Dynamically load json type in python

In python, the json module is included in the package, but the yaml module is made by a third party. In the program I am currently writing, to absorb the difference in the environment

If you have a yaml module and you have a yaml config file, read it.
If it doesn't work, try reading with json.
If that doesn't work, give an exception

The behavior is adopted. Since yaml and json have similar interfaces, I treated them exactly the same as below,

try:
    import yaml
    with open("config.yaml") as y:
        config = yaml.load("config.yaml")
except ImportError:
    import json
    try:
        with open("config.json") as j:
            config = json.load(j)
    except AttributeError:
        logger.error("please set config file in json or yaml !!")
        raise


#Use the value of config below ...

There was a pitfall here. In the case of yaml, the key type is automatically guessed and read, but json is always str. In other words, if the config file has the following contents

`config.yaml`


1: "hoge"
key: "fuga"

`config.json`


{
    "1": "hoge",
    "key": "fuga"
}

In either case, you can access " fuga " withconfig ["key"], but"hoge"isconfig ["1"]for json andconfig [" for yaml. Must be accessed with 1].

To avoid this, give a function as a hook at json.load ().

def jsonkeys_to_int_always(x): #If you can set all key types to int, click here.
    if isinstance(x, dict):
            return {int(k):v for k,v in x.items()}
    return x

def jsonkey_to_int_when_possible(x): #Click here if you want to behave in the same way as yaml
    d = {}
    if isinstance(x, dict):
        for k,v in x.items():
            try:
                k = int(k)
            except ValueError:
                pass
            d[k] = v
        return d

config = json.load("config.json", object_hook=jsonkeys_to_int_when_possible)

It seems that you can absorb the difference by Specify the value in yaml, but I don't know the details.