RuleEngine

An engine which allows to define rules on product dimensions such as category/region/properties to calculate discounts without requiring a code change.

Technologies used

1. Scala
2. Spray (For REST APIs)
3. MongoDB(Casbah)
4. Redis

Setup and Examples

Head over to the link for setup and sample examples
https://github.com/sudan/RuleEngine/wiki

Overview

Basic idea behind building this rule engine is to create a minimal module to define rules for ecommerce products over dimensions such as region, category and product specific properties which are used as a criteria to calculate discounts.

Hierarchy

Sample Rule Schema

{
  "id": "<rule id>", // Generated for creation and can be used for updation,
  "name": "<rule name>",
  "description": "<rule description>",
  "version": "<rule version>",
  "owner": "<rule owner>", // a string field which denotes the owner who created this
  "discount": "<rule discount>", // a double value which denotes discount
  "boost": "<rule boost>", // an integer which is used to decide preference when two rules collide,
  "properties": {
    "property1": ["value11", "value12" ...],
    "property2": ["value21", "value22" ...]
  },
  "region_list": {
    "countries": [<list of countries>],
    "states": [<list of states>],
    "cities": [<list of cities>],
    "areas": [<list of areas>],
    "pincodes": [<list of pincodes>]
  },
  "category_list": {
    "main_categories": [<list of main categories>],
    "sub_categories": [<list of sub categories>],
    "verticals": [<list of verticals>],
    "product_ids": [<list of product ids>]
  }
}

Collection of rules form a campaign

Campaign Schema

{
  "id": "<campaign id>", // generated on creation and can be used for updation
  "rule_ids": [<list of rule ids>],
  "start_date": "<campaign start date>",
  "end_date": "<campaign end date>",
  "relationships": [
    <rule_id1> AND <rule_id2>
  ] // To be used across dimensions which combines rules instead of picking one
}

Collection of campaigns form a sale

{
  "id": "<sale id>", // generated on creation and can be used for updation
  "campaign_ids": [<list of campaign ids]
}

By default rules will be in dormant state unless the sale is explictly started.

Rules can be defined on

1. Region - Which includes countries, states, cities, areas and pincodes. Order of priority is bottom up where pincodes are given the most priority and countries are given the least to narrow down
2. Category - Which includes main categories, sub categories, verticals and product ids. Order of priority is bottom up where product ids are given the most priority and main categories are given the least to narrow down
3. List of properties where each is a key to list of values

Given a product with information on buyer’s region and product properties (only relevant ones) and category information, it fetches all rules which match region (based on priority mentioned above), which match category (based on priority mentioned above) and properties and apply the following conditions to fetch ruleIds

Region = r (rule ids returned for region based on order of priority)
Category = c (rule ids returned for category based on order of priority)
Property = p (rule ids returned for properties)
  rule_ids = r intersect c
  if properties.isNotEmpty?
    rule_ids = rule_ids intersect p
  if not rule_ids:
    rule_ids1 = check for region global (pincode -> area -> city -> state -> country ) : order of priority
    rule_ids2 = check for category global (product_id -> vertical -> sub_category -> main_category) : order of priority
    rule_ids3 = check for property global
    rule_ids = rule_ids1 + rule_ids2 + rule_ids3

If there is a single rule id, discount corresponding to the rule is returned. If there are multiple rules, boost parameter is used to decide which rule’s discount takes preference. There is also a provision to combine rules which can be defined at the campaign level

Persistent information is stored in mongodb and once the sale is started, it flushes redis and builds rules with discounting information in queryable format for faster response in redis (FYI : This is a one time process which takes time since it fetches all campaigns, rules from mongodb and populates redis). Post that all reads of discount calculation happens from redis

What is not solved

1. This module doesnt serve as a cache for discounts for given input. ie. calculation is done each time from redis.
2. Edits requires re-building information stored in redis and can be a bit time consuming. So this module is not meant for real time edits but its meant for real time calculation of discounts
3. Collision of names not handled. Ideal way would be each of category and region should be represented by a unique integer while creating rules.
4. Order of priority in a category/region always takes priority over composition of rules. Composition works on different dimensions and not on same dimensions