Configure default value when key doesn't exist? about addict

[mewwts]

Hi Chris!
Thanks for your input. You could call prune() and/or to_dict() before shipping it off to a module that doesn't handle the addict instances well.

Now, here's the problem we're facing with returning something but a Dict. In addict the key feature, for me at least, is the fact that you can create nested dictionaries using attributes.

>>> a = Dict()
>>> a.b.c.d.e = 2
>>> a
{'a': {'b': {'c': {'d': {'e': 2}}}}} 

This would not be possible (with the current implementation at least) if we were returning something else than an addict-instance, since when a unknown key is specified, we do not know if the intent was to set a value or getting one!

Does this sound OK to you?

[sabhiram]

We can always expose another member function. Perhaps something to the effect of:

def get(self, key, default_value = None):
    ...

@carmstrong - How do you envision the specifying of said default value? Would it be dict wide? or unique to each key?

Perhaps when the dict is created we can tell it how to initialize default values for various types.

[carmstrong]

Perhaps when the dict is created we can tell it how to initialize default values for various types.

Exactly! A dict-wide default would work just swell.

Your proposal is exactly what I'm envisioning. I really want to use Dict(), but I need default values to be sane Python. Using to_dict() isn't really possible for me because going this route in the first place was to avoid getting a KeyError when a key didn't exist.

[sabhiram]

Hmm, I thought about this a bit more today. Here is another issue, as it stands today: we can legally do this with a Dict:

prop = Dict(a=2, b={'a': 2}, c=[{'a':2}])

So if we would try to set "str" to "DEFAULT" like so:

prop = Dict(a=2, b={'a': 2}, c=[{'a':2}], str="DEFAULT")

The code would just set a key called "str" to "DEFAULT" :)

[mewwts]

Well for what it's whort @sabhiram, you can use get today, as it belongs to the dict. So to get a default-value back, like you would with a normal dict-instance, you could simply do

>>>> my_dict.get('my_nonexistent_key', None)
None

Also, it is trivial to to chance the constructor signature to something like

def __init__(*args, default=None, **kwargs):

(The syntax might be off. I know it's different in 2.7 and 3.4)

@carmstrong, would you mind writing out an example for how you use this with mongoengine and when they fail using it? It would be useful to see.

[carmstrong]

would you mind writing out an example for how you use this with mongoengine and when they fail using it? It would be useful to see.

Sure. So my issue is that I'm using xmltodict to convert a large XML payload to a dict. I then create a new dict with addict: Dict(xml). This works great, because I'm accessing deep member properties of the dict when I'm not sure their parent (or grandparent) even exists. For example:

my_dict.Result.Pet.Cat.Name

works great using addict, because if any of those sub-Dicts doesn't exist, I just get a {} back. With a regular dict, I'd get a KeyError which I don't want to have to try/except on every single call to the dict (there are dozens in this file). However, when using this with mongoengine, it doesn't know how to handle an empty Dict. Here's an example of how I'm using everything together:

    parsed = xmltodict.parse(xml)
    res = Dict(parsed)
    result = MyModel()

    errors = []
    for error in res.Results.Error:
      errors.append(ProcessingError(
        error_id = error['@id'],
        code = error.ErrorCode,
        message = error.ErrorMessage
      ))
    result.processing_errors = errors

In this example, ProcessingError is a mongoengine model:

class ProcessingError(EmbeddedDocument):
  error_id = IntField()
  code = StringField()
  message = StringField()

I'm getting an error when mongoengine tries to cast the error_id to an int, because in this case it's not provided, and is {}. mongoengine has no idea what a Dict is.

I hope that makes sense! Very much appreciate the replies.

[mewwts]

I'm away this weekend, so my responses might be few and short. Hope it wont hinder your progress!
Could this problem, for now, be bypassed by doing

errors = []
for error in res.Results.Error:
    errors.append(ProcessingError(
        error_id = error.get('@id', None),
        code = error.get('ErrorCode', None),
        message = error.get('ErrorMessage', None)
    ))

or am I wrong? Have a great weekend, and thanks for using, and contributing to addict!

[carmstrong]

Could this problem, for now, be bypassed

You're right! I couldn't do that before I started using addict because some sub-dict may not exist and throw a KeyError, but now that I'm using a Dict I can just use .get on the final dict's key. Thanks for the idea!

[mewwts]

Great, @carmstrong!
Do you find that this is a good enough solution so that we may consider to close this issue?

[carmstrong]

Do you find that this is a good enough solution so that we may consider to close this issue?

I still think it'd be nice to have a default method so I could just set it once on the parent Dict creation rather than the 100 or so times I'm accessing a sub-Dict, but I suppose that's up to you guys.

[mewwts]

Alright, @carmstrong, your usecase is a bit different from my typical one, and I see where you're coming from. I'll leave this open. I think it might be a candidate feature for future releases.

[perpetual-hydrofoil]

It'd be interesting to allow the Dict() constructor to accept a param that would allow any arbitrary class (or perhaps a function that'd return some value or object) to serve as a default for unset values.

[mewwts]

Closing this as there has been no activity for a long time.