是否可以将字符串或列表以外的数据作为 python 中的子进程的参数传递？

Question

I can start a sub-process with a list as arguments for example like this:

data = []
data.append("test")
subprocess.run(['python', 'myprocess.py'] + data)

My question is, if it is possible to pass bigger amounts of data to the sub-process? For example, if I have loaded a bunch of images and want to pass them directly as numpy arrays without having to store them to file and pass the paths as strings to load them again in the sub-process?

Answer 1

Argument vectors are arrays of C strings at the operating system level (see man execve ). You can only pass data on the argv that can be represented as a C string (note that C strings are NUL-terminated, so no NULs allowed). You can certainly serialize data into a C string on one side and deserialize it on the other side, but that's work.

Moreover, there's a maximum amount of storage available that's shared between environment variables and command-line arguments -- so the more/larger environment variables you have, the shorter the combined length of all your command-line arguments can be.

Answer 2

No you cannot. The subprocess module is only a wrapper around the system call used to create a process and pass it a command line . In most system, that command line is just a sequence of space separated words, hence the list of strings.

If you want to pass anything else, you will have to serialize the data on one side, pass it from one process to the other one through pipes or shared memory (if available) or environment variables and deserialize it in the second process. pickle can be an easy tool to serialize simple data.

Alternatively, you could have a look at the multiprocess modules that does those things under the hood between Python processes