Validators hold state if you vlad multiple times in the same context about vladiate

[dmcclory]

@di, what's happening is, we're initializing a Vlad object inside a function, like so: (in a python notebook)

def run_validators():
    Vlad(source=LocalFile(_output_path('plan_benefits.csv')), validators=FunValidator.validators).validate()

if there's a problem, the output looks correct the first time we call run_validators:

ERROR:vlad_logger:  SetValidator failed 4 time(s) on field: 'Subject to Cool Thing?'
ERROR:vlad_logger:    Invalid fields: ['YES', 'N/A']

if we run the same function again & again though, the error counts start increasing. (And the UniquenessValidators start to break)

ERROR:vlad_logger:  SetValidator failed 8 time(s) on field: 'Subject to Cool Thing?'
ERROR:vlad_logger:    Invalid fields: ['YES', 'N/A']
ERROR:vlad_logger:  UniqueValidator failed 40296 time(s) on field: 'Plan ID'
ERROR:vlad_logger:    Invalid fields: ['('fjsdlfjdsf', '30', '2')', '('dfklsdjfdf', '47', '2')', '('sdfjdskfjd', '28', '2')', '('sdkfjdfdk', '41', '2')'.....

[di]

I think doing validators=copy.deepcopy(CignaPlanBenefitInfoValidator.validators) instead will solve this issue.

[dmcclory]

Yeah, I'll definitely try that out. It requires a lot of knowledge about Vladiate though.

I think this would be nicer API:

PlanInfoValidator(source=LocalFile(_output_path('plan_info.csv'))).validate()

[csojinb]

Is there any reason why you'd want to pass validators into a new Vlad with state already attached? If not, we could flush them or something in the Vlad instatiation.

[di]

I decided this is a bug, because it is using the default validators argument (an empty dict) without checking to see if the class variable exists or not. Pushing a fix shortly.

[di]

Is there any reason why you'd want to pass validators into a new Vlad with state already attached?

I guess if you wanted to override them for some reason, you would expect to be able to do this.