Update to new 1.18 Feature API

Author: ruivieira

README.md

@@ -1,4 +1,4 @@
-![version](https://img.shields.io/badge/version-0.0.9-green) ![TrustyAI](https://img.shields.io/badge/TrustyAI-1.17-green) [![Tests](https://github.com/trustyai-python/module/actions/workflows/workflow.yml/badge.svg)](https://github.com/trustyai-python/examples/actions/workflows/workflow.yml)
+![version](https://img.shields.io/badge/version-0.0.9-green) ![TrustyAI](https://img.shields.io/badge/TrustyAI-1.18-green) [![Tests](https://github.com/trustyai-python/module/actions/workflows/workflow.yml/badge.svg)](https://github.com/trustyai-python/examples/actions/workflows/workflow.yml)
 
 # python-trustyai
 

setup.py

@@ -7,7 +7,7 @@
 with open(os.path.join(here, 'README.md'), encoding='utf-8') as f:
     long_description = f.read()
 
-TRUSTY_VERSION = "1.17.0.Final"
+TRUSTY_VERSION = "1.18.0.Final"
 
 
 class PostInstall(install):
@@ -27,7 +27,7 @@ def run(self):
 
 setup(
     name="trustyai",
-    version="0.0.9",
+    version="0.1.0",
     description="Python bindings to the TrustyAI explainability library",
     long_description=long_description,
     long_description_content_type="text/markdown",

tests/test_conversions.py

@@ -72,3 +72,21 @@ def test_feature_function():
     assert f3.name == "f-3"
     assert f3.value.as_string() == "foo"
     assert f3.type == Type.CATEGORICAL
+
+
+def test_feature_domains():
+    """Test domains"""
+    f1 = feature(name="f-1", value=1.0, dtype="number")
+    assert f1.name == "f-1"
+    assert f1.value.as_number() == 1.0
+    assert f1.type == Type.NUMBER
+    assert f1.domain is None
+    assert f1.is_constrained
+
+    f2 = feature(name="f-2", value=2.0, dtype="number", domain=(0.0, 10.0))
+    assert f2.name == "f-2"
+    assert f2.value.as_number() == 2.0
+    assert f2.type == Type.NUMBER
+    assert f2.domain
+    print(f2.domain)
+    assert not f2.is_constrained

tests/test_counterfactualexplainer.py

@@ -1,11 +1,10 @@
 # pylint: disable=import-error, wrong-import-position, wrong-import-order, R0801
 """Test suite for counterfactual explanations"""
 
-from common import *
-from pytest import approx
-
 from java.util import Random
+from pytest import approx
 
+from common import *
 from trustyai.explainers import CounterfactualExplainer
 from trustyai.model import (
     counterfactual_prediction,
@@ -24,17 +23,15 @@ def test_non_empty_input():
 
     goal = [output(name="f-num1", dtype="number", value=10.0, score=0.0)]
     features = [
-        feature(name=f"f-num{i}", value=i * 2.0, dtype="number")
+        feature(name=f"f-num{i}", value=i * 2.0, dtype="number", domain=(0.0, 1000.0))
         for i in range(n_features)
     ]
-    domains = [(0.0, 1000.0)] * n_features
 
     model = TestUtils.getSumSkipModel(0)
 
     prediction = counterfactual_prediction(
         input_features=features,
         outputs=goal,
-        domains=domains
     )
 
     counterfactual_result = explainer.explain(prediction, model)
@@ -48,9 +45,8 @@ def test_counterfactual_match():
     goal = [output(name="inside", dtype="bool", value=True, score=0.0)]
 
     features = [
-        feature(name=f"f-num{i + 1}", value=10.0, dtype="number") for i in range(4)
+        feature(name=f"f-num{i + 1}", value=10.0, dtype="number", domain=(0.0, 1000.0)) for i in range(4)
     ]
-    domains = [(0.0, 1000.0)] * 4
 
     center = 500.0
     epsilon = 10.0
@@ -60,7 +56,6 @@ def test_counterfactual_match():
     prediction = counterfactual_prediction(
         input_features=features,
         outputs=goal,
-        domains=domains
     )
     model = TestUtils.getSumThresholdModel(center, epsilon)
     result = explainer.explain(prediction, model)
@@ -86,55 +81,17 @@ def test_counterfactual_match_python_model():
     n_features = 5
 
     features = [
-        feature(name=f"f-num{i + 1}", value=10.0, dtype="number") for i in range(n_features)
+        feature(name=f"f-num{i + 1}", value=10.0, dtype="number", domain=(0.0, 1000.0)) for i in range(n_features)
     ]
-    domains = [(0.0, 1000.0)] * n_features
 
     explainer = CounterfactualExplainer(steps=1000)
 
