>>> import zarr
>>> z = zarr.zeros((10000, 10000), chunks=(1000, 1000), dtype='i4')
<zarr.core.Array (10000, 10000) int32>
The code above creates a 2-dimensional array of 32-bit integers with 10000 rows
and 10000 columns, divided into chunks where each chunk has 1000 rows and 1000
columns (and so there will be 100 chunks in total).
For a complete list of array creation routines see the zarr.creation
module documentation.
Reading and writing data
Zarr arrays support a similar interface to NumPy arrays for reading and writing
data. For example, the entire array can be filled with a scalar value:
>>> z[:] = 42
Regions of the array can also be written to, e.g.:
>>> import numpy as np
>>> z[0, :] = np.arange(10000)
>>> z[:, 0] = np.arange(10000)
The contents of the array can be retrieved by slicing, which will load the
requested region into memory as a NumPy array, e.g.:
>>> z[0, 0]
>>> z[-1, -1]
>>> z[0, :]
array([ 0, 1, 2, ..., 9997, 9998, 9999], dtype=int32)
>>> z[:, 0]
array([ 0, 1, 2, ..., 9997, 9998, 9999], dtype=int32)
>>> z[:]
array([[ 0, 1, 2, ..., 9997, 9998, 9999],
[ 1, 42, 42, ..., 42, 42, 42],
[ 2, 42, 42, ..., 42, 42, 42],
[9997, 42, 42, ..., 42, 42, 42],
[9998, 42, 42, ..., 42, 42, 42],
[9999, 42, 42, ..., 42, 42, 42]], dtype=int32)
Persistent arrays
In the examples above, compressed data for each chunk of the array was stored in
main memory. Zarr arrays can also be stored on a file system, enabling
persistence of data between sessions. For example:
>>> z1 = zarr.open('data/example.zarr', mode='w', shape=(10000, 10000),
... chunks=(1000, 1000), dtype='i4')
The array above will store its configuration metadata and all compressed chunk
data in a directory called ‘data/example.zarr’ relative to the current working
directory. The zarr.convenience.open()
function provides a convenient way
to create a new persistent array or continue working with an existing
array. Note that although the function is called “open”, there is no need to
close an array: data are automatically flushed to disk, and files are
automatically closed whenever an array is modified.
Persistent arrays support the same interface for reading and writing data,
e.g.:
>>> z1[:] = 42
>>> z1[0, :] = np.arange(10000)
>>> z1[:, 0] = np.arange(10000)
Check that the data have been written and can be read again:
>>> z2 = zarr.open('data/example.zarr', mode='r')
>>> np.all(z1[:] == z2[:])
If you are just looking for a fast and convenient way to save NumPy arrays to
disk then load back into memory later, the functions
zarr.convenience.save()
and zarr.convenience.load()
may be
useful. E.g.:
>>> a = np.arange(10)
>>> zarr.save('data/example.zarr', a)
>>> zarr.load('data/example.zarr')
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
Please note that there are a number of other options for persistent array
storage, see the section on Storage alternatives below.
Resizing and appending
A Zarr array can be resized, which means that any of its dimensions can be
increased or decreased in length. For example:
>>> z = zarr.zeros(shape=(10000, 10000), chunks=(1000, 1000))
>>> z[:] = 42
>>> z.resize(20000, 10000)
>>> z.shape
(20000, 10000)
Note that when an array is resized, the underlying data are not rearranged in
any way. If one or more dimensions are shrunk, any chunks falling outside the
new array shape will be deleted from the underlying store.
