}, chunk={arr_chunk_shape}, shard={arr_shard_shape}")
             out_slices = slices_from_chunks(normalize_chunks(dest_chunks, shape=shape))
             out_slices_partitioned = tuple(partition_all(100000, out_slices))
+            logger.info(f"{rel_path}: {len(out_slices)} total slices, {len(out_slices_partitioned)} batch(es)")
 
             for idx, part in enumerate(out_slices_partitioned):
                 logger.info(f"{rel_path}: {idx + 1} / {len(out_slices_partitioned)}")

src/zarrify/to_zarr.py

@@ -102,24 +102,26 @@ def to_zarr(src : str,
         codec (str, optional): compression codec name ('zstd', 'gzip', 'blosc'). Defaults to 'zstd'.
         codec_level (int | None, optional): codec compression level. None uses per-codec default.
     """
+    logger.info(f"Initializing dataset from {src}")
     dataset = init_dataset(src, axes, scale, translation, units, optimize_reads)
+    logger.info(f"Dataset type: {type(dataset).__name__}")
 
     # Handle N5Group separately as it has custom zarr creation logic
     codec_dict = build_codec(codec, codec_level)
 
     if isinstance(dataset, N5Group):
-        # N5 handles zarr creation internally due to tree structure complexity
-        # RAM check is done per-array inside write_to_zarr (arrays can have different shapes)
+        logger.info("Detected N5Group — scaling workers and starting write")
         client.cluster.scale(workers)
         dataset.write_to_zarr(dest, client, zarr_chunks, shard_shape=shard_shape, codec=codec_dict)
         client.cluster.scale(0)
         return
 
     if isinstance(dataset, Zarr2Group):
-        # RAM check is done per-array inside write_to_zarr
+        logger.info("Detected Zarr2Group — scaling workers and starting write")
         client.cluster.scale(workers)
         dataset.write_to_zarr(str(dest), client, zarr_chunks, shard_shape=shard_shape, codec=codec_dict)
         client.cluster.scale(0)
+        logger.info("Zarr2Group write complete")
         return
     else:
         logger.info(f"Input dataset: {type(dataset)}")