Merge pull request #70 from AutomatedProcessImprovement/update_rules

Update rules

Author: SimpleSquirrelz

prosimos/all_attributes.py

@@ -15,6 +15,7 @@ def __init__(self, global_attributes, case_attributes, event_attributes):
         global_case_attribute_names = global_attribute_names.intersection(all_case_attribute_names)
         global_event_attribute_names = global_attribute_names.intersection(all_event_attribute_names)
 
+        self.global_attribute_initial_values = global_attributes.get_values_calculated()
         self.global_attributes = self._extract_global_attributes(global_attributes, global_case_attribute_names, global_event_attribute_names)
 
         self.global_case_attributes, self.case_attributes = \

prosimos/branch_condition_parser.py

@@ -7,6 +7,7 @@
     BranchConditionWithRule
 )
 
+
 class BranchConditionParser:
     def __init__(self, json_data_with_branching_conditions):
         self.data = json_data_with_branching_conditions
@@ -17,25 +18,20 @@ def parse(self):
             or_rules_json = curr_branch_condition_rule["rules"]
             branch_id = curr_branch_condition_rule["id"]
 
-            # print(f"OR RULES JSON: {str(or_rules_json)}")
             or_rules: List[AndBranchConditionRule] = []
             for or_rule in or_rules_json:
                 and_rules: List[BranchConditionRule] = []
-                # print(f"OR RULE: {str(or_rule)}")
                 for and_rule_json in or_rule:
-                    # print(f"AND RULE JSON: {str(and_rule_json)}")
                     rule: BranchConditionRule = BranchConditionRule(
                         and_rule_json["attribute"],
                         and_rule_json["comparison"],
                         and_rule_json["value"],
                     )
                     and_rules.append(rule)
-                # print(f"AND RULES: {str(or_rule)}")
                 all_and_rules = AndBranchConditionRule(and_rules)
                 or_rules.append(all_and_rules)
 
             all_or_rules = OrBranchConditionRule(or_rules)
-            # print(f"ALL OR RULES: {all_or_rules}\n\n")
 
             branching_conditions.append(BranchConditionWithRule(all_or_rules, branch_id))
 

prosimos/branch_condition_rules.py

@@ -26,7 +26,6 @@ def _parse_value(self, value):
         if self.condition == "in":
             min_boundary = float(value[0])
             max_boundary = sys.maxsize if value[1] == "inf" else float(value[1])
-
             return min_boundary, max_boundary
         else:
             return value
@@ -35,12 +34,34 @@ def is_rule_true(self, all_case_values):
         if self.attribute not in all_case_values:
             return False
 
-        case_value = all_case_values[self.attribute]
+        raw_case_value = all_case_values[self.attribute]
+        if self.condition in [">", ">=", "<", "<=", "!="]:
+            try:
+                case_value = float(raw_case_value)
+                comparison_value = float(self.value)
+            except ValueError:
+                case_value = raw_case_value
+                comparison_value = self.value
+        else:
+            case_value = raw_case_value
+            comparison_value = self.value
+
         if self.condition == "in":
-            evaluator = InOperatorEvaluator(self.value, case_value)
-            return evaluator.eval()
+            min_value, max_value = comparison_value
+            return min_value <= case_value <= max_value
+        elif self.condition == ">":
+            return case_value > comparison_value
+        elif self.condition == ">=":
+            return case_value >= comparison_value
+        elif self.condition == "<":
+            return case_value < comparison_value
+        elif self.condition == "<=":
+            return case_value <= comparison_value
+        elif self.condition == "!=":
+            return case_value != comparison_value
         else:
-            return self.value == case_value
+            return case_value == comparison_value
+
 
 
 class AndBranchConditionRule:

prosimos/case_attributes.py

@@ -52,12 +52,15 @@ def get_next_value(self):
             return self.value.generate_sample(1)[0]
 
     def validate(self):
+        epsilon = 1e-6
+
         if self.case_atrr_type == CASE_ATTR_TYPE.DISCRETE:
             actual_sum_probabilities = reduce(lambda acc, item: acc + item, self.value["probabilities"], 0) 
 
-            if actual_sum_probabilities != 1:
-                raise InvalidCaseAttributeException(f"Case attribute ${self.name}: probabilities' sum should be equal to 1") 
-        
+            if not (1 - epsilon <= actual_sum_probabilities <= 1 + epsilon):
+                raise InvalidCaseAttributeException(
+                    f"Case attribute {self.name}: probabilities' sum should be equal to 1")
+
         return True
 