For convenience, Zarr arrays also provide an append()
method, which can be
used to append data to any axis. E.g.:
>>> a = np.arange(10000000, dtype='i4').reshape(10000, 1000)
>>> z = zarr.array(a, chunks=(1000, 100))
>>> z.shape
(10000, 1000)
>>> z.append(a)
(20000, 1000)
>>> z.append(np.vstack([a, a]), axis=1)
(20000, 2000)
>>> z.shape
(20000, 2000)
Compressors
A number of different compressors can be used with Zarr. A separate package
called NumCodecs is available which provides a common interface to various
compressor libraries including Blosc, Zstandard, LZ4, Zlib, BZ2 and
LZMA. Different compressors can be provided via the compressor
keyword
argument accepted by all array creation functions. For example:
>>> from numcodecs import Blosc
>>> compressor = Blosc(cname='zstd', clevel=3, shuffle=Blosc.BITSHUFFLE)
>>> data = np.arange(100000000, dtype='i4').reshape(10000, 10000)
>>> z = zarr.array(data, chunks=(1000, 1000), compressor=compressor)
>>> z.compressor
Blosc(cname='zstd', clevel=3, shuffle=BITSHUFFLE, blocksize=0)
This array above will use Blosc as the primary compressor, using the Zstandard
algorithm (compression level 3) internally within Blosc, and with the
bit-shuffle filter applied.
When using a compressor, it can be useful to get some diagnostics on the
compression ratio. Zarr arrays provide a info
property which can be used to
print some diagnostics, e.g.:
>>> z.info
Type : zarr.core.Array
Data type : int32
Shape : (10000, 10000)
Chunk shape : (1000, 1000)
Order : C
Read-only : False
Compressor : Blosc(cname='zstd', clevel=3, shuffle=BITSHUFFLE,
: blocksize=0)
Store type : zarr.storage.KVStore
No. bytes : 400000000 (381.5M)
No. bytes stored : 3379344 (3.2M)
Storage ratio : 118.4
Chunks initialized : 100/100
If you don’t specify a compressor, by default Zarr uses the Blosc
compressor. Blosc is generally very fast and can be configured in a variety of
ways to improve the compression ratio for different types of data. Blosc is in
fact a “meta-compressor”, which means that it can use a number of different
compression algorithms internally to compress the data. Blosc also provides
highly optimized implementations of byte- and bit-shuffle filters, which can
improve compression ratios for some data. A list of the internal compression
libraries available within Blosc can be obtained via:
>>> from numcodecs import blosc
>>> blosc.list_compressors()
['blosclz', 'lz4', 'lz4hc', 'snappy', 'zlib', 'zstd']
In addition to Blosc, other compression libraries can also be used. For example,
here is an array using Zstandard compression, level 1:
>>> from numcodecs import Zstd
>>> z = zarr.array(np.arange(100000000, dtype='i4').reshape(10000, 10000),
... chunks=(1000, 1000), compressor=Zstd(level=1))
>>> z.compressor
Zstd(level=1)
Here is an example using LZMA with a custom filter pipeline including LZMA’s
built-in delta filter:
>>> import lzma
>>> lzma_filters = [dict(id=lzma.FILTER_DELTA, dist=4),
... dict(id=lzma.FILTER_LZMA2, preset=1)]
>>> from numcodecs import LZMA
>>> compressor = LZMA(filters=lzma_filters)
>>> z = zarr.array(np.arange(100000000, dtype='i4').reshape(10000, 10000),
... chunks=(1000, 1000), compressor=compressor)
>>> z.compressor
LZMA(format=1, check=-1, preset=None, filters=[{'dist': 4, 'id': 3}, {'id': 33, 'preset': 1}])
The default compressor can be changed by setting the value of the
zarr.storage.default_compressor
variable, e.g.:
>>> import zarr.storage
>>> from numcodecs import Zstd, Blosc
>>> # switch to using Zstandard
... zarr.storage.default_compressor = Zstd(level=1)
>>> z = zarr.zeros(100000000, chunks=1000000)
>>> z.compressor
Zstd(level=1)
>>> # switch back to Blosc defaults
... zarr.storage.default_compressor = Blosc()
To disable compression, set compressor=None
when creating an array, e.g.:
>>> z = zarr.zeros(100000000, chunks=1000000, compressor=None)
>>> z.compressor is None
Filters
In some cases, compression can be improved by transforming the data in some
way. For example, if nearby values tend to be correlated, then shuffling the
bytes within each numerical value or storing the difference between adjacent
values may increase compression ratio. Some compressors provide built-in filters
that apply transformations to the data prior to compression. For example, the
Blosc compressor has built-in implementations of byte- and bit-shuffle filters,
and the LZMA compressor has a built-in implementation of a delta
filter. However, to provide additional flexibility for implementing and using
filters in combination with different compressors, Zarr also provides a
mechanism for configuring filters outside of the primary compressor.
