Update Zarr driver to v0.1.6 (#121)

- Update the Zarr driver to v0.1.6. This release fixes a bug where
trailing whitespace in the external metadata fails to validate.
- Break the Zarr tests out into a separate file.

Author: aliddell

.github/workflows/release.yml

@@ -111,6 +111,7 @@ jobs:
       - name: Test artifact
         run: |
           python -m pytest -k test_basic --color=yes --cov-report=xml --cov=acquire --maxfail=5 --log-cli-level=0
+          python -m pytest -k test_zarr --color=yes --cov-report=xml --cov=acquire --maxfail=5 --log-cli-level=0
 
       - name: Upload wheels
         uses: actions/upload-artifact@v3

.github/workflows/test_pr.yml

@@ -56,6 +56,7 @@ jobs:
           ZARR_V3_SHARDING: "1"
         run: |
           python -m pytest -k test_basic --color=yes --cov-report=xml --cov=acquire --maxfail=5 --log-cli-level=0
+          python -m pytest -k test_zarr --color=yes --cov-report=xml --cov=acquire --maxfail=5 --log-cli-level=0
 
   dcam:
     name: Python ${{ matrix.python }} (DCAM)

drivers.json

@@ -1,6 +1,6 @@
 {
   "acquire-driver-common": "0.1.5",
-  "acquire-driver-zarr": "0.1.5",
+  "acquire-driver-zarr": "0.1.6",
   "acquire-driver-egrabber": "0.1.4",
   "acquire-driver-hdcam": "0.1.6",
   "acquire-driver-spinnaker": "0.1.1"

tests/test_basic.py

@@ -7,15 +7,8 @@
 
 import acquire
 from acquire import DeviceKind, DeviceState, Runtime, Trigger, PropertyType
-import dask.array as da
-import numcodecs.blosc as blosc
 import pytest
 import tifffile
-import zarr
-from ome_zarr.io import parse_url
-from ome_zarr.reader import Reader
-from skimage.transform import downscale_local_mean
-import numpy as np
 
 
 @pytest.fixture(scope="module")
@@ -282,308 +275,6 @@ def meta(iframe: int) -> Dict[Any, Any]:
             assert meta(i)["frame_id"] == i
 
