Merge pull request #799 from DHI/dataset_from_numpy

Dataset_from_numpy

Author: ecomodeller

docs/examples/dfsu/merge_subdomains.qmd

@@ -132,11 +132,11 @@ combined_geometry.plot()
 ## Merge everything into dataset
 
 ```{python}
-ds_out = mikeio.Dataset(
+ds_out = mikeio.Dataset.from_numpy(
     data=merged_data,  # n_items, timesteps, n_elements
-    items=items, 
+    items=items,
     time=time_steps,
-    geometry=combined_geometry
+    geometry=combined_geometry,
 )
 ```
 

docs/user-guide/dataset.qmd

@@ -34,7 +34,7 @@ Selecting a specific item "itemA" (at position 0) from a Dataset ds can be done
 * `ds[0]` - returns "itemA" DataArray
 * `ds.itemA` - returns "itemA" DataArray
 
-We recommend the use *named* items for readability. 
+We recommend to use *named* items for readability. 
 
 ```{python}
 ds.Surface_elevation

mikeio/_track.py

@@ -156,7 +156,7 @@ def is_EOF(step: int) -> bool:
     for item_info in items:
         items_out.append(item_info)
 
-    return Dataset(data_list, times, items_out)
+    return Dataset.from_numpy(data=data_list, time=times, items=items_out)
 
 
 def _get_track_data_from_dataset(track: Dataset) -> tuple[pd.DatetimeIndex, np.ndarray]:

mikeio/dataset/_dataarray.py

@@ -269,8 +269,6 @@ def _parse_geometry(
             if dims == ("time", "x"):
                 return Grid1D(nx=shape[1], dx=1.0 / (shape[1] - 1))
 
-            warnings.warn("Geometry is required for ndim >=1")
-
         axis = 1 if "time" in dims else 0
         # dims_no_time = tuple([d for d in dims if d != "time"])
         # shape_no_time = shape[1:] if ("time" in dims) else shape

mikeio/dataset/_dataset.py

@@ -64,8 +64,8 @@ class Dataset:
         By providing a mapping of data arrays, the remaining parameters are not needed
     time:
         a pandas.DatetimeIndex with the time instances of the data
-    item:
-        an ItemInfo with name, type and unit
+    items:
+        a list of ItemInfo with name, type and unit
     geometry:
         a geometry object e.g. Grid2D or GeometryFM2D
     zn:
@@ -113,6 +113,10 @@ def __init__(
         dt: float = 1.0,
     ):
         if not self._is_DataArrays(data):
+            warnings.warn(
+                "Supplying data as a list of numpy arrays is deprecated. Use Dataset.from_numpy",
+                FutureWarning,
+            )
             data = self._create_dataarrays(
                 data=data,
                 time=time,
@@ -122,12 +126,36 @@ def __init__(
                 dims=dims,
                 dt=dt,
             )
-        self._data_vars: MutableMapping[str, DataArray] = self._init_from_DataArrays(
+        self._data_vars = self._init_from_DataArrays(
             data,  # type: ignore
             validate=validate,
-        )  # type: ignore
+        )
         self.plot = _DatasetPlotter(self)
 
+    @staticmethod
+    def from_numpy(
+        data: Sequence[NDArray[np.floating]],
+        time: pd.DatetimeIndex | None = None,
+        items: Sequence[ItemInfo] | None = None,
+        *,
+        geometry: Any | None = None,
+        zn: NDArray[np.floating] | None = None,
+        dims: tuple[str, ...] | None = None,
+        validate: bool = True,
+        dt: float = 1.0,
+    ) -> Dataset:
+        das = Dataset._create_dataarrays(
+            data=data,
+            time=time,
+            items=items,
+            geometry=geometry,
+            zn=zn,
+            dims=dims,
+            dt=dt,
+        )
+
+        return Dataset(das)
+
     @staticmethod
     def _is_DataArrays(data: Any) -> bool:
         """Check if input is Sequence/Mapping of DataArrays."""
@@ -179,8 +207,6 @@ def _init_from_DataArrays(
             for da in data_vars.values():
                 first._is_compatible(da, raise_error=True)
 
-        # TODO is it necessary to keep track of item names?
-        self.__itemattr: set[str] = set()
         for key, value in data_vars.items():
             self._set_name_attr(key, value)
 
@@ -644,15 +670,11 @@ def rename(self, mapper: Mapping[str, str], inplace: bool = False) -> "Dataset":
 
     def _set_name_attr(self, name: str, value: DataArray) -> None:
         name = _to_safe_name(name)
-        if name not in self.__itemattr:
-            self.__itemattr.add(name)  # keep track of what we insert
         setattr(self, name, value)
 
     def _del_name_attr(self, name: str) -> None:
         name = _to_safe_name(name)
-        if name in self.__itemattr:
-            self.__itemattr.remove(name)
-            delattr(self, name)
+        delattr(self, name)
 
     @overload
     def __getitem__(self, key: Hashable | int) -> DataArray: ...
@@ -1359,7 +1381,7 @@ def _concat_time(
                 zn[idx2, :] = other._zn
                 zn[idx1, :] = self._zn
 
-        return Dataset(
+        return Dataset.from_numpy(
             newdata, time=newtime, items=ds.items, geometry=ds.geometry, zn=zn
         )
 

mikeio/dfs/_dfs0.py

@@ -175,7 +175,7 @@ def read(
             fitems = [self.items[it] for it in item_numbers]
         else:
             fitems = self.items
-        ds = Dataset(fdata, ftime, fitems, validate=False)
+        ds = Dataset.from_numpy(fdata, time=ftime, items=fitems, validate=False)
 
         # select time steps
         self._n_timesteps = dfs.FileInfo.TimeAxis.NumberOfTimeSteps

mikeio/dfs/_dfs1.py

@@ -175,7 +175,7 @@ def read(
 
         self._dfs.Close()
 
-        return Dataset(
+        return Dataset.from_numpy(
             data=data_list,
             time=time,
             items=items,

mikeio/dfs/_dfs2.py

@@ -243,7 +243,7 @@ def read(
         else:
             dims = ("time", "y", "x")
 
-        return Dataset(
+        return Dataset.from_numpy(
             data_list,
             time=time,
             items=items,

mikeio/dfs/_dfs3.py

@@ -267,7 +267,7 @@ def read(
 
         time = pd.to_datetime(t_seconds, unit="s", origin=self.start_time)
         items = _get_item_info(dfs.ItemInfo, item_numbers)
-        return Dataset(
+        return Dataset.from_numpy(
             data_list,
             time=time,
             items=items,

mikeio/dfsu/_dfsu.py

@@ -414,10 +414,10 @@ def read(
         time = pd.to_datetime(t_rel, unit="s", origin=self.start_time)
         item_infos = _get_item_info(dfs.ItemInfo, item_numbers)
 
-        return Dataset(
+        return Dataset.from_numpy(
             data_list,
-            time,
-            item_infos,
+            time=time,
+            items=item_infos,
             geometry=geometry,
             dims=dims,
             validate=False,