 

prosimos/event_attributes.py

@@ -1,19 +1,22 @@
+import pprint
 from enum import Enum
-from functools import reduce
 from random import choices
 from typing import Dict
 import ast
 import operator as op
+import numpy as np
 
 from pix_framework.statistics.distribution import DurationDistribution
 
 from prosimos.exceptions import InvalidEventAttributeException
+import math
 
 
 class EVENT_ATTR_TYPE(Enum):
     DISCRETE = "discrete"
     CONTINUOUS = "continuous"
     EXPRESSION = "expression"
+    DTREE = "dtree"
 
 
 operators = {
@@ -24,18 +27,67 @@ class EVENT_ATTR_TYPE(Enum):
     ast.FloorDiv: op.floordiv
 }
 
+def fix(mean):
+    return DurationDistribution(name="fix", mean=mean, var=0.0, std=0.0, minimum=mean, maximum=mean).generate_sample(1)[0]
 
-def parse_discrete_value(value_info_arr):
-    prob_arr = []
-    options_arr = []
-    for item in value_info_arr:
-        options_arr.append(item["key"])
-        prob_arr.append(float(item["value"]))
 
-    return {
-        "options": options_arr,
-        "probabilities": prob_arr
-    }
+def uniform(minimum, maximum):
+    return DurationDistribution(name="uniform", minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+def norm(mean, std, minimum=None, maximum=None):
+    return DurationDistribution(name="norm", mean=mean, std=std, minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+def triang(c, minimum, maximum):
+    return DurationDistribution(name="triang", mean=c, minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+def expon(mean, minimum=None, maximum=None):
+    return DurationDistribution(name="expon", mean=mean, minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+def lognorm(mean, var, minimum=None, maximum=None):
+    return DurationDistribution(name="lognorm", mean=mean, var=var, minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+def gamma(mean, var, minimum=None, maximum=None):
+    return DurationDistribution(name="gamma", mean=mean, var=var, minimum=minimum, maximum=maximum).generate_sample(1)[0]
+
+
+distributions = {
+    'fix': fix,
+    'uniform': uniform,
+    'norm': norm,
+    'triang': triang,
+    'expon': expon,
+    'lognorm': lognorm,
+    'gamma': gamma
+}
+
+math_functions = {name: getattr(math, name) for name in dir(math) if callable(getattr(math, name))}
+
+
+def parse_discrete_value(value_info):
+    if isinstance(value_info, list):
+        prob_arr = []
+        options_arr = []
+        for item in value_info:
+            options_arr.append(item["key"])
+            prob_arr.append(float(item["value"]))
+
+        return {
+            "type": "discrete",
+            "options": options_arr,
+            "probabilities": prob_arr
+        }
+    elif isinstance(value_info, dict):
+        return {
+            "type": "markov",
+            "transitions": value_info
+        }
+    else:
+        raise ValueError("Unsupported value_info format for discrete value")
 
 
 def parse_continuous_value(value_info) -> "DurationDistribution":
@@ -57,7 +109,8 @@ def _eval(node):
             elif isinstance(node, ast.UnaryOp):
                 return operators[type(node.op)](_eval(node.operand))
             elif isinstance(node, ast.Name):
-                return vars_dict[node.id]
+                if node.id in vars_dict:
+                    return vars_dict[node.id]
             elif isinstance(node, ast.Str):
                 return node.s
             elif isinstance(node, ast.Compare):
@@ -67,14 +120,33 @@ def _eval(node):
                     return all(_eval(value) for value in node.values)
                 elif type(node.op) is ast.Or:
                     return any(_eval(value) for value in node.values)
+            elif isinstance(node, ast.Call):
+                func_name = node.func.id
+                args = [_eval(arg) for arg in node.args]
+                if func_name in distributions:
+                    return distributions[func_name](*args)
+                elif func_name in math_functions:
+                    args = [_eval(arg) for arg in node.args]
+                    try:
+                        return math_functions[func_name ](*args)
+                    except OverflowError:
+                        return np.finfo(np.float32).max
+                    except ValueError as e:
+                        return 0
             else:
-                return None
+                return 0
         except (SyntaxError, ZeroDivisionError, TypeError, KeyError):
-            return None
+            return 0
 
     return _eval(tree.body)
 
 
+def evaluate_dtree(dtree, vars_dict):
+    for conditions, formula in dtree:
+        if conditions is True or all(eval_expr(cond, vars_dict) for cond in conditions):
+            return eval_expr(formula, vars_dict)
+    return None
+
 class EventAttribute:
     def __init__(self, event_id, name, event_attr_type, value):
         self.event_id: str = event_id
@@ -87,27 +159,76 @@ def __init__(self, event_id, name, event_attr_type, value):
             self.value = parse_continuous_value(value)
         elif self.event_attr_type == EVENT_ATTR_TYPE.EXPRESSION:
             self.value = value
+        elif self.event_attr_type == EVENT_ATTR_TYPE.DTREE:
+            self.value = value
         else:
             raise Exception(f"Not supported event attribute {type}")
 
         self.validate()
 