Here is an example using a delta filter with the Blosc compressor:
>>> from numcodecs import Blosc, Delta
>>> filters = [Delta(dtype='i4')]
>>> compressor = Blosc(cname='zstd', clevel=1, shuffle=Blosc.SHUFFLE)
>>> data = np.arange(100000000, dtype='i4').reshape(10000, 10000)
>>> z = zarr.array(data, chunks=(1000, 1000), filters=filters, compressor=compressor)
>>> z.info
Type : zarr.core.Array
Data type : int32
Shape : (10000, 10000)
Chunk shape : (1000, 1000)
Order : C
Read-only : False
Filter [0] : Delta(dtype='<i4')
Compressor : Blosc(cname='zstd', clevel=1, shuffle=SHUFFLE, blocksize=0)
Store type : zarr.storage.KVStore
No. bytes : 400000000 (381.5M)
No. bytes stored : 1290562 (1.2M)
Storage ratio : 309.9
Chunks initialized : 100/100
For more information about available filter codecs, see the Numcodecs documentation.
Groups
Zarr supports hierarchical organization of arrays via groups. As with arrays,
groups can be stored in memory, on disk, or via other storage systems that
support a similar interface.
To create a group, use the zarr.group()
function:
>>> root = zarr.group()
<zarr.hierarchy.Group '/'>
Groups have a similar API to the Group class from h5py. For example, groups can contain other groups:
>>> foo = root.create_group('foo')
>>> bar = foo.create_group('bar')
Groups can also contain arrays, e.g.:
>>> z1 = bar.zeros('baz', shape=(10000, 10000), chunks=(1000, 1000), dtype='i4')
<zarr.core.Array '/foo/bar/baz' (10000, 10000) int32>
Arrays are known as “datasets” in HDF5 terminology. For compatibility with h5py,
Zarr groups also implement the create_dataset()
and require_dataset()
methods, e.g.:
>>> z = bar.create_dataset('quux', shape=(10000, 10000), chunks=(1000, 1000), dtype='i4')
<zarr.core.Array '/foo/bar/quux' (10000, 10000) int32>
Members of a group can be accessed via the suffix notation, e.g.:
>>> root['foo']
<zarr.hierarchy.Group '/foo'>
The ‘/’ character can be used to access multiple levels of the hierarchy in one
call, e.g.:
>>> root['foo/bar']
<zarr.hierarchy.Group '/foo/bar'>
>>> root['foo/bar/baz']
<zarr.core.Array '/foo/bar/baz' (10000, 10000) int32>
The zarr.hierarchy.Group.tree()
method can be used to print a tree
representation of the hierarchy, e.g.:
>>> root.tree()
└── foo
└── bar
├── baz (10000, 10000) int32
└── quux (10000, 10000) int32
The zarr.convenience.open()
function provides a convenient way to create or
re-open a group stored in a directory on the file-system, with sub-groups stored in
sub-directories, e.g.:
>>> root = zarr.open('data/group.zarr', mode='w')
<zarr.hierarchy.Group '/'>
>>> z = root.zeros('foo/bar/baz', shape=(10000, 10000), chunks=(1000, 1000), dtype='i4')
<zarr.core.Array '/foo/bar/baz' (10000, 10000) int32>
Groups can be used as context managers (in a with
statement).
