Duck array documentation improvements

doc/howdoi.rst

@@ -42,7 +42,7 @@ How do I ...
    * - extract the underlying array (e.g. NumPy or Dask arrays)
      - :py:attr:`DataArray.data`
    * - convert to and extract the underlying NumPy array
-     - :py:attr:`DataArray.values`
+     - :py:attr:`DataArray.to_numpy`
    * - convert to a pandas DataFrame
      - :py:attr:`Dataset.to_dataframe`
    * - sort values

doc/internals/duck-arrays-integration.rst

@@ -1,23 +1,55 @@
 
-.. _internals.duck_arrays:
+.. _internals.duckarrays:
 
 Integrating with duck arrays
 =============================
 
 .. warning::
 
-    This is a experimental feature.
+    This is an experimental feature. Please report any bugs or other difficulties on `xarray's issue tracker <https://github.com/pydata/xarray/issues>`_.
 
-Xarray can wrap custom :term:`duck array` objects as long as they define numpy's
-``shape``, ``dtype`` and ``ndim`` properties and the ``__array__``,
-``__array_ufunc__`` and ``__array_function__`` methods.
+Xarray can wrap custom numpy-like arrays (":term:`duck array`\s") - see the :ref:`user guide documentation <userguide.duckarrays>`.
+This page is intended for developers who are interested in wrapping a new custom array type with xarray.
+
+Duck array requirements
+~~~~~~~~~~~~~~~~~~~~~~~
+
+Xarray does not explicitly check that required methods are defined by the underlying duck array object before
+attempting to wrap the given array. However, a wrapped array type should at a minimum define these attributes:
+
+* ``shape`` property,
+* ``dtype`` property,
+* ``ndim`` property,
+* ``__array__`` method,
+* ``__array_ufunc__`` method,
+* ``__array_function__`` method.
+
+These need to be defined consistently with :py:class:`numpy.ndarray`, for example the array ``shape``
+property needs to obey `numpy's broadcasting rules <https://numpy.org/doc/stable/user/basics.broadcasting.html>`_
+(see also the `Python Array API standard's explanation <https://data-apis.org/array-api/latest/API_specification/broadcasting.html>`_
+of these same rules).
+
+Python Array API standard support
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+As an integration library xarray benefits greatly from the standardization of duck-array libraries' APIs, and so is a
+big supporter of the `Python Array API Standard <https://data-apis.org/array-api/latest/>`_. .
+
+We aim to support any array libraries that follow the Array API standard out-of-the-box. However, xarray does occasionally
+call some numpy functions which are not (yet) part of the standard (e.g. :py:meth:`xarray.DataArray.pad` calls :py:func:`numpy.pad`).
+See `xarray issue #7848 <https://github.com/pydata/xarray/issues/7848>`_ for a list of such functions. We can still support dispatching on these functions through
+the array protocols above, it just means that if you exclusively implement the methods in the Python Array API standard
+then some features in xarray will not work.
+
+Custom inline reprs
+~~~~~~~~~~~~~~~~~~~
 
 In certain situations (e.g. when printing the collapsed preview of
 variables of a ``Dataset``), xarray will display the repr of a :term:`duck array`
 in a single line, truncating it to a certain number of characters. If that
 would drop too much information, the :term:`duck array` may define a
 ``_repr_inline_`` method that takes ``max_width`` (number of characters) as an
-argument:
+argument
 
 .. code:: python
 

doc/internals/extending-xarray.rst

@@ -1,4 +1,6 @@
 
+.. _internals.accessors:
+
 Extending xarray using accessors
 ================================
 

doc/internals/index.rst

@@ -8,6 +8,12 @@ stack, NumPy and pandas. It is written in pure Python (no C or Cython
 extensions), which makes it easy to develop and extend. Instead, we push
 compiled code to :ref:`optional dependencies<installing>`.
 
