Executes python scripts in different environments without writing temp files to disk, it can pass and receive any type of variable, multiprocessing can be executed from everywhere, not only at top-level
Project description
Executes python scripts in different environments without writing temp files to disk, it can pass and receive any variable, multiprocessing can be executed from everywhere, not only at top-level
pip install callpyfile
# Create a file called callc1.py
import sys
from callpyfile import run_py_file # use this function to call the pyfile
from a_cv_imwrite_imread_plus import open_image_in_cv
# an example with OpenCV
pics = [
"https://github.com/hansalemaos/screenshots/raw/main/gimages/elephant.png",
"https://github.com/hansalemaos/screenshots/raw/main/gimages/dl.png",
"https://github.com/hansalemaos/screenshots/raw/main/gimages/house.png",
]
allpics = []
allpics_ = [open_image_in_cv(x) for x in pics]
for p in range(100):
allpics.extend(allpics_)
# Execute the py file
resu = run_py_file(
variables={
"imagelist": allpics
}, # The keys will be the global variables in the called pyfile
pyfile=r"C:\Users\Gamer\anaconda3\envs\dfdir\callc2.py", # The pyfile
activate_env_command="", # if you want to activate another environment, you can pass something like this (Anaconda): activate_env_command = ['activate.bat', 'otherenv']
pythonexe=sys.executable, # python.exe from your active environment
raise_exception=False, # if raise_exception is False, the function returns None if an Exception occurs during the Execution
)
# Create another file called callc2.py
import time
import cv2
from pathos.multiprocessing import ProcessingPool as Pool # By the way: good stuff
import numpy as np
from callpyfile import to_stdout # import the decorator
# Some CV stuff
def do_cv2_stuff(img):
kernel = np.ones((5, 5), np.float32) / 25
dst = cv2.filter2D(img, -1, kernel)
return dst
@to_stdout(kill_if_exception=True) # optional, kills the process using os._exit()
def exemu():
pool = Pool(nodes=5)
res = pool.amap(
do_cv2_stuff, imagelist
) # imagelist is not defined yet, but it doesn't matter because the decorator will create global variables from the dict you'have passed: {"imagelist": allpics},
while not res.ready():
time.sleep(0.001)
allb = res.get()
return allb
if __name__ == "__main__":
exemu()
# Run callc1.py
print(allpics[0][:1])
[[[255 255 255]
[255 255 255]
[255 255 255]
...
[ 54 53 50]
[ 57 55 52]
[ 59 57 54]]]
print(resu[0][:1])
[[[255 255 255]
[255 255 255]
[255 255 255]
...
[ 59 58 55]
[ 56 54 51]
[ 56 55 52]]]
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
callpyfile-0.10.tar.gz
(4.9 kB
view hashes)
Built Distribution
Close
Hashes for callpyfile-0.10-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8d610b9f8fc47f2651d2d3c2f015ee9b25c7139a5b98eaa9c3555bcae7e62e95 |
|
MD5 | 4115c63c8351b0fefcd54b698b9c30b1 |
|
BLAKE2b-256 | 776b4ba65179e6cbb9018278a7b0df7b11c365b2845b5bafa32b795c58d05748 |