Python中的全局实例与本地实例

Question

I'm a self-taught programmer without much CS background. I read a lot of discussions here about the topic of global variables in Python, but I'm still unclear about when the use of global is 'good programming practice'.

Let's assume I write a Python script that performs a couple of long computations on each sample in a folder. I create a simple class that holds the results of each computation, so once the script is finished I can print a summary file.

class SampleInfo():

    def __init__(self):
        self.dict = {}

    def add_sample(self, sample):
        self.dict[sample] = {}

    def add_value(self, sample, value)
        self.dict[sample] = value

I see 2 ways to write the main script:

A) Using the global statement:

import os

def LongComputation(sample):
    # do some stuff that results in a value
    Info.add_value(sample, value)

def main():
    global Info
    Info = SampleInfo()
    for sample in os.listdir("."):
        LongComputation(sample)

main()

B) Without the global statement, passing the instance to the function:

import os

def LongComputation(sample, info_instance=None):
    # do some stuff that results in a value
    info_instance.add_value(sample, value)

def main():
    Info = SampleInfo()
    for sample in os.listdir("."):
        LongComputation(sample, info_instance=Info)

main()

Especially when I have several types of Info classes and many LongComputation functions, option B) becomes very messy. On the other hand, using the global statement is discouraged.

Should I use B) over A)? Why?

Answer 1

I don't see why you need to use a global here, you can just compute and return the value before adding it to the Info object. Also the class is overkill when a simple dictionary works just fine.

import os

def LongComputation(sample):
    # do some stuff that results in a value
    return value

def main():
    Info = {}
    for sample in os.listdir("."):
        Info[sample] = LongComputation(sample)

main()

Answer 2

When to use globals:

1. when it's a constant
2. when it's a singleton
3. when it's a private cache

And maybe, when your script is a quick, throwaway hack ;-)