As a software system grows in complexity and usage, it can become burdensome if every change to the logic/behavior of the system also requires you to write and deploy new code. The goal of this business rules engine is to provide a simple interface allowing anyone to capture new rules and logic defining the behavior of a system, and a way to then process those rules on the backend.
You might, for example, find this is a useful way for analysts to define marketing logic around when certain customers or items are eligible for a discount or to automate emails after users enter a certain state or go through a particular sequence of events.
Variables represent values in your system, usually the value of some particular object. You create rules by setting threshold conditions such that when a variable is computed that triggers the condition some action is taken.
You define all the available variables for a certain kind of object in your code, and then later dynamically set the conditions and thresholds for those.
For example:
class ProductVariables(BaseVariables):
def __init__(self, product):
self.product = product
@numeric_rule_variable
def current_inventory(self):
return self.product.current_inventory
@numeric_rule_variable(label='Days until expiration')
def expiration_days(self)
last_order = self.product.orders[-1]
return (last_order.expiration_date - datetime.date.today()).days
@string_rule_variable()
def current_month(self):
return datetime.datetime.now().strftime("%B")
@select_rule_variable(options=Products.top_holiday_items())
def goes_well_with(self):
return products.related_productsThese are the actions that are available to be taken when a condition is triggered.
For example:
class ProductActions(BaseActions):
def __init__(self, product):
self.product = product
@rule_action(params={"sale_percentage": FIELD_NUMERIC})
def put_on_sale(self, sale_percentage):
self.product.price = (1.0 - sale_percentage) * self.product.price
self.product.save()
@rule_action(params={"number_to_order": FIELD_NUMERIC})
def order_more(self, number_to_order):
ProductOrder.objects.create(product_id=self.product.id,
quantity=number_to_order)If you need a select field for an action parameter, another -more verbose- syntax is available:
class ProductActions(BaseActions):
def __init__(self, product):
self.product = product
@rule_action(params=[{'fieldType': FIELD_SELECT,
'name': 'stock_state',
'label': 'Stock state',
'options': [
{'label': 'Available', 'name': 'available'},
{'label': 'Last items', 'name': 'last_items'},
{'label': 'Out of stock', 'name': 'out_of_stock'}
]}])
def change_stock_state(self, stock_state):
self.product.stock_state = stock_state
self.product.save()A rule is just a JSON object that gets interpreted by the business-rules engine.
Note that the JSON is expected to be auto-generated by a UI, which makes it simple for anyone to set and tweak business rules without knowing anything about the code. The javascript library used for generating these on the web can be found here.
An example of the resulting python lists/dicts is:
rules = [
# expiration_days < 5 AND current_inventory > 20
{
"conditions":{
"all":[
{
"name":"expiration_days",
"operator":"less_than",
"value":5
},
{
"name":"current_inventory",
"operator":"greater_than",
"value":20
}
]
},
"actions":[
{
"name":"put_on_sale",
"params":{
"sale_percentage":0.25
}
}
]
},
# NOT expiration_days < 5 AND current_inventory > 20
{
"conditions":{
"not":{
"all":[
{
"name":"expiration_days",
"operator":"less_than",
"value":5
},
{
"name":"current_inventory",
"operator":"greater_than",
"value":20
}
]
}
},
"actions":[
{
"name":"put_on_sale",
"params":{
"sale_percentage":0.25
}
}
]
},
# current_inventory < 5 OR (current_month = "December" AND current_inventory < 20)
{
"actions":[
{
"name":"order_more",
"params":{
"number_to_order":40
}
}
],
"conditions":{
"any":[
{
"name":"current_inventory",
"operator":"less_than",
"value":5
},
{
"all":[
{
"name":"current_month",
"operator":"equal_to",
"value":"December"
},
{
"name":"current_inventory",
"operator":"less_than",
"value":20
}
]
}
]
}
}
]To e.g. send to your client so it knows how to build rules
from business_rules import export_rule_data
export_rule_data(ProductVariables, ProductActions)that returns
{"variables": [
{ "name": "expiration_days",
"label": "Days until expiration",
"field_type": "numeric",
"options": []},
{ "name": "current_month",
"label": "Current Month",
"field_type": "string",
"options": []},
{ "name": "goes_well_with",
"label": "Goes Well With",
"field_type": "select",
"options": ["Eggnog", "Cookies", "Beef Jerkey"]}
],
"actions": [
{ "name": "put_on_sale",
"label": "Put On Sale",
"params": {"sale_percentage": "numeric"}},
{ "name": "order_more",
"label": "Order More",
"params": {"number_to_order": "numeric"}}
],
"variable_type_operators": {
"numeric": [ {"name": "equal_to",
"label": "Equal To",
"input_type": "numeric"},
{"name": "less_than",
"label": "Less Than",
"input_type": "numeric"},
{"name": "greater_than",
"label": "Greater Than",
"input_type": "numeric"}],
"string": [ { "name": "equal_to",
"label": "Equal To",
"input_type": "text"},
{ "name": "non_empty",
"label": "Non Empty",
"input_type": "none"}]
}
}from business_rules import run_all
rules = _some_function_to_receive_from_client()
for product in Products.objects.all():
run_all(rule_list=rules,
defined_variables=ProductVariables(product),
defined_actions=ProductActions(product),
stop_on_first_trigger=True
)The type represents the type of the value that will be returned for the variable and is necessary since there are different available comparison operators for different types, and the front-end that's generating the rules needs to know which operators are available.
