Making distributed compatible with the delayed parameter and changing the default value of delayed (False if distributed is False, True otherwise).

Author: Thomas Zilio

zcollection/collection/__init__.py

@@ -622,7 +622,7 @@ def update(
         filters: PartitionFilter | None = None,
         trim: bool = True,
         npartitions: int | None = None,
-        delayed: bool = True,
+        delayed: bool | None = None,
         distributed: bool = True,
         **kwargs,
     ) -> None:
@@ -656,6 +656,7 @@ def update(
                 default, it is equal to the number of Dask workers available
                 when calling this method.
             delayed: Whether to load data in a dask array or not.
+                Default value is True if distributed is True, False otherwise.
             distributed: Whether to use dask or not. Default To True.
             **kwargs: The keyword arguments to pass to the function.
 
@@ -673,9 +674,8 @@ def update(
         if not callable(func):
             raise TypeError('func must be a callable')
 
-        # Delayed has to be False if dask is disabled
-        if not distributed:
-            delayed = False
+        if delayed is None:
+            delayed = distributed
 
         variables = variables or _infer_callable(
             self, func, filters, delayed, selected_variables, *args, **kwargs)

zcollection/collection/abc.py

@@ -644,7 +644,7 @@ def _wrap(
     def load(
         self,
         *,
-        delayed: bool = True,
+        delayed: bool | None = None,
         filters: PartitionFilter = None,
         indexer: Indexer | None = None,
         selected_variables: Iterable[str] | None = None,
@@ -656,6 +656,7 @@ def load(
 
         Args:
             delayed: Whether to load data in a dask array or not.
+                Default value is True if distributed is True, False otherwise.
             filters: The predicate used to filter the partitions to load. To
                 get more information on the predicate, see the documentation of
                 the :meth:`partitions` method.
@@ -683,11 +684,10 @@ def load(
             ...     filters=lambda keys: keys["year"] == 2019 and
             ...     keys["month"] == 3 and keys["day"] % 2 == 0)
         """
-        array: dataset.Dataset | None = None
+        if delayed is None:
+            delayed = distributed
 
-        # Delayed has to be False if dask is disabled
-        if not distributed:
-            delayed = False
+        array: dataset.Dataset | None = None
 
         if indexer is None:
             arrays = self._load_partitions(

zcollection/view/__init__.py

@@ -422,7 +422,7 @@ def drop_variable(
     def load(
         self,
         *,
-        delayed: bool = True,
+        delayed: bool | None = None,
         filters: collection.PartitionFilter = None,
         indexer: collection.Indexer | None = None,
         selected_variables: Iterable[str] | None = None,
@@ -433,6 +433,7 @@ def load(
 
         Args:
             delayed: Whether to load data in a dask array or not.
+                Default value is True if distributed is True, False otherwise.
             filters: The predicate used to filter the partitions to select.
                 To get more information on the predicate, see the
                 documentation of the :meth:`Collection.partitions
@@ -452,9 +453,8 @@ def load(
             >>> view.load(filters="time == '2020-01-01'")
             >>> view.load(filters=lambda x: x["time"] == "2020-01-01")
         """
-        # Delayed has to be False if dask is disabled
-        if not distributed:
-            delayed = False
+        if delayed is None:
+            delayed = distributed
 
         array: dataset.Dataset | None = None
         datasets: list[tuple[dataset.Dataset, str] | None]