 
-def test_write_external_metadata_to_zarr(
-    runtime: Runtime, request: pytest.FixtureRequest
-):
-    dm = runtime.device_manager()
-    p = runtime.get_configuration()
-    p.video[0].camera.identifier = dm.select(
-        DeviceKind.Camera, "simulated.*sin.*"
-    )
-    p.video[0].camera.settings.shape = (33, 47)
-    p.video[0].storage.identifier = dm.select(DeviceKind.Storage, "Zarr")
-    p.video[0].max_frame_count = 4
-    p.video[0].storage.settings.filename = f"{request.node.name}.zarr"
-    metadata = {"hello": "world"}
-    p.video[0].storage.settings.external_metadata_json = json.dumps(metadata)
-    p.video[0].storage.settings.pixel_scale_um = (0.5, 4)
-    p.video[0].storage.settings.chunk_dims_px.width = 33
-    p.video[0].storage.settings.chunk_dims_px.height = 47
-    p.video[0].storage.settings.chunk_dims_px.planes = 4
-
-    p = runtime.set_configuration(p)
-
-    nframes = 0
-    runtime.start()
-    while nframes < p.video[0].max_frame_count:
-        if packet := runtime.get_available_data(0):
-            nframes += packet.get_frame_count()
-            packet = None
-    runtime.stop()
-
-    assert p.video[0].storage.settings.filename
-    store = parse_url(p.video[0].storage.settings.filename)
-    assert store
-    reader = Reader(store)
-    nodes = list(reader())
-
-    # ome-ngff supports multiple images, in separate directories but we only
-    # wrote one.
-    multi_scale_image_node = nodes[0]
-
-    # ome-ngff always stores multi-scale images, but we only have a single
-    # scale/level.
-    image_data = multi_scale_image_node.data[0]
-    assert image_data.shape == (
-        p.video[0].max_frame_count,
-        1,
-        p.video[0].camera.settings.shape[1],
-        p.video[0].camera.settings.shape[0],
-    )
-
-    multi_scale_image_metadata = multi_scale_image_node.metadata
-
-    axes = multi_scale_image_metadata["axes"]
-    axis_names = tuple(a["name"] for a in axes)
-    assert axis_names == ("t", "c", "y", "x")
-
-    axis_types = tuple(a["type"] for a in axes)
-    assert axis_types == ("time", "channel", "space", "space")
-
-    axis_units = tuple(a.get("unit") for a in axes)
-    assert axis_units == (None, None, "micrometer", "micrometer")
-
-    # We only have one multi-scale level and one transform.
-    transform = multi_scale_image_metadata["coordinateTransformations"][0][0]
-    pixel_scale_um = tuple(
-        transform["scale"][axis_names.index(axis)] for axis in ("x", "y")
-    )
-    assert pixel_scale_um == p.video[0].storage.settings.pixel_scale_um
-
-    # ome-zarr only reads attributes it recognizes, so use a plain zarr reader
-    # to read external metadata instead.
-    group = zarr.open(p.video[0].storage.settings.filename)
-    assert group["0"].attrs.asdict() == metadata
-
-
-@pytest.mark.parametrize(
-    ("compressor_name",),
-    [
-        ("zstd",),
-        ("lz4",),
-    ],
-)
-def test_write_compressed_zarr(
-    runtime: Runtime, request: pytest.FixtureRequest, compressor_name
-):
-    filename = f"{request.node.name}.zarr"
-    filename = filename.replace("[", "_").replace("]", "_")
-
-    dm = runtime.device_manager()
-    p = runtime.get_configuration()
-    p.video[0].camera.identifier = dm.select(
-        DeviceKind.Camera, "simulated.*empty.*"
-    )
-    p.video[0].camera.settings.shape = (64, 48)
-    p.video[0].camera.settings.exposure_time_us = 1e4
-    p.video[0].storage.identifier = dm.select(
-        DeviceKind.Storage,
-        f"ZarrBlosc1{compressor_name.capitalize()}ByteShuffle",
-    )
-    p.video[0].max_frame_count = 70
-    p.video[0].storage.settings.filename = filename
-    metadata = {"foo": "bar"}
-    p.video[0].storage.settings.external_metadata_json = json.dumps(metadata)
-    runtime.set_configuration(p)
-
-    runtime.start()
-    runtime.stop()
-
-    # load from Zarr
-    group = zarr.open(p.video[0].storage.settings.filename)
-    data = group["0"]
-
-    assert data.compressor.cname == compressor_name
-    assert data.compressor.clevel == 1
-    assert data.compressor.shuffle == blosc.SHUFFLE
-
-    assert data.shape == (
-        p.video[0].max_frame_count,
-        1,
-        p.video[0].camera.settings.shape[1],
-        p.video[0].camera.settings.shape[0],
-    )
-    assert data.attrs.asdict() == metadata
-
-    # load from Dask
-    data = da.from_zarr(p.video[0].storage.settings.filename, component="0")
-    assert data.shape == (
-        p.video[0].max_frame_count,
-        1,
-        p.video[0].camera.settings.shape[1],
-        p.video[0].camera.settings.shape[0],
-    )
-
-
-@pytest.mark.parametrize(
-    ("number_of_frames", "expected_number_of_chunks", "compression"),
-    [
-        (64, 4, None),
-        (64, 4, {"codec": "zstd", "clevel": 1, "shuffle": 1}),
-        (65, 8, None),  # rollover
-        (65, 8, {"codec": "blosclz", "clevel": 2, "shuffle": 2}),  # rollover
-    ],
-)
-def test_write_zarr_with_chunking(
-    runtime: acquire.Runtime,
-    request: pytest.FixtureRequest,
-    number_of_frames: int,