+The pages in this section are intended for:
+
+* Contributors to xarray who wish to better understand some of the internals,
+* Developers who wish to extend xarray with domain-specific logic, perhaps to support a new scientific community of users,
+* Developers who wish to interface xarray with their existing tooling, e.g. by creating a plugin for reading a new file format, or wrapping a custom array type.
+
 
 .. toctree::
    :maxdepth: 2

doc/user-guide/data-structures.rst

@@ -19,7 +19,8 @@ DataArray
 :py:class:`xarray.DataArray` is xarray's implementation of a labeled,
 multi-dimensional array. It has several key properties:
 
-- ``values``: a :py:class:`numpy.ndarray` holding the array's values
+- ``values``: a :py:class:`numpy.ndarray` or
+  :ref:`numpy-like array <userguide.duckarrays>` holding the array's values
 - ``dims``: dimension names for each axis (e.g., ``('x', 'y', 'z')``)
 - ``coords``: a dict-like container of arrays (*coordinates*) that label each
   point (e.g., 1-dimensional arrays of numbers, datetime objects or
@@ -46,7 +47,8 @@ Creating a DataArray
 The :py:class:`~xarray.DataArray` constructor takes:
 
 - ``data``: a multi-dimensional array of values (e.g., a numpy ndarray,
-  :py:class:`~pandas.Series`, :py:class:`~pandas.DataFrame` or ``pandas.Panel``)
+  a :ref:`numpy-like array <userguide.duckarrays>`, :py:class:`~pandas.Series`,
+  :py:class:`~pandas.DataFrame` or ``pandas.Panel``)
 - ``coords``: a list or dictionary of coordinates. If a list, it should be a
   list of tuples where the first element is the dimension name and the second
   element is the corresponding coordinate array_like object.

doc/user-guide/duckarrays.rst

@@ -1,30 +1,183 @@
 .. currentmodule:: xarray
 
+.. _userguide.duckarrays:
+
 Working with numpy-like arrays
 ==============================
 
+NumPy-like arrays (often known as :term:`duck array`\s) are drop-in replacements for the :py:class:`numpy.ndarray`
+class but with different features, such as propagating physical units or a different layout in memory.
+Xarray can often wrap these array types, allowing you to use labelled dimensions and indexes whilst benefiting from the
+additional features of these array libraries.
+
+Some numpy-like array types that xarray already has some support for:
+
+* `Cupy <https://cupy.dev/>`_ - GPU support (see `cupy-xarray <https://cupy-xarray.readthedocs.io>`_),
+* `Sparse <https://sparse.pydata.org/en/stable/>`_ - for performant arrays with many zero elements,
+* `Pint <https://pint.readthedocs.io/en/latest/>`_ - for tracking the physical units of your data (see `pint-xarray <https://pint-xarray.readthedocs.io>`_),
+* `Dask <https://docs.dask.org/en/stable/>`_ - parallel computing on larger-than-memory arrays (see :ref:`using dask with xarray <dask>`),
+* `Cubed <https://github.com/tomwhite/cubed/tree/main/cubed>`_ - another parallel computing framework that emphasises reliability (see `cubed-xarray <https://github.com/cubed-xarray>`_).
+
 .. warning::
 
-   This feature should be considered experimental. Please report any bug you may find on
-   xarray’s github repository.
+   This feature should be considered somewhat experimental. Please report any bugs you find on
+   `xarray’s issue tracker <https://github.com/pydata/xarray/issues>`_.
+
+.. note::
+
+    For information on wrapping dask arrays see :ref:`dask`. Whilst xarray wraps dask arrays in a similar way to that
+    described on this page, chunked array types like :py:class:`dask.array.Array` implement additional methods that require
+    slightly different user code (e.g. calling ``.chunk`` or ``.compute``).
+
+Why "duck"?
+-----------
+
+Why is it also called a "duck" array? This comes from a common statement of object-oriented programming -
+"If it walks like a duck, and quacks like a duck, treat it like a duck". In other words, a library like xarray that
+is capable of using multiple different types of arrays does not have to explicitly check that each one it encounters is
+permitted (e.g. ``if dask``, ``if numpy``, ``if sparse`` etc.). Instead xarray can take the more permissive approach of simply
+treating the wrapped array as valid, attempting to call the relevant methods (e.g. ``.mean()``) and only raising an
+error if a problem occurs (e.g. the method is not found on the wrapped class). This is much more flexible, and allows
+objects and classes from different libraries to work together more easily.
+
+What is a numpy-like array?
+---------------------------
+
+A "numpy-like array" (also known as a "duck array") is a class that contains array-like data, and implements key
+numpy-like functionality such as indexing, broadcasting, and computation methods.
+
+For example, the `sparse <https://sparse.pydata.org/en/stable/>`_ library provides a sparse array type which is useful for representing nD array objects like sparse matrices
+in a memory-efficient manner. We can create a sparse array object (of the :py:class:`sparse.COO` type) from a numpy array like this:
+
+.. ipython:: python
+
+    from sparse import COO
+
+    x = np.eye(4, dtype=np.uint8)  # create diagonal identity matrix
+    s = COO.from_numpy(x)
+    s
 