If the underlying store has a close
method, it will be called on exit.
For more information on groups see the zarr.hierarchy
and
zarr.convenience
API docs.
Array and group diagnostics
Diagnostic information about arrays and groups is available via the info
property. E.g.:
>>> root = zarr.group()
>>> foo = root.create_group('foo')
>>> bar = foo.zeros('bar', shape=1000000, chunks=100000, dtype='i8')
>>> bar[:] = 42
>>> baz = foo.zeros('baz', shape=(1000, 1000), chunks=(100, 100), dtype='f4')
>>> baz[:] = 4.2
>>> root.info
Name : /
Type : zarr.hierarchy.Group
Read-only : False
Store type : zarr.storage.MemoryStore
No. members : 1
No. arrays : 0
No. groups : 1
Groups : foo
>>> foo.info
Name : /foo
Type : zarr.hierarchy.Group
Read-only : False
Store type : zarr.storage.MemoryStore
No. members : 2
No. arrays : 2
No. groups : 0
Arrays : bar, baz
>>> bar.info
Name : /foo/bar
Type : zarr.core.Array
Data type : int64
Shape : (1000000,)
Chunk shape : (100000,)
Order : C
Read-only : False
Compressor : Blosc(cname='lz4', clevel=5, shuffle=SHUFFLE, blocksize=0)
Store type : zarr.storage.MemoryStore
No. bytes : 8000000 (7.6M)
No. bytes stored : 33240 (32.5K)
Storage ratio : 240.7
Chunks initialized : 10/10
>>> baz.info
Name : /foo/baz
Type : zarr.core.Array
Data type : float32
Shape : (1000, 1000)
Chunk shape : (100, 100)
Order : C
Read-only : False
Compressor : Blosc(cname='lz4', clevel=5, shuffle=SHUFFLE, blocksize=0)
Store type : zarr.storage.MemoryStore
No. bytes : 4000000 (3.8M)
No. bytes stored : 23943 (23.4K)
Storage ratio : 167.1
Chunks initialized : 100/100
Groups also have the zarr.hierarchy.Group.tree()
method, e.g.:
>>> root.tree()
└── foo
├── bar (1000000,) int64
└── baz (1000, 1000) float32
If you’re using Zarr within a Jupyter notebook (requires
ipytree), calling tree()
will generate an
interactive tree representation, see the repr_tree.ipynb notebook
for more examples.
User attributes
Zarr arrays and groups support custom key/value attributes, which can be useful for
storing application-specific metadata. For example:
>>> root = zarr.group()
>>> root.attrs['foo'] = 'bar'
>>> z = root.zeros('zzz', shape=(10000, 10000))
>>> z.attrs['baz'] = 42
>>> z.attrs['qux'] = [1, 4, 7, 12]
>>> sorted(root.attrs)
['foo']
>>> 'foo' in root.attrs
>>> root.attrs['foo']
'bar'
>>> sorted(z.attrs)
['baz', 'qux']
>>> z.attrs['baz']
>>> z.attrs['qux']
[1, 4, 7, 12]
Internally Zarr uses JSON to store array attributes, so attribute values must be
JSON serializable.
Advanced indexing
As of version 2.2, Zarr arrays support several methods for advanced or “fancy”
indexing, which enable a subset of data items to be extracted or updated in an
array without loading the entire array into memory.
Note that although this functionality is similar to some of the advanced
indexing capabilities available on NumPy arrays and on h5py datasets, the Zarr
API for advanced indexing is different from both NumPy and h5py, so please
read this section carefully. For a complete description of the indexing API,
see the documentation for the zarr.core.Array
class.