     prediction = counterfactual_prediction(
         input_features=features,
         outputs=goal,
-        domains=domains
     )
 
     model = Model(sum_skip_model)
 
     result = explainer.explain(prediction, model)
     assert sum([entity.as_feature().value.as_number() for entity in result.entities]) == approx(GOAL_VALUE, rel=3)
-
-
-def test_default_constraints():
-    goal = [output(name="sum-but-0", dtype="number", value=1000, score=1.0)]
-
-    n_features = 5
-
-    features = [
-        feature(name=f"f-num{i + 1}", value=10.0, dtype="number") for i in range(n_features)
-    ]
-    domains = [(0.0, 1000.0)] * n_features
-
-    prediction = counterfactual_prediction(
-        input_features=features,
-        outputs=goal,
-        domains=domains
-    )
-
-    assert len(prediction.constraints) == n_features
-
-    n_features = 10
-
-    features = [
-        feature(name=f"f-num{i + 1}", value=10.0, dtype="number") for i in range(n_features)
-    ]
-    domains = [(0.0, 1000.0)] * n_features
-
-    constaints = [False] * n_features
-    prediction = counterfactual_prediction(
-        input_features=features,
-        outputs=goal,
-        domains=domains,
-        constraints=constaints
-    )
-
-    assert len(prediction.constraints) == n_features

trustyai/__init__.py

@@ -1,24 +1,25 @@
 # pylint: disable = import-error, import-outside-toplevel, dangerous-default-value, invalid-name, R0801
 """Main TrustyAI Python bindings"""
-from typing import List
-import site
 import os
+import site
 import uuid
+from typing import List
+
 import jpype
 import jpype.imports
 from jpype import _jcustomizer, _jclass
 
-TRUSTY_VERSION = "1.17.0.Final"
+TRUSTY_VERSION = "1.18.0.Final"
 DEFAULT_DEP_PATH = os.path.join(site.getsitepackages()[0], "trustyai", "dep")
 
 CORE_DEPS = [
     f"{DEFAULT_DEP_PATH}/org/kie/kogito/explainability-core/{TRUSTY_VERSION}/*",
     f"{DEFAULT_DEP_PATH}/org/slf4j/slf4j-api/1.7.30/slf4j-api-1.7.30.jar",
     f"{DEFAULT_DEP_PATH}/org/apache/commons/commons-lang3/3.12.0/commons-lang3-3.12.0.jar",
-    f"{DEFAULT_DEP_PATH}/org/optaplanner/optaplanner-core/8.17.0.Final/"
-    f"optaplanner-core-8.17.0.Final.jar",
+    f"{DEFAULT_DEP_PATH}/org/optaplanner/optaplanner-core-impl/8.18.0.Final/"
+    f"optaplanner-core-impl-8.18.0.Final.jar",
     f"{DEFAULT_DEP_PATH}/org/apache/commons/commons-math3/3.6.1/commons-math3-3.6.1.jar",
-    f"{DEFAULT_DEP_PATH}/org/kie/kie-api/8.17.0.Beta/kie-api-8.17.0.Beta.jar",
+    f"{DEFAULT_DEP_PATH}/org/kie/kie-api/8.18.0.Beta/kie-api-8.18.0.Beta.jar",
     f"{DEFAULT_DEP_PATH}/io/micrometer/micrometer-core/1.8.2/micrometer-core-1.8.2.jar",
 ]
 

trustyai/model/__init__.py

@@ -1,11 +1,14 @@
 # pylint: disable = import-error, too-few-public-methods, invalid-name, duplicate-code
 """General model classes"""
-from typing import List, Optional, Tuple
 import uuid as _uuid
+from typing import List, Optional
 
 from java.lang import Long
-from java.util.concurrent import CompletableFuture, ForkJoinPool
+from java.util.concurrent import CompletableFuture
 from jpype import JImplements, JOverride, _jcustomizer, _jclass
+from org.kie.kogito.explainability.local.counterfactual.entities import (
+    CounterfactualEntity,
+)
 from org.kie.kogito.explainability.model import (
     CounterfactualPrediction as _CounterfactualPrediction,
     DataDistribution,
@@ -21,8 +24,9 @@
     Value as _Value,
     Type as _Type,
 )
-from org.kie.kogito.explainability.local.counterfactual.entities import (
-    CounterfactualEntity,
+
+from org.kie.kogito.explainability.model.domain import (
+    EmptyFeatureDomain as _EmptyFeatureDomain,
 )
 
 from trustyai.model.domain import feature_domain
@@ -165,6 +169,19 @@ def value(self):
         """Return value"""
         return self.getValue()
 
+    @property
+    def domain(self):
+        """Return domain"""
+        _domain = self.getDomain()
+        if isinstance(_domain, _EmptyