# dask.array.blockwise

### dask.array.blockwise(func, out_ind, \*args, name=None, token=None, dtype=None, adjust_chunks=None, new_axes=None, align_arrays=True, concatenate=None, meta=None, \*\*kwargs)

Tensor operation: Generalized inner and outer products

A broad class of blocked algorithms and patterns can be specified with a
concise multi-index notation.  The `blockwise` function applies an in-memory
function across multiple blocks of multiple inputs in a variety of ways.
Many dask.array operations are special cases of blockwise including
elementwise, broadcasting, reductions, tensordot, and transpose.

* **Parameters:**
  **func**
  : Function to apply to individual tuples of blocks

  **out_ind**
  : Block pattern of the output, something like ‘ijk’ or (1, 2, 3)

  **\*args**
  : You may also pass literal arguments, accompanied by None index
    e.g. (x, ‘ij’, y, ‘jk’, z, ‘i’, some_literal, None)

  **\*\*kwargs**
  : Extra keyword arguments to pass to function

  **dtype**
  : Datatype of resulting array.

  **concatenate**
  : If true concatenate arrays along dummy indices, else provide lists

  **adjust_chunks**
  : Dictionary mapping index to function to be applied to chunk sizes

  **new_axes**
  : New indexes and their dimension lengths

  **align_arrays: bool**
  : Whether or not to align chunks along equally sized dimensions when
    multiple arrays are provided.  This allows for larger chunks in some
    arrays to be broken into smaller ones that match chunk sizes in other
    arrays such that they are compatible for block function mapping. If
    this is false, then an error will be thrown if arrays do not already
    have the same number of blocks in each dimension.

### Examples

2D embarrassingly parallel operation from two arrays, x, and y.

```pycon
>>> import operator, numpy as np, dask.array as da
>>> x = da.from_array([[1, 2],
...                    [3, 4]], chunks=(1, 2))
>>> y = da.from_array([[10, 20],
...                    [0, 0]])
>>> z = blockwise(operator.add, 'ij', x, 'ij', y, 'ij', dtype='f8')
>>> z.compute()
array([[11, 22],
       [ 3,  4]])
```

Outer product multiplying a by b, two 1-d vectors

```pycon
>>> a = da.from_array([0, 1, 2], chunks=1)
>>> b = da.from_array([10, 50, 100], chunks=1)
>>> z = blockwise(np.outer, 'ij', a, 'i', b, 'j', dtype='f8')
>>> z.compute()
array([[  0,   0,   0],
       [ 10,  50, 100],
       [ 20, 100, 200]])
```

z = x.T

```pycon
>>> z = blockwise(np.transpose, 'ji', x, 'ij', dtype=x.dtype)
>>> z.compute()
array([[1, 3],
       [2, 4]])
```

The transpose case above is illustrative because it does transposition
both on each in-memory block by calling `np.transpose` and on the order
of the blocks themselves, by switching the order of the index `ij -> ji`.

We can compose these same patterns with more variables and more complex
in-memory functions

z = X + Y.T

```pycon
>>> z = blockwise(lambda x, y: x + y.T, 'ij', x, 'ij', y, 'ji', dtype='f8')
>>> z.compute()
array([[11,  2],
       [23,  4]])
```

Any index, like `i` missing from the output index is interpreted as a
contraction (note that this differs from Einstein convention; repeated
indices do not imply contraction.)  In the case of a contraction the passed
function should expect an iterable of blocks on any array that holds that
index.  To receive arrays concatenated along contracted dimensions instead
pass `concatenate=True`.

Inner product multiplying a by b, two 1-d vectors

```pycon
>>> def sequence_dot(a_blocks, b_blocks):
...     result = 0
...     for a, b in zip(a_blocks, b_blocks):
...         result += a.dot(b)
...     return result
```

```pycon
>>> z = blockwise(sequence_dot, '', a, 'i', b, 'i', dtype='f8')
>>> z.compute()
np.int64(250)
```

Add new single-chunk dimensions with the `new_axes=` keyword, including
the length of the new dimension.  New dimensions will always be in a single
chunk.

```pycon
>>> def f(a):
...     return a[:, None] * np.ones((1, 5))
```

```pycon
>>> z = blockwise(f, 'az', a, 'a', new_axes={'z': 5}, dtype=a.dtype)
```

New dimensions can also be multi-chunk by specifying a tuple of chunk
sizes.  This has limited utility as is (because the chunks are all the
same), but the resulting graph can be modified to achieve more useful
results (see `da.map_blocks`).

```pycon
>>> z = blockwise(f, 'az', a, 'a', new_axes={'z': (5, 5)}, dtype=x.dtype)
>>> z.chunks
((1, 1, 1), (5, 5))
```

If the applied function changes the size of each chunk you can specify this
with a `adjust_chunks={...}` dictionary holding a function for each index
that modifies the dimension size in that index.

```pycon
>>> def double(x):
...     return np.concatenate([x, x])
```

```pycon
>>> y = blockwise(double, 'ij', x, 'ij',
...               adjust_chunks={'i': lambda n: 2 * n}, dtype=x.dtype)
>>> y.chunks
((2, 2), (2,))
```

Include literals by indexing with None

```pycon
>>> z = blockwise(operator.add, 'ij', x, 'ij', 1234, None, dtype=x.dtype)
>>> z.compute()
array([[1235, 1236],
       [1237, 1238]])
```

<!-- !! processed by numpydoc !! -->