Indexing with coordinate arrays
Items from a Zarr array can be extracted by providing an integer array of
coordinates. E.g.:
>>> z = zarr.array(np.arange(10) ** 2)
>>> z[:]
array([ 0, 1, 4, 9, 16, 25, 36, 49, 64, 81])
>>> z.get_coordinate_selection([2, 5])
array([ 4, 25])
Coordinate arrays can also be used to update data, e.g.:
>>> z.set_coordinate_selection([2, 5], [-1, -2])
>>> z[:]
array([ 0, 1, -1, 9, 16, -2, 36, 49, 64, 81])
For multidimensional arrays, coordinates must be provided for each dimension,
e.g.:
>>> z = zarr.array(np.arange(15).reshape(3, 5))
>>> z[:]
array([[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14]])
>>> z.get_coordinate_selection(([0, 2], [1, 3]))
array([ 1, 13])
>>> z.set_coordinate_selection(([0, 2], [1, 3]), [-1, -2])
>>> z[:]
array([[ 0, -1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, -2, 14]])
For convenience, coordinate indexing is also available via the vindex
property, as well as the square bracket operator, e.g.:
>>> z.vindex[[0, 2], [1, 3]]
array([-1, -2])
>>> z.vindex[[0, 2], [1, 3]] = [-3, -4]
>>> z[:]
array([[ 0, -3, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, -4, 14]])
>>> z[[0, 2], [1, 3]]
array([-3, -4])
When the indexing arrays have different shapes, they are broadcast together.
That is, the following two calls are equivalent:
>>> z[1, [1, 3]]
array([6, 8])
>>> z[[1, 1], [1, 3]]
array([6, 8])
Indexing with a mask array
Items can also be extracted by providing a Boolean mask. E.g.:
>>> z = zarr.array(np.arange(10) ** 2)
>>> z[:]
array([ 0, 1, 4, 9, 16, 25, 36, 49, 64, 81])
>>> sel = np.zeros_like(z, dtype=bool)
>>> sel[2] = True
>>> sel[5] = True
>>> z.get_mask_selection(sel)
array([ 4, 25])
>>> z.set_mask_selection(sel, [-1, -2])
>>> z[:]
array([ 0, 1, -1, 9, 16, -2, 36, 49, 64, 81])
Here’s a multidimensional example:
>>> z = zarr.array(np.arange(15).reshape(3, 5))
>>> z[:]
array([[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14]])
>>> sel = np.zeros_like(z, dtype=bool)
>>> sel[0, 1] = True
>>> sel[2, 3] = True
>>> z.get_mask_selection(sel)
array([ 1, 13])
>>> z.set_mask_selection(sel, [-1, -2])
>>> z[:]
array([[ 0, -1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, -2, 14]])
For convenience, mask indexing is also available via the vindex
property,
e.g.:
>>> z.vindex[sel]
array([-1, -2])
>>> z.vindex[sel] = [-3, -4]
>>> z[:]
array([[ 0, -3, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, -4, 14]])
Mask indexing is conceptually the same as coordinate indexing, and is
implemented internally via the same machinery. Both styles of indexing allow
selecting arbitrary items from an array, also known as point selection.
