Using 0/1 for boolean/integer parameter produces error

[hdoupe]

I came across this issue while trying to get to 100% test coverage in PR #1952. If you have a parameter that can be either boolean or an integer that is 0 or 1, you will get an error when the parameter is set to 0 or 1.

The code below is using the master branch of Tax-Calculator.

import taxcalc

dd = taxcalc.Policy.default_data(start_year=2017, metadata=True)
print('is boolean value', dd['_ALD_InvInc_ec_base_RyanBrady']['boolean_value'])
print('is integer value', dd['_ALD_InvInc_ec_base_RyanBrady']['integer_value'])

reform = {2017: {'_ALD_InvInc_ec_base_RyanBrady': [1]}}
print('reform', reform)

pol = taxcalc.Policy()
pol.ignore_reform_errors()
pol.implement_reform(reform)

which gives output:

is boolean value True
is integer value True
reform {2017: {'_ALD_InvInc_ec_base_RyanBrady': [1]}}
Traceback (most recent call last):
  File "int_bool_test.py", line 12, in <module>
    pol.implement_reform(reform)
  File "/Users/henrydoupe/Documents/Tax-Calculator/taxcalc/policy.py", line 232, in implement_reform
    raise ValueError('\n' + self.reform_errors)
ValueError: 
ERROR: 2017 _ALD_InvInc_ec_base_RyanBrady value 1 is not boolean

One possible fix for this is:

# Policy._validate_parameter_names_types line 490
pval_is_bool = type(pval) == bool or pval in (0, 1)

instead of:

# Policy._validate_parameter_names_types line 490
pval_is_bool = type(pval) == bool

[martinholmer]

@hdoupe, The _ALD_InvInc_ec_base_RyanBrady parameter is boolean, so it must have a value of True or False. It cannot have a numerical value following the merge of pull request #1960. So the "fix" is to specify the reform not this way:

reform = {2017: {'_ALD_InvInc_ec_base_RyanBrady': [1]}}

but this way:

reform = {2017: {'_ALD_InvInc_ec_base_RyanBrady': [True]}}

[hdoupe]

Ok, I see. So, for all boolean parameters, the JSON file should have "integer_type" be false and "boolean_type" be true?

[martinholmer]

@hdoupe said:

Ok, I see. So, for all boolean parameters, the JSON file should have "integer_type" be false and "boolean_type" be true?

Yes and I'm now preparing a pull request that makes those changes to the JSON parameter files and to the source code to accommodate the new and less confusing data type information.

[hdoupe]

Ok, thanks @martinholmer.

[martinholmer]

Pull request #1970 clarifies the documentation of parameter datatypes, the need for which was indicated by @hdoupe's reading of the old documentation in issue #1969.