Easy-to-use Python library for inter-process communications.
Using pip:
pip install easyipc --user
From source:
git clone https://github.com/luiscarlosgph/easyipc.git
cd easyipc
python setup.py build
python setup.py install --user
Server:
import numpy as np
import easyipc
server = easyipc.Pipe('haha')
server.listen() # Blocks until a client is connected
whatever_object = {'Hello': 'This is an example'}
# Send an object
server.send_whatever(whatever_object) # Blocks until all data is sent
# Send a numpy.ndarray: you can send numpy arrays with send_whatever(), but this is faster
arr = np.ones((32, 3, 1080, 1920), dtype=np.float32)
server.send_array(arr) # Block until all data is sent
Client:
import numpy as np
import easyipc
client = easyipc.Pipe('haha')
client.connect() # Blocks until it can connect to a server
# Receive an object
whatever_object = client.recv_whatever() # Blocks until it receives an object
print(whatever_object)
# Receive a numpy.ndarray
arr = client.recv_array() # Blocks until it receives an array
Too see some more examples click here.
In the examples folder, two timing scripts can be found: pipe_timing_report.py and zmq_timing_report.py. They both performed the same operation, a round trip of a numpy array of shape (32, 3, 1700, 1700) whose size if approximately 1GB. The results obtained on an Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz machine running Ubuntu 16.04 are as follows:
| Script | Execution time |
|---|---|
| EasyIPC (pipe_timing_report.py) | 1.81s |
| ZeroMQ (zmq_timing_report.py) | 2.07s |
To run the tests execute:
python setup.py test
This code uses FIFOs (particularly os.mkfifo), which are supported by Unix-like systems only. This code has been tested in Ubuntu 16.04 and Ubuntu 18.04.