Orthogonal indexing
Zarr arrays also support methods for orthogonal indexing, which allows
selections to be made along each dimension of an array independently. For
example, this allows selecting a subset of rows and/or columns from a
2-dimensional array. E.g.:
>>> z = zarr.array(np.arange(15).reshape(3, 5))
>>> z[:]
array([[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14]])
>>> z.get_orthogonal_selection(([0, 2], slice(None))) # select first and third rows
array([[ 0, 1, 2, 3, 4],
[10, 11, 12, 13, 14]])
>>> z.get_orthogonal_selection((slice(None), [1, 3])) # select second and fourth columns
array([[ 1, 3],
[ 6, 8],
[11, 13]])
>>> z.get_orthogonal_selection(([0, 2], [1, 3])) # select rows [0, 2] and columns [1, 4]
array([[ 1, 3],
[11, 13]])
Data can also be modified, e.g.:
>>> z.set_orthogonal_selection(([0, 2], [1, 3]), [[-1, -2], [-3, -4]])
>>> z[:]
array([[ 0, -1, 2, -2, 4],
[ 5, 6, 7, 8, 9],
[10, -3, 12, -4, 14]])
For convenience, the orthogonal indexing functionality is also available via the
oindex
property, e.g.:
>>> z = zarr.array(np.arange(15).reshape(3, 5))
>>> z.oindex[[0, 2], :] # select first and third rows
array([[ 0, 1, 2, 3, 4],
[10, 11, 12, 13, 14]])
>>> z.oindex[:, [1, 3]] # select second and fourth columns
array([[ 1, 3],
[ 6, 8],
[11, 13]])
>>> z.oindex[[0, 2], [1, 3]] # select rows [0, 2] and columns [1, 4]
array([[ 1, 3],
[11, 13]])
>>> z.oindex[[0, 2], [1, 3]] = [[-1, -2], [-3, -4]]
>>> z[:]
array([[ 0, -1, 2, -2, 4],
[ 5, 6, 7, 8, 9],
[10, -3, 12, -4, 14]])
Any combination of integer, slice, 1D integer array and/or 1D Boolean array can
be used for orthogonal indexing.
If the index contains at most one iterable, and otherwise contains only slices and integers,
orthogonal indexing is also available directly on the array:
>>> z = zarr.array(np.arange(15).reshape(3, 5))
>>> all(z.oindex[[0, 2], :] == z[[0, 2], :])
Block Indexing
As of version 2.16.0, Zarr also support block indexing, which allows
selections of whole chunks based on their logical indices along each dimension
of an array. For example, this allows selecting a subset of chunk aligned rows and/or
columns from a 2-dimensional array. E.g.:
>>> import zarr
>>> import numpy as np
>>> z = zarr.array(np.arange(100).reshape(10, 10), chunks=(3, 3))
Retrieve items by specifying their block coordinates:
>>> z.get_block_selection(1)
array([[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[50, 51, 52, 53, 54, 55, 56, 57, 58, 59]])
Equivalent slicing:
>>> z[3:6]
array([[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[50, 51, 52, 53, 54, 55, 56, 57, 58, 59]])
For convenience, the block selection functionality is also available via the
blocks property, e.g.:
>>> z.blocks[1]
array([[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[50, 51, 52, 53, 54, 55, 56, 57, 58, 59]])
Block index arrays may be multidimensional to index multidimensional arrays.
For example:
>>> z.blocks[0, 1:3]
array([[ 3, 4, 5, 6, 7, 8],
[13, 14, 15, 16, 17, 18],
[23, 24, 25, 26, 27, 28]])
Data can also be modified. Let’s start by a simple 2D array:
>>> import zarr
>>> import numpy as np
>>> z = zarr.zeros((6, 6), dtype=int, chunks=2)
Set data for a selection of items:
>>> z.set_block_selection((1, 0), 1)
>>> z[...]
array([[0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0],
[1, 1, 0, 0, 0, 0],
[1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0]])
For convenience, this functionality is also available via the blocks
property.
E.g.:
>>> z.blocks[:, 2] = 7
>>> z[...]
array([[0, 0, 0, 0, 7, 7],
[0, 0, 0, 0, 7, 7],
[1, 1, 0, 0, 7, 7],
[1, 1, 0, 0, 7, 7],
[0, 0, 0, 0, 7, 7],
[0, 0, 0, 0, 7, 7]])
Any combination of integer and slice can be used for block indexing:
>>> z.blocks[2, 1:3]
array([[0, 0, 7, 7],
[0, 0, 7, 7]])
Indexing fields in structured arrays
All selection methods support a fields
parameter which allows retrieving or
replacing data for a specific field in an array with a structured dtype. E.g.:
>>> a = np.array([(b'aaa', 1, 4.2),
... (b'bbb', 2, 8.4),
... (b'ccc', 3, 12.6)],
... dtype=[('foo', 'S3'), ('bar', 'i4'), ('baz', 'f8')])
>>> z = zarr.array(a)
>>> z['foo']
array([b'aaa', b'bbb', b'ccc'],
dtype='|S3')
>>> z['baz']
array([ 4.2, 8.4, 12.6])
>>> z.get_basic_selection(slice(0, 2), fields='bar')
array([1, 2], dtype=int32)
>>> z.get_coordinate_selection([0, 2], fields=['foo', 'baz'])
array([(b'aaa', 4.2), (b'ccc', 12.6)],
dtype=[('foo', 'S3'), ('baz', '<f8')])
Storage alternatives
Zarr can use any object that implements the MutableMapping
interface from
the collections
module in the Python standard library as the store for a
group or an array.
