Add extra docs to CodecStack.encode_decode_data_array

Author: juntyr

.github/workflows/ci.yml

@@ -6,7 +6,10 @@ on:
       - main
   pull_request:
     branches:
-      - '*'
+      - "*"
+
+env:
+  CLICOLOR: 1
 
 jobs:
   check:
@@ -21,6 +24,16 @@ jobs:
         with:
           args: check
 
+  spelling:
+    name: Spellcheck
+    runs-on: ubuntu-latest
+    steps:
+    - name: Checkout the Repository
+      uses: actions/checkout@v4
+
+    - name: Spellcheck repo
+      uses: crate-ci/typos@v1.32.0
+
   fmt:
     name: Formatting
     runs-on: ubuntu-latest
@@ -50,7 +63,7 @@ jobs:
         run: uv sync --all-extras --dev && uv pip install .
 
       - name: Run tests
-        run: uv run pytest
+        run: uv run pytest -v -W error
 
   mypy:
     name: Typecheck

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "numcodecs-combinators"
-version = "0.2.7"
+version = "0.2.8"
 description = "Combinator codecs for the `numcodecs` buffer compression API"
 readme = "README.md"
 license = { file = "LICENSE" }

src/numcodecs_combinators/stack.py

@@ -167,7 +167,14 @@ def encode_decode(self, buf: Buffer) -> Buffer:
 
     def encode_decode_data_array(self, da: "xr.DataArray") -> "xr.DataArray":
         """
-        Encode, then decode the data array in `da`.
+        Encode, then decode each chunk (independently) in the data array `da`.
+
+        Since each chunk is encoded *independently*, this method may cause
+        chunk boundary artifacts. Do *not* use this method if the codec
+        requires access to the entire data at once or if it needs to access
+        a neighbourhood of points across the chunk boundary. In these cases,
+        it is preferable to use
+        `da.copy(data=stack.encode_decode(da.values))` instead.
 
         The encode-decode computation may be deferred until the
         [`compute`][xarray.DataArray.compute] method is called on the result.