Python uses the PYTHONPATH
environment-variable to determine in which folders it should look for modules.
You can play around with it by modifying sys.path
, which works nicely for pure Python-Modules.
But when a module uses shared object files or static libraries, it looks for those in LD_LIBRARY_PATH
(on linux), but this can't be changed as easily and is platform dependent as far as I know.
The quick-fix for this problem is of course to set the environment-variable or invoke the script like LD_LIBRARY_PATH=. ./script.py
, but then you'll have to set it again for every new shell you open.
Also, the .so
files in my case will always be in the same directory as the .py
file, but may very well be moved to another absolute path, so I'd like to set them automatically every time I invoke the script.
How can I edit the path in which the Python interpreter looks for libraries platform-independently on runtime?
EDIT:
I already tried os.environ['LD_LIBRARY_PATH'] = os.getcwd()
, but to no avail.
Python, when gets the values of environment variables as in
os.environ[‘LD_LIBRARY_PATH’]
oros.environ[‘PATH’]
, it copies the values, into a dictionary, from it's parent process's environment, generally bash (bash process's environment get's carried to the child process, the python running instance).you can see this environment variable section with
env
command output from bash.you can also see/read this env data from
/proc/<pid>/environ
, by introducing an infinite loop(while 1: pass
) after modifying any environment variable.If you see/read this variable value/data from
/proc/<pid>/environ
after modifying it inside the python script, you would get to see that the real variable's data doesn't get modified, though the python script shows a modified dictionary key value, updated.What actually happens when you modify an env variable inside python script, as in
os.environ['LD_LIBRARY_PATH']='/<new_location>'
, is that it just updates the value in local dictionary, which is not mapped to process's env variable section. Hence it won't propagate all the way back to reflect in current process's environment, because ONLY a local dictionary was modified/updated/populated.Hence if we want to have the new environment variable to be reflected, we should overwrite the memory image of the process with new environment variable data, using
execv
.Example:
Limitation: Ideally, python should not allow such modification of
os.environ
variables. But because there is no constant dictionary data type, it allows modification of the data variable. There is absolutely no use of modifying the values, as it does nothing useful to reflect in running process's real environment, unlessexecv
is used.I would use:
This sets the
LD_LIBRARY_PATH
environment variable for the duration/lifetime of the execution of the current process only.EDIT: it looks like this needs to be set before starting Python: Changing LD_LIBRARY_PATH at runtime for ctypes
So I'd suggest going with a wrapper
.sh
(or.py
if you insist) script. Also, as @chepner pointed out, you might want to consider installing your.so
files in a standard location (within the virtualenv).See also Setting LD_LIBRARY_PATH from inside Python
My solution to this problem is to put this as the first line of a Python script (instead of the usual shebang):
And here is how this works: