Question regarding the Python import system and the importer (finder, loader) protocol:

In this example on Real Python, a custom importer for importing CSV files is implemented by specializing find_spec and exec_module. What puzzles me is the use of exec_module here, according to my understanding

• create_module is responsible for creating and returning a module object
• exec_module is responsible for executing the module’s code in its namespace.

Since the goal here is to hook into module creation and not module code execution, to me create_module would be the logical candidate for implementing this logic.

I tried to create_module for a toy JSON importer and it works just fine:

from importlib.machinery import ModuleSpec
import json
import pathlib
import sys


class JSONImporter:

    def __init__(self, json_path):
        self.json_path = json_path

    @classmethod
    def find_spec(cls, name, path, target=None):
        *_, module_name = name.rpartition(".")
        json_file_name = f"{module_name}.json"

        directories = sys.path if path is None else path

        for directory in directories:
            json_path = pathlib.Path(directory) / json_file_name

            if json_path.exists():
                return ModuleSpec(name, cls(json_path))

    def create_module(self, spec):
        try:
            with open(self.json_path) as f:
                json_object = json.load(f)
        except json.JSONDecodeError:
            return None
        else:
            return json_object

    def exec_module(self, module):
        pass


sys.meta_path.append(JSONImporter)

Obviously the implementation in the Real Python article also does the job, but it seems to me that their implementation using exec_module is a bit awkward. Am I missing something?