wea.py

What is wea?

Giving the package a meaning - wea stands for Wrapped Exchange Array. If you want to share array-packed data with different processes, remote nodes or different language executables ( yes, that's the vision ), wea is aiming to be a lean, lightweight and convenient alternative to Protocol Buffers and Co.

It's inspired and adopted partly from Julia’s InterProcessCommunication WrappedArray.

The wrapped exchange array can be accessed like a numpy array because under the hood, numpy is applied.

If this sounds good to you, just give it a try.

Getting started

Install the package from Pypi

pip install wea

Quick API guide

Shared memory

import wea
import numpy as np

...
wa = wea.shared_memory.create_shared_array('/awesome-1', np.dtype('float64'), (10, 2))
wa[:] = my_new_data[:]
...

Creating a shared memory segment

In order to create a new shared memory segment, use the following snippet

import wea
import numpy as np

type = np.dtype('float64')
dims = (10, 2)
wa = wea.shared_memory.create_shared_array('/awesome-1', type, dims)
wa[:] = np.random.randn(dims[0], dims[1])

If creating was not possible because the segment already exists , a FileExistsError exception will be thrown.

Attaching to an existing shared memory array

If a wrapped exchange array was already created, you can attach to it simply by

import wea
import numpy as np

wa = wea.shared_memory.attach_shared_array('/awesome-1')
wa[:] = np.random.randn(dims[0], dims[1])

The metadata of the array are stored in the shared memory header segment and will be retrieved for the numpy array creation.

If attaching was not possible because the segment does not exist so far, a FileNotFoundError exception will be thrown.

Bytearray buffer memory

import wea
import numpy as np

...
wa = wea.buffered_memory.create_buffered_array(np.dtype('float64'), (10, 2))
wa[:] = my_new_data[:]
buf = wa.exchange_buffer
share(buf) # where share calls your prefered communication protocol
...

Creating a buffered memory segment

In order to create a new buffered memory segment, use the following snippet

import wea
import numpy as np

type = np.dtype('float64')
dims = (10, 2)
wa = wea.buffered_memory.create_buffered_array(type, dims)
wa[:] = np.random.randn(dims[0], dims[1])
buf: bytearray = wa.exchange_buffer

Actually it copies the content from the numpy array into the buffer. Thus, the current behavior is like a deep copy.

Loading from an existing buffered memory segment

If a wrapped exchange array was already created, you can load from it simply by

import wea
import numpy as np

buf: bytearray = receive() # where receive via your prefered communication protocol
wa = wea.buffered_memory.load_buffered_array(buf)

The metadata of the array are stored in the buffered memory header segment and will be retrieved for the numpy array creation.

Contributing

I welcome any contributions, enhancements, and bug-fixes. Open an issue on GitHub and submit a pull request.

License

wea.py is 100% free and open-source, under the MIT license. Use it however you want.

This package is Treeware. If you use it in production, then we ask that you buy the world a tree to thank us for our work. By contributing to the Treeware forest you’ll be creating employment for local families and restoring wildlife habitats.