Python str subclass with lazy evaluation of its value

I am building a command-line program that uses argparse.
In the (assumed-to-be) rare case of a wrong call, argparse will show a description string supplied when creating the ArgumentParser.

I want this description to show the version number of my program.
I want to extract this from the pyproject.toml file via tomllib.
Since this is an expensive operation (and even more so since I want to learn how to do it), I would like the description string to be evaluated lazily: only when it is actually to be printed.

I have not yet found a way to do it even though I am willing to build a one-trick-pony object specialized for this particular value:

• collections.UserString could provide the lazy evaluation (via overriding __getattribute__ for the data attribute), but, alas, some code in argparse uses re.sub() on it, which appears to check isinstance(x, str), which a UserString does not fulfill.
• a subclass of str can override any operation done on a string — but not perform lazy evaluation for a plain use of the entire string. (Is this true?)
• if ArgumentParser would use str(description) instead of description when it is about to print the description, one could supply an object that performs the lazy evaluation in its __str__ method. But, alas, ArgumentParser does not do this.

Is there any approach that does the job?

>Solution :

In a roundabout way: don’t set description, and only set it just before ArgumentParser would format the help.

import argparse
import time


class MyArgParser(argparse.ArgumentParser):
    def format_help(self):
        if not self.description:
            print("Now thinking hard...")
            time.sleep(3)
            self.description = "An application!"
        return super().format_help()


ap = MyArgParser()
ap.add_argument("--foo", help="foo")
args = ap.parse_args()
print(args)

This program is fast when run without arguments, or with --foo=bar, and takes its sweet time when you run with --help.