NotImplementedError when accessing validator errors while using oneof list of dict

[kardaj]

Use-case abstract

When I run the following snippet, using the 1.1 version, I get a NotImplementedError

from cerberus import Validator
schema = {
    'x': {
        "type": "dict",
        "oneof": [{
            "schema": {
                'v': {'type': 'string'}
            }}, {
            "schema": {
                'v': {'type': 'list', 'schema': {'type': 'dict', 'schema': {
                    'a': {'type': 'string', 'required': True}
                }}},
            }}]
    }}

v = Validator(schema)
v.validate({'x': {'v': ['a']}})
v.errors

In this example, the problem is that a is supposed to be a dict. Instead, a string was found.

Support request

After looking into the code, I found out it's due to following section in the insert_logic_error method of the BasicErrorHandler.

for child_error in child_errors:
    if child_error.is_logic_error:
        raise NotImplementedError
    elif child_error.is_group_error:
        raise NotImplementedError

Replacing it with the next snippet fixes my problem without breaking any of the existing tests of the project.

for child_error in child_errors:
    if child_error.is_logic_error:
        self.insert_logic_error(child_error)
    elif child_error.is_group_error:
        self.insert_group_error(child_error)

But I guess this behavior is not an accident and there's a reason for the NotImplementedError being there. Am I missing something?

[funkyfuture]

see #278

[boramalper]

Same issue here! Can be reproduced using:

import cerberus

generic_message_schema = {
    b"t": {"type": "binary", "empty": False, "required": True},
}

common_response_schema = {
    b"y": {"type": "binary", "allowed": [b"r"], "required": True}
}

common_response_arguments_schema = {
    b"id": {"type": "binary", "minlength": 20, "maxlength": 20, "required": True}
}

get_peers_response_schema = {
    **generic_message_schema,
    **common_response_schema,
    b"r": {"required": True, "schema": {
        **common_response_arguments_schema,
        b"token": {"type": "binary", "empty": False, "required": True},
        b"values": {"type": "list", "excludes": b"nodes", "required": True, "schema": {
            "type": "bytes",
            "minlength": 6, "maxlength": 6
        }},
        b"nodes": {"type": "binary", "excludes": b"values", "required": True, "minlength": 26,
                    "maxlength": 8 * 26}
    }}
}

print(">>>", cerberus.Validator().validate({}, get_peers_response_schema))

$ pip3 show cerberus
Name: Cerberus
Version: 1.1
Summary: Lightweight, extensible schema and data validation tool for Python dictionaries.
Home-page: http://github.com/nicolaiarocci/cerberus
Author: Nicola Iarocci
Author-email: nicola@nicolaiarocci.com
License: ISC
Location: /home/bora/.pyenv/versions/3.6.0/lib/python3.6/site-packages
Requires: 

[drcraig]

Any chance of the fix for this being released soon? I'm hitting exactly this error in what is really a perfect use case for a one-of rule, and am struggling to figure out a way to refactor my schema to avoid it.