All decorators can optionally take a label:
label- A human-readable label to show on the frontend. By default we just split the variable name on underscores and capitalize the words.
The available types and decorators are:
numeric - an integer, float, or python Decimal.
@numeric_rule_variable operators:
equal_togreater_thanless_thangreater_than_or_equal_toless_than_or_equal_to
Note: to compare floating point equality we just check that the difference is less than some small epsilon
string - a python bytestring or unicode string.
@string_rule_variable operators:
equal_tostarts_withends_withcontainsmatches_regexnon_empty
boolean - a True or False value.
@boolean_rule_variable operators:
is_trueis_false
select - a set of values, where the threshold will be a single item.
@select_rule_variable operators:
containsdoes_not_contain
select_multiple - a set of values, where the threshold will be a set of items.
@select_multiple_rule_variable operators:
contains_allis_contained_byshares_at_least_one_element_withshares_exactly_one_element_withshares_no_elements_with
dataframe - A pandas dataframe
@dataframe_rule_variable operators:
containsdoes_not_containcontains_case_insensitivedoes_not_contain_case_insensitiveequal_tonot_equal_tostarts_withends_withmatches_regexnot_matches_regexgreater_thanless_thangreater_than_or_equal_toless_than_or_equal_tonon_emptynon_empty_within_except_last_rowemptyempty_within_except_last_rowcontains_allnot_contains_allexistsnot_existshas_equal_lengthhas_not_equal_lengthequal_to_case_insensitivenot_equal_to_case_insensitiveis_contained_byis_not_contained_byis_contained_by_case_insensitiveis_not_contained_by_case_insensitivelonger_thanlonger_than_or_equal_toshorter_thanshorter_than_or_equal_toinvalid_datedate_equal_todate_not_equal_todate_less_thandate_less_than_or_equal_todate_greater_thandate_greater_than_or_equal_tois_complete_dateis_incomplete_dateis_unique_setis_not_unique_setis_ordered_setis_not_ordered_setis_not_unique_relationshipis_unique_relationshipis_not_valid_relationshipis_valid_relationshipis_not_valid_referenceis_valid_referencenon_conformant_value_data_typeconformant_value_data_typenon_conformant_value_lengthconformant_value_lengthhas_next_corresponding_recorddoes_not_have_next_corresponding_recordpresent_on_multiple_rows_withinnot_present_on_multiple_rows_withinadditional_columns_emptyadditional_columns_not_emptyhas_different_valueshas_same_valuesreferences_correct_codelistdoes_not_reference_correct_codelistdoes_not_use_valid_codelist_termsuses_valid_codelist_termsis_ordered_byis_not_ordered_byis_target_sorted_byis_target_not_sorted_byvalue_has_multiple_referencesvalue_does_not_have_multiple_referencessuffix_equal_tosuffix_not_equal_toprefix_equal_toprefix_not_equal_to
Open up a pull request, making sure to add tests for any new functionality. To set up the dev environment (assuming you're using virtualenvwrapper):
$ mkvirtualenv business-rules
$ pip install -r dev-requirements.txt
$ nosetests