Guide

Installation

You can install simplezarr via pip (or most other Python package managers). Python 3.12 or higher is required.

pip install simplezarr

Opening a zarr file

# Create a store
store = simplezarr.LocalStore(filename)

# Open as zarr
group = simplezarr.open_zarr(store)

# Navigate to array
array = group["foo/array"]

Get some info

group.print_metadata()
group.print_structure()
group.attributes
group.children

array.print_metadata()
array.dtype
array.nbytes
array.ndim
array.shape
array.chunk_shape
array.chunk_grid_shape
# etc.

Read some data

# Get a numpy array (blocking)
a = array[:200, :200].get_now()

# Load two regions in parallel
f1 = array[:200, :200].get_soon()
f2 = array[1000:1200, :200].get_soon()
a1, a2 = [f1.result(), f2.result()]

Write some data

# Put random data in a region (blocking)
a = np.random.uniform(size=(200, 200))
array[:200, :200].set_now(a)

# Zero out two regions in parallel
f1 = array[:200, :200].set_soon(0)
f2 = array[1000:1200, :200].set_soon(0)

# You want to wait so that any errors during writing are raised
[f1.result(), f2.result()]

Read and write whole chunks

# Load first chunk as numpy array
a = array.chunks[0, 0].get_now()

# Load a series as chunk as a single array (in parallel, blocking)
a = array.chunks[0, :].get_now()

# Writing and `_soon` works the same

Read and write individual chunks

# Lower-level function to read individual chunks
a = array.get_chunk_now((0, 0))

# Read parallel
f1 = array.get_chunk_soon((0, 0))
f2 = array.get_chunk_soon((10, 8))
a1, a2 = [f1.result(), f2.result()]

# Writing
array.set_chunk_now((0, 1), a)

# Write in parallel
f1 = array.set_chunk_soon((0, 2), a)
f2 = array.set_chunk_soon((0, 3), a)
[f1.result(), f2.result()]

More functionality

That’s it really … any other functionality is built on top of this, and provided via dedicated modules in simplezarr.utils.