-    expected_number_of_chunks: int,
-    compression: Optional[dict],
-):
-    dm = runtime.device_manager()
-
-    p = runtime.get_configuration()
-    p.video[0].camera.identifier = dm.select(
-        DeviceKind.Camera, "simulated.*empty.*"
-    )
-    p.video[0].camera.settings.shape = (1920, 1080)
-    p.video[0].camera.settings.exposure_time_us = 1e4
-    p.video[0].camera.settings.pixel_type = acquire.SampleType.U8
-    p.video[0].storage.identifier = dm.select(
-        DeviceKind.Storage,
-        "Zarr",
-    )
-    p.video[0].storage.settings.filename = f"{request.node.name}.zarr"
-    p.video[0].max_frame_count = number_of_frames
-
-    p.video[0].storage.settings.chunk_dims_px.width = 1920 // 2
-    p.video[0].storage.settings.chunk_dims_px.height = 1080 // 2
-    p.video[0].storage.settings.chunk_dims_px.planes = 64
-
-    runtime.set_configuration(p)
-
-    runtime.start()
-    runtime.stop()
-
-    group = zarr.open(p.video[0].storage.settings.filename)
-    data = group["0"]
-
-    assert data.chunks == (64, 1, 1080 // 2, 1920 // 2)
-
-    assert data.shape == (
-        number_of_frames,
-        1,
-        p.video[0].camera.settings.shape[1],
-        p.video[0].camera.settings.shape[0],
-    )
-    assert data.nchunks == expected_number_of_chunks
-
-
-def test_write_zarr_multiscale(
-    runtime: acquire.Runtime,
-    request: pytest.FixtureRequest,
-):
-    filename = f"{request.node.name}.zarr"
-    filename = filename.replace("[", "_").replace("]", "_")
-
-    dm = runtime.device_manager()
-
-    p = runtime.get_configuration()
-    p.video[0].camera.identifier = dm.select(
-        DeviceKind.Camera, "simulated.*empty.*"
-    )
-    p.video[0].camera.settings.shape = (1920, 1080)
-    p.video[0].camera.settings.exposure_time_us = 1e4
-    p.video[0].camera.settings.pixel_type = acquire.SampleType.U8
-    p.video[0].storage.identifier = dm.select(
-        DeviceKind.Storage,
-        "Zarr",
-    )
-    p.video[0].storage.settings.filename = filename
-    p.video[0].storage.settings.pixel_scale_um = (1, 1)
-    p.video[0].max_frame_count = 100
-
-    p.video[0].storage.settings.chunk_dims_px.width = (
-        p.video[0].camera.settings.shape[0] // 3
-    )
-    p.video[0].storage.settings.chunk_dims_px.height = (
-        p.video[0].camera.settings.shape[1] // 3
-    )
-    p.video[0].storage.settings.chunk_dims_px.planes = 64
-
-    p.video[0].storage.settings.enable_multiscale = True
-
-    runtime.set_configuration(p)
-
-    runtime.start()
-    runtime.stop()
-
-    reader = Reader(parse_url(filename))
-    zgroup = list(reader())[0]
-    # loads each layer as a dask array from the Zarr dataset
-    data = [
-        da.from_zarr(filename, component=str(i))
-        for i in range(len(zgroup.data))
-    ]
-    assert len(data) == 3
-
-    image = data[0][0, 0, :, :].compute()  # convert dask array to numpy array
-
-    for d in data:
-        assert (
-            np.linalg.norm(image - d[0, 0, :, :].compute()) == 0
-        )  # validate against the same method from scikit-image
-        image = downscale_local_mean(image, (2, 2)).astype(np.uint8)
-
-
-@pytest.mark.parametrize(
-    ("number_of_frames", "expected_number_of_chunks", "codec"),
-    [
-        (64, 4, None),
-        (64, 4, "zstd"),
-        (65, 8, None),  # rollover
-        (65, 8, "lz4"),  # rollover
-    ],
-)
-def test_write_zarr_v3(
-    runtime: acquire.Runtime,
-    request: pytest.FixtureRequest,
-    number_of_frames: int,
-    expected_number_of_chunks: int,
-    codec: Optional[str],
-):
-    dm = runtime.device_manager()
-
-    p = runtime.get_configuration()
-    p.video[0].camera.identifier = dm.select(
-        DeviceKind.Camera, "simulated.*empty.*"
-    )
-
-    p.video[0].camera.settings.shape = (1920, 1080)
-    p.video[0].camera.settings.exposure_time_us = 1e4
-    p.video[0].camera.settings.pixel_type = acquire.SampleType.U8
-    p.video[0].storage.identifier = dm.select(
-        DeviceKind.Storage,
-        f"ZarrV3Blosc1{codec.capitalize()}ByteShuffle" if codec else "ZarrV3",
-    )
-    p.video[0].storage.settings.filename = f"{request.node.name}.zarr"
-    p.video[0].max_frame_count = number_of_frames
-
-    p.video[0].storage.settings.chunk_dims_px.width = 1920 // 2
-    p.video[0].storage.settings.chunk_dims_px.height = 1080 // 2
-    p.video[0].storage.settings.chunk_dims_px.planes = 64
-
-    runtime.set_configuration(p)
-
-    runtime.start()
-    runtime.stop()
-
-    store = zarr.DirectoryStoreV3(p.video[0].storage.settings.filename)
-    group = zarr.open(store=store, mode="r")
-    data = group["0"]
-
-    assert data.chunks == (64, 1, 1080 // 2, 1920 // 2)
-
-    assert data.shape == (
-        number_of_frames,
-        1,
-        p.video[0].camera.settings.shape[1],
-        p.video[0].camera.settings.shape[0],
-    )
-    assert data.nchunks == expected_number_of_chunks
-
-
 @pytest.mark.skip(
     reason="Runs into memory limitations on github ci."
     + " See https://github.com/acquire-project/cpx/issues/147"