Some pre-defined storage classes are provided in the zarr.storage
module. For example, the zarr.storage.DirectoryStore
class provides a
MutableMapping
interface to a directory on the local file system. This is
used under the hood by the zarr.convenience.open()
function. In other words,
the following code:
>>> z = zarr.open('data/example.zarr', mode='w', shape=1000000, dtype='i4')
…is short-hand for:
>>> store = zarr.DirectoryStore('data/example.zarr')
>>> z = zarr.create(store=store, overwrite=True, shape=1000000, dtype='i4')
…and the following code:
>>> root = zarr.open('data/example.zarr', mode='w')
…is short-hand for:
>>> store = zarr.DirectoryStore('data/example.zarr')
>>> root = zarr.group(store=store, overwrite=True)
Any other compatible storage class could be used in place of
zarr.storage.DirectoryStore
in the code examples above. For example,
here is an array stored directly into a ZIP archive, via the
zarr.storage.ZipStore
class:
>>> store = zarr.ZipStore('data/example.zip', mode='w')
>>> root = zarr.group(store=store)
>>> z = root.zeros('foo/bar', shape=(1000, 1000), chunks=(100, 100), dtype='i4')
>>> z[:] = 42
>>> store.close()
Re-open and check that data have been written:
>>> store = zarr.ZipStore('data/example.zip', mode='r')
>>> root = zarr.group(store=store)
>>> z = root['foo/bar']
>>> z[:]
array([[42, 42, 42, ..., 42, 42, 42],
[42, 42, 42, ..., 42, 42, 42],
[42, 42, 42, ..., 42, 42, 42],
[42, 42, 42, ..., 42, 42, 42],
[42, 42, 42, ..., 42, 42, 42],
[42, 42, 42, ..., 42, 42, 42]], dtype=int32)
>>> store.close()
Note that there are some limitations on how ZIP archives can be used, because items
within a ZIP archive cannot be updated in place. This means that data in the array
should only be written once and write operations should be aligned with chunk
boundaries. Note also that the close()
method must be called after writing
any data to the store, otherwise essential records will not be written to the
underlying ZIP archive.
Another storage alternative is the zarr.storage.DBMStore
class, added
in Zarr version 2.2. This class allows any DBM-style database to be used for
storing an array or group. Here is an example using a Berkeley DB B-tree
database for storage (requires bsddb3 to be installed):
>>> import bsddb3
>>> store = zarr.DBMStore('data/example.bdb', open=bsddb3.btopen)
>>> root = zarr.group(store=store, overwrite=True)
>>> z = root.zeros('foo/bar', shape=(1000, 1000), chunks=(100, 100), dtype='i4')
>>> z[:] = 42
>>> store.close()
Also added in Zarr version 2.2 is the zarr.storage.LMDBStore
class which
enables the lightning memory-mapped database (LMDB) to be used for storing an array or
group (requires lmdb to be installed):
>>> store = zarr.LMDBStore('data/example.lmdb')
>>> root = zarr.group(store=store, overwrite=True)
>>> z = root.zeros('foo/bar', shape=(1000, 1000), chunks=(100, 100), dtype='i4')
>>> z[:] = 42
>>> store.close()
In Zarr version 2.3 is the