Python: Unable to loop if key.value() is Null

I am working on a use case to remove the “All Users” group from a folder, and I am using all_content_metadata_accesses.

In Looker, if a folder doesn’t have any groups as its editor or viewer, then group_id is null (json), so I expect to loop and write some conditions with group_id, similar to this:

testing_loop = [{'a':1, 'group_id': None }, {'a':1, 'group_id':1}]  
for i in testing_loop:
  if i['group_id'] is not None:
    print('group_id is not Null')
  else: 
    print('group_id is Null')

However, I am unable to loop or do anything with the key-value pair ‘group_id’ if the value is None because we have an exception handling a key with a None value:

accesses = sdk.all_content_metadata_accesses(content_metadata_id=767)
test_dict = accesses[0] # Purposely choose accesses[0] because I know group_id here is Null 
print(type(test_dict))
print(test_dict['id']) # OK
print(test_dict['group_id']) # KeyError: 'group_id'

__getitem__(self, key)
     63             ret = ret.value
     64         if ret is None:
---> 65             raise KeyError(key)
     66         return ret
     67 

KeyError: 'group_id'

I’ll post a reproduction case in our internal tool.

[joeldodge79]

TLDR; I believe the if ret is None: raise KeyError(key) logic is wrong and I’ll see about removing it.

also sorry I missed the notification

this is more complicated. This “dictionary” access is a facade in front of the actual model/instance interface the python sdk presents.

When we wrote the initial [de]serialization for the model/instance interface we hit a snag where we if someone wrote this and sent it:

user = User(first_name="Jill")

we would send null for all unspecified json properties:

{"first_name": "Jill", "last_name": null, ...}

In order to work around this we decided to remove all None properties from a model before serializing it so the resulting json would only have what you specified. In the rare case that you meant to send a json null, then you’d do:

user = User(first_name="Jill", last_name=EXPLICIT_NULL) # defined in models.py

But that’s all about creating input to be sent to the API. Here we’re really talking about a response object/dict (though keeping in mind that often the response obj/dict is re-used as input). I think we should NOT be raising a KeyError if the field is None.

This raises another question that is applicable to both the model/instance and dictionary interfaces: for a sparse response (e.g. fields parameter was used on the api call) how can we distinguish between a field missing from the json payload versus one that is present and the value is null?

We currently don’t distinguish when we populate a response model/instance and, because the dictionary implementation is just a proxy to the model/instance, then we should expose what it would do. It’s gonna look kind of odd though:

user = User()
user.first_name  # None
user['first_name']  # None

[josephaxisa]

@lanlooker sorry for the delay on this, I missed the notification. The behavior you’re experiencing is not unique to the SDK unfortunately; accessing a dictionary item that does not exist throws a KeyError in Python. However, I can think of a couple of ways to get around this. Consider the below dictionary:

x = { 'foo': 'foo_val' }

1. Use the in operator to check if the key exists before accessing it

if 'bar' in x:
  print(x['bar'])

1. Use dict.get(key[, default]) (docs)

print(x.get('foo')) # foo_val
print(x.get('bar', None)) # None
print(x.get('bar', 'N/A')) # N/A
print(x.get('bar')) # None

Hope this helps!