-NumPy-like arrays (:term:`duck array`) extend the :py:class:`numpy.ndarray` with
-additional features, like propagating physical units or a different layout in memory.
+This sparse object does not attempt to explicitly store every element in the array, only the non-zero elements.
+This approach is much more efficient for large arrays with only a few non-zero elements (such as tri-diagonal matrices).
+Sparse array objects can be converted back to a "dense" numpy array by calling :py:meth:`sparse.COO.todense`.
 
-:py:class:`DataArray` and :py:class:`Dataset` objects can wrap these duck arrays, as
-long as they satisfy certain conditions (see :ref:`internals.duck_arrays`).
+Just like :py:class:`numpy.ndarray` objects, :py:class:`sparse.COO` arrays support indexing
+
+.. ipython:: python
+
+    s[1, 1]  # diagonal elements should be ones
+    s[2, 3]  # off-diagonal elements should be zero
+
+broadcasting,
+
+.. ipython:: python
+
+    x2 = np.zeros(
+        (4, 1), dtype=np.uint8
+    )  # create second sparse array of different shape
+    s2 = COO.from_numpy(x2)
+    (s * s2)  # multiplication requires broadcasting
+
+and various computation methods
+
+.. ipython:: python
+
+    s.sum(axis=1)
+
+This numpy-like array also supports calling so-called `numpy ufuncs <https://numpy.org/doc/stable/reference/ufuncs.html#available-ufuncs>`_
+("universal functions") on it directly:
+
+.. ipython:: python
+
+    np.sum(s, axis=1)
+
+
+Notice that in each case the API for calling the operation on the sparse array is identical to that of calling it on the
+equivalent numpy array - this is the sense in which the sparse array is "numpy-like".
 
 .. note::
 
-    For ``dask`` support see :ref:`dask`.
+    For discussion on exactly which methods a class needs to implement to be considered "numpy-like", see :ref:`internals.duckarrays`.
+
+Wrapping numpy-like arrays in xarray
+------------------------------------
+
+:py:class:`DataArray`, :py:class:`Dataset`, and :py:class:`Variable` objects can wrap these numpy-like arrays.
 
+Constructing xarray objects which wrap numpy-like arrays
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-Missing features
-----------------
-Most of the API does support :term:`duck array` objects, but there are a few areas where
-the code will still cast to ``numpy`` arrays:
+The primary way to create an xarray object which wraps a numpy-like array is to pass that numpy-like array instance directly
+to the constructor of the xarray class. The :ref:`page on xarray data structures <data structures>` shows how :py:class:`DataArray` and :py:class:`Dataset`
+both accept data in various forms through their ``data`` argument, but in fact this data can also be any wrappable numpy-like array.
 
-- dimension coordinates, and thus all indexing operations:
+For example, we can wrap the sparse array we created earlier inside a new DataArray object:
+
+.. ipython:: python
+
+    s_da = xr.DataArray(s, dims=["i", "j"])
+    s_da
+
+We can see what's inside - the printable representation of our xarray object (the repr) automatically uses the printable
+representation of the underlying wrapped array.
+
+Of course our sparse array object is still there underneath - it's stored under the ``.data`` attribute of the dataarray:
+
+.. ipython:: python
+
+    s_da.data
+
+Array methods
+~~~~~~~~~~~~~
+
+We saw above that numpy-like arrays provide numpy methods. Xarray automatically uses these when you call the corresponding xarray method:
+
+.. ipython:: python
+
+    s_da.sum(dim="j")
+
+Converting wrapped types
+~~~~~~~~~~~~~~~~~~~~~~~~
+
+If you want to change the type inside your xarray object you can use :py:meth:`DataArray.as_numpy`:
+
+.. ipython:: python
+
+    s_da.as_numpy()
+
+This returns a new :py:class:`DataArray` object, but now wrapping a normal numpy array.
+
+If instead you want to convert to numpy and return that numpy array you can use either :py:meth:`DataArray.to_numpy` or
+:py:meth:`DataArray.values`, where the former is strongly preferred. The difference is in the way they coerce to numpy - :py:meth:`~DataArray.values`
+always uses :py:func:`numpy.asarray` which will fail for some array types (e.g. ``cupy``), whereas :py:meth:`~DataArray.to_numpy`
+uses the correct method depending on the array type.
+
+.. ipython:: python
+
+    s_da.to_numpy()
+
+.. ipython:: python
+    :okexcept:
+
+    s_da.values
+
+This illustrates the difference between :py:meth:`~DataArray.data` and :py:meth:`~DataArray.values`,
+which is sometimes a point of confusion for new xarray users.
+Explicitly: :py:meth:`DataArray.data` returns the underlying numpy-like array, regardless of type, whereas
+:py:meth:`DataArray.values` converts the underlying array to a numpy array before returning it.
+(This is another reason to use :py:meth:`~DataArray.to_numpy` over :py:meth:`~DataArray.values` - the intention is clearer.)
+
+Conversion to numpy as a fallback
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+If a wrapped array does not implement the corresponding array method then xarray will often attempt to convert the
+underlying array to a numpy array so that the operation can be performed. You may want to watch out for this behavior,
+and report any instances in which it causes problems.
+
+Most of xarray's API does support using :term:`duck array` objects, but there are a few areas where
+the code will still convert to ``numpy`` arrays:
+
+- Dimension coordinates, and thus all indexing operations:
 
