More tests

.github/workflows/ci-tests.yml

@@ -17,6 +17,9 @@ jobs:
       run:
         shell: bash -l {0}
 
+    env:
+      IRIS_TEST_DATA_VERSION: "2.20"
+
     steps:
       # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
       - uses: actions/checkout@v3
@@ -50,6 +53,14 @@ jobs:
           conda info
           conda list --explicit
 
+      - name: "Fetch Iris-test-data"
+        run: |
+          wget --quiet https://github.com/SciTools/iris-test-data/archive/v${IRIS_TEST_DATA_VERSION}.zip -O iris-test-data.zip
+          unzip -q iris-test-data.zip
+          mkdir --parents ${GITHUB_WORKSPACE}/iris_test_data
+          mv iris-test-data-${IRIS_TEST_DATA_VERSION} ${GITHUB_WORKSPACE}/iris_test_data_download
+
       - name: "Run tests"
         run: |
-          PYTHONPATH=./tests:$PYTHONPATH pytest -v ./tests
+          ls ${GITHUB_WORKSPACE}/iris_test_data_download/test_data
+          OVERRIDE_TEST_DATA_REPOSITORY=${GITHUB_WORKSPACE}/iris_test_data_download/test_data PYTHONPATH=./tests:$PYTHONPATH pytest -v ./tests

lib/ncdata/_core.py

@@ -295,7 +295,7 @@ def _print_value(self):
 
         # Convert numpy non-string scalars to simple Python values, in string output.
         if getattr(value, "shape", None) in ((0,), (1,), ()):
-            op = {"i": int, "f": float}[value.dtype.kind]
+            op = {"i": int, "u": int, "f": float}[value.dtype.kind]
             value = op(value.flatten()[0])
 
         return repr(value)

lib/ncdata/dataset_like.py

@@ -90,6 +90,9 @@ class Nc4DatasetLike(_Nc4DatalikeWithNcattrs):
 
     _local_instance_props = ("_ncdata", "variables", "dimensions")
 
+    # Needed for Iris to recognise the dataset format.
+    file_format = "NETCDF4"
+
     def __init__(self, ncdata: NcData = None):
         if ncdata is None:
             ncdata = NcData()  # an empty dataset
@@ -134,13 +137,26 @@ def createVariable(
             msg = f'creating duplicate variable "{varname}".'
             raise ValueError(msg)
         # Add a variable into the underlying NcData object.
+
+        # N.B. to correctly mirror netCDF4, a variable must be created with all-masked
+        # content.  For this we need to decode the dims + work out the shape.
+        # NOTE: simplistic version here, as we don't support groups.
+        shape = tuple(
+            self._ncdata.dimensions[dim_name].size for dim_name in dimensions
+        )
+        # Note: does *not* allocate a full array in memory ...until you modify it.
+        initial_allmasked_data = np.ma.masked_array(
+            np.zeros(shape, dtype=datatype), mask=True
+        )
+
         ncvar = NcVariable(
             name=varname,
             dimensions=dimensions,
+            data=initial_allmasked_data,
             group=self._ncdata,
         )
-        # Note: initially has no data (or attributes), since this is how netCDF4 expects
-        #  to do it.
+        # Note: no valid data is initially assigned, since that is how the netCDF4 API
+        # does it.
         self._ncdata.variables[varname] = ncvar
         # Create a netCDF4-like "wrapper" variable + install that here.
         nc4var = Nc4VariableLike._from_ncvariable(ncvar, dtype=datatype)
@@ -155,10 +171,9 @@ def close(self):  # noqa: D102
 
     @staticmethod
     def filepath() -> str:  # noqa: D102
-        #
         # Note: for now, let's just not care about this.
         # we *might* need this to be an optional defined item on an NcData ??
-        # .. or, we ight need to store an xarray "encoding" somewhere ?
+        # .. or, we might need to store an xarray "encoding" somewhere ?
         # TODO: more thought here ?
         # return self.ncdata.encoding.get("source", "")
         return "<Nc4DatasetLike>"
@@ -225,7 +240,11 @@ def __getitem__(self, keys):  # noqa: D105
             raise IndexError(keys)
         if self.ndim == 0:
             return self._ncdata.data
-        return self._ncdata.data[keys]
+        array = self._ncdata.data[keys]
+        if hasattr(array, "compute"):
+            # When accessed as a data variable, we must realise lazy data.
+            array = array.compute()
+        return array
 
     # The __setitem__ is not required for normal saving.
     # The saver will assign ._data_array instead