-    def get_next_value(self, all_attributes={}):
+    def get_next_value(self, all_attributes):
         if self.event_attr_type == EVENT_ATTR_TYPE.DISCRETE:
-            one_choice_arr = choices(self.value["options"], self.value["probabilities"])
-            return one_choice_arr[0]
+            if self.value["type"] == "markov":
+                current_value = all_attributes.get(self.name, None)
+                next_state = self.get_next_markov_state(current_value)
+                if next_state is not None:
+                    all_attributes[self.name] = next_state
+                    return next_state
+                return current_value
+            else:
+                one_choice_arr = choices(self.value["options"], self.value["probabilities"])
+                return one_choice_arr[0]
+
         elif self.event_attr_type == EVENT_ATTR_TYPE.EXPRESSION:
-            return eval_expr(self.value, all_attributes)
+            result = eval_expr(self.value, all_attributes)
+            if isinstance(result, (int, float, np.number)) and not isinstance(result, bool):
+                if result == 0:  # Specifically handle zero without adjusting to tiny (in case of any errors in eval)
+                    return 0
+                elif result == float('inf'):
+                    return np.finfo(np.float32).max
+                elif result == -float('inf'):
+                    return np.finfo(np.float32).min
+                elif abs(result) < np.finfo(np.float32).tiny:
+                    return np.finfo(np.float32).tiny
+                elif abs(result) > np.finfo(np.float32).max:
+                    return np.finfo(np.float32).max if result > 0 else np.finfo(np.float32).min
+                else:
+                    return result
+            else:
+                return result
+        elif self.event_attr_type == EVENT_ATTR_TYPE.DTREE:
+            result = evaluate_dtree(self.value, all_attributes)
+            if result is not None:
+                return result
+            return 0
         else:
             return self.value.generate_sample(1)[0]
 
+    def get_next_markov_state(self, current_value):
+        transitions = self.value["transitions"]
+        if current_value in transitions:
+            current_transitions = transitions[current_value]
+            options, probabilities = zip(*current_transitions.items())
+            return choices(options, probabilities)[0]
+        return current_value
+
     def validate(self):
-        if self.event_attr_type == EVENT_ATTR_TYPE.DISCRETE:
-            actual_sum_probabilities = reduce(lambda acc, item: acc + item, self.value["probabilities"], 0)
+        epsilon = 1e-6
 
-            if actual_sum_probabilities != 1:
-                raise InvalidEventAttributeException(
-                    f"Event attribute ${self.name}: probabilities' sum should be equal to 1")
+        if self.event_attr_type == EVENT_ATTR_TYPE.DISCRETE:
+            if self.value["type"] == "discrete":
+                actual_sum_probabilities = sum(self.value["probabilities"])
+
+                if not (1 - epsilon <= actual_sum_probabilities <= 1 + epsilon):
+                    raise InvalidEventAttributeException(
+                        f"Event attribute {self.name}: probabilities' sum should be equal to 1")
+            elif self.value["type"] == "markov":
+                for state, transitions in self.value["transitions"].items():
+                    actual_sum_probabilities = sum(transitions.values())
+                    if not (1 - epsilon <= actual_sum_probabilities <= 1 + epsilon):
+                        raise InvalidEventAttributeException(
+                            f"Event attribute {self.name}, state {state}: "
+                            f"probabilities' sum should be equal to 1")
 
         return True
 

prosimos/gateway_condition_choice.py

@@ -1,10 +1,15 @@
 class GatewayConditionChoice:
-    def __init__(self, candidates_list, rules_list):
+    def __init__(self, candidates_list, rules_list, default_path=None):
         self.candidates_list = candidates_list
         self.rules_list = rules_list
+        self.default_path = default_path
 
-    def get_outgoing_flow(self, case_attributes):
-        return [candidate for candidate, rule in zip(self.candidates_list, self.rules_list)
-                if rule.is_true(case_attributes)]
+    def set_default(self, default_path):
+        self.default_path = default_path
 
+    def get_outgoing_flow(self, attributes):
+        return [candidate for candidate, rule in zip(self.candidates_list, self.rules_list)
+                if rule.is_true(attributes)]
 
+    def get_default_path(self):
+        return self.default_path