Refactor API for NGFF datasets (#31)

* base class for ngff nodes * fix typo * sketch image container and position node * create new image in position * add docstrings * combine ome-zarr reader and writer * fix docstring length * append channel method * mixin class for file mode methods * handle metadata parsing in node classes * move ome-zarr-py and hypothesis to dev dependency * move hypothesis to dev dependency * match signature * fix constructors * fix metadata writing * fix default chunking * fix context manager * fix channel axis check * fix shape check * update OMEZarr example * rename ngff file * rename imports * showcase `ImageArray.numpy()` * update docstring * remove old writing code and add new well node * update well constructor and docstring * add row node * update docstring * elevate iterator * dummy metadata parsing for the row node * update docstrings * update example introduction * clarify read-only example * fix hcs metadata * add back dump metadata method as placeholder * plate node * fix typo * init plate meta in its own method * update docstring * replace old HCSWriter with a wrapper of plate node * remove reader adaptor * remove OMEZarrReader * update deprecation message * update docstring * rename example files * rearrange metadata attributes * fix plate metadata * fix metadata loading * update hcs example * fix omero metadata * rename test directory * rename test file * change error type * test util functions * fix linting and formatting * test OMEZarr * warning against invalid plate metadata * warn -> warning * warning against channel name not found * test `HCSZarr.open()` * handle missing axes and channel names * load axes from metadata if possible * fix check attribute logic * fix typo * test HCSZarr * return early if position is not found * match error type * restrict example count and loosen deadline * combine pytest commands * fix fixture call * remove unused import * update example comment * fix linting * fix linting * fix linting * Merge branch 'main' into ngff-dataset * Revert "Merge branch 'main' into ngff-dataset" This reverts commit c17b81e. * isort * remove empty files from merging * fix linting * simple refactors and an example that fails * revert wrong example changes * fix linting and formatting * simplify code * revert to conditional operator * use string as array key in examples * rename OMEZarr to OMEZarrFOV * Update docstrings * test append channel * check image shape with test * test create position * test well metadata * updated hcs_zarr example * move open store function to top level * rename open_store to aid auto-completion * wrap zarr.Group.tree() for ngff nodes * remove mixin and container classes * test against new api * update examples * refined examples and delete_channel stub * rename (ome_zarr -> singleFOV_zarr) * proposal for simplified API for tiled acquisition * Revert 3 commits This reverts commits 6d0ba4c, b9516b1, 61237f9. * rename single fov example file * modify example store path and array name * edit HCS example * rename hcs example file --------- Co-authored-by: Shalin Mehta <2934183+mattersoflight@users.noreply.github.com>
czbiohub-sf · Feb 13, 2023 · 4a9e6fc · 4a9e6fc
1 parent d8521e0
commit 4a9e6fc
Show file tree

Hide file tree

Showing 25 changed files with 1,862 additions and 1,446 deletions.
diff --git a/.github/workflows/pr.yml b/.github/workflows/pr.yml
@@ -51,8 +51,9 @@ jobs:
           python -m pip install --upgrade pip
           pip install flake8
       - name: Check code with Flake8
+        # E203 conflicts with black
         run: |
-          flake8 iohub
+          flake8 iohub --extend-ignore=E203
 
   isort:
     name: isort Check
@@ -72,7 +73,7 @@ jobs:
         run: |
           python -m pip install --upgrade pip
           pip install isort
-      - name: Check code with Flake8
+      - name: Check code with isort
         run: |
           isort --check iohub
 