   * :py:meth:`Dataset.sel` and :py:meth:`DataArray.sel`
   * :py:meth:`Dataset.loc` and :py:meth:`DataArray.loc`
@@ -33,7 +186,7 @@ the code will still cast to ``numpy`` arrays:
     :py:meth:`DataArray.reindex` and :py:meth:`DataArray.reindex_like`: duck arrays in
     data variables and non-dimension coordinates won't be casted
 
-- functions and methods that depend on external libraries or features of ``numpy`` not
+- Functions and methods that depend on external libraries or features of ``numpy`` not
   covered by ``__array_function__`` / ``__array_ufunc__``:
 
   * :py:meth:`Dataset.ffill` and :py:meth:`DataArray.ffill` (uses ``bottleneck``)
@@ -49,17 +202,25 @@ the code will still cast to ``numpy`` arrays:
     :py:class:`numpy.vectorize`)
   * :py:func:`apply_ufunc` with ``vectorize=True`` (uses :py:class:`numpy.vectorize`)
 
-- incompatibilities between different :term:`duck array` libraries:
+- Incompatibilities between different :term:`duck array` libraries:
 
   * :py:meth:`Dataset.chunk` and :py:meth:`DataArray.chunk`: this fails if the data was
     not already chunked and the :term:`duck array` (e.g. a ``pint`` quantity) should
-    wrap the new ``dask`` array; changing the chunk sizes works.
-
+    wrap the new ``dask`` array; changing the chunk sizes works however.
 
 Extensions using duck arrays
 ----------------------------
-Here's a list of libraries extending ``xarray`` to make working with wrapped duck arrays
-easier:
+
+Whilst the features above allow many numpy-like array libraries to be used pretty seamlessly with xarray, it often also
+makes sense to use an interfacing package to make certain tasks easier.
+
+For example the `pint-xarray package <https://pint-xarray.readthedocs.io>`_ offers a custom ``.pint`` accessor (see :ref:`internals.accessors`) which provides
+convenient access to information stored within the wrapped array (e.g. ``.units`` and ``.magnitude``), and makes makes
+creating wrapped pint arrays (and especially xarray-wrapping-pint-wrapping-dask arrays) simpler for the user.
+
+We maintain a list of libraries extending ``xarray`` to make working with particular wrapped duck arrays
+easier. If you know of more that aren't on this list please raise an issue to add them!
 
 - `pint-xarray <https://pint-xarray.readthedocs.io>`_
 - `cupy-xarray <https://cupy-xarray.readthedocs.io>`_
+- `cubed-xarray <https://github.com/cubed-xarray>`_

doc/whats-new.rst

@@ -38,6 +38,8 @@ Bug fixes
 Documentation
 ~~~~~~~~~~~~~
 
+- Expanded the page on wrapping numpy-like "duck" arrays.
+  (:pull:`7911`) By `Tom Nicholas <https://github.com/TomNicholas>`_.
 
 Internal Changes
 ~~~~~~~~~~~~~~~~
@@ -98,7 +100,6 @@ Bug fixes
 Documentation
 ~~~~~~~~~~~~~
 
-
 Internal Changes
 ~~~~~~~~~~~~~~~~