diff --git a/examples/hcs_ome_zarr_writer.py b/examples/hcs_ome_zarr_writer.py
diff --git a/examples/multi_fov_hcs_ome_zarr.py b/examples/multi_fov_hcs_ome_zarr.py
@@ -0,0 +1,61 @@
+# %%
+# This script writes a high content screening (HCS) OME-Zarr dataset
+# with a single FOV and a single scaling level per well,
+# and adds an extra well-position to an existing dataset.
+# It can be run as a plain Python script, or as interactive cells in some IDEs.
+
+import os
+
+import numpy as np
+
+from iohub.ngff import open_ome_zarr
+
+# %%
+# Set storage path
+
+store_path = f'{os.path.expanduser("~/")}hcs.zarr'
+
+# %%
+# Write 5D data to multiple wells.
+# Integer path names will be automatically converted to strings.
+# While the NGFF specification (and iohub) allows for arbitrary names,
+# the ome-zarr-py library and the napari-ome-zarr viewer
+# can only load positions and arrays with name '0' for the whole plate.
+
+position_list = (
+    ("A", "1", "0"),
+    ("H", 1, "0"),
+    ("H", "12", "CannotVisualize"),
+    ("Control", "Blank", 0),
+)
+
+with open_ome_zarr(
+    store_path,
+    layout="hcs",
+    mode="w-",
+    channel_names=["DAPI", "GFP", "Brightfield"],
+) as dataset:
+    # Create and write to positions
+    # This affects the tile arrangement in visualization
+    for row, col, fov in position_list:
+        position = dataset.create_position(row, col, fov)
+        position["0"] = np.random.randint(
+            0, np.iinfo(np.uint16).max, size=(5, 3, 2, 32, 32), dtype=np.uint16
+        )
+    # Print dataset summary
+    dataset.print_tree()
+
+# %%
+# Append a channel to all the positions
+
+with open_ome_zarr(store_path, mode="r+") as dataset:
+    for name, position in dataset.positions():
+        print(f"Appending a channel to position: {name}")
+        position.append_channel("Segmentation", resize_arrays=True)
+        position["0"][:, 3] = np.random.randint(
+            0, np.iinfo(np.uint16).max, size=(5, 2, 32, 32), dtype=np.uint16
+        )
+    dataset.print_tree()
+
+# %%
+# Try viewing the images with napari-ome-zarr
diff --git a/examples/ome_zarr_writer.py b/examples/ome_zarr_writer.py
diff --git a/examples/single_fov_ome_zarr.py b/examples/single_fov_ome_zarr.py
@@ -0,0 +1,70 @@
+# %%
+# This script shows how to create a single-FOV, single-scale OME-Zarr dataset,
+# read data in read-only mode,
+# append an extra time point to an existing dataset,
+# and adding a new channel to an existing dataset.
+# It can be run as a plain Python script,
+# or as interactive cells in some IDEs.
+
+import os
+
+import numpy as np
+
+from iohub.ngff import open_ome_zarr
+
+# %%
+# Set storage path
+
+store_path = f'{os.path.expanduser("~/")}hcs.zarr'
+
+# %%
+# Write 5D data to a new Zarr store
+
+tczyx = np.random.randint(
+    0, np.iinfo(np.uint16).max, size=(5, 2, 3, 32, 32), dtype=np.uint16
+)
+
+with open_ome_zarr(
+    store_path, layout="fov", mode="a", channel_names=["DAPI", "GFP"]
+) as dataset:
+    dataset["img"] = tczyx
+
+# %%
+# Opening in read-only mode prevents writing
+
+with open_ome_zarr(store_path, layout="auto", mode="r") as dataset:
+    img = dataset["img"]
+    print(img)
+    print(img.numpy())
+    try:
+        img[0, 0, 0, 0, 0] = 0
+    except Exception as e:
+        print(f"Writing was rejected: {e}")
+
+# %%
+# Append a new timepoint to an existing dataset
+
+new_1czyx = np.random.randint(
+    0, np.iinfo(np.uint16).max, size=(1, 2, 3, 32, 32), dtype=np.uint16
+)
+
+with open_ome_zarr(store_path, layout="fov", mode="r+") as dataset:
+    img = dataset["img"]
+    print(img.shape)
+    img.append(new_1czyx, axis=0)
+    print(img.shape)
+
+# %%
+# Add a new channel and write a Z-stack
+
+new_zyx = np.random.randint(
+    0, np.iinfo(np.uint16).max, size=(3, 32, 32), dtype=np.uint16
+)
+
+dataset = open_ome_zarr(store_path, mode="r+")
+dataset.append_channel("New", resize_arrays=True)
+dataset["img"][0, 2] = new_zyx
+dataset.close()
+
+# %%
+# Try viewing the images with napari-ome-zarr
diff --git a/iohub/lf_utils.py b/iohub/lf_utils.py
@@ -12,7 +12,8 @@ def channel_display_settings(
     clim: Tuple[float, float, float, float] = None,
     first_chan: bool = False,
 ):
-    """This will create a dictionary used for OME-zarr metadata.  Allows custom contrast limits and channel
+    """This will create a dictionary used for OME-zarr metadata.
+    Allows custom contrast limits and channel.
     names for display. Defaults everything to grayscale.
 
     Parameters
@@ -22,7 +23,9 @@ def channel_display_settings(
     clim : Tuple[float, float, float, float], optional
         Contrast limits (start, end, min, max)
     first_chan : bool, optional
-        Whether or not this is the first channel of the dataset (display will be set to active), by default False
+        Whether or not this is the first channel of the dataset
+        (display will be set to active),
+        by default False
 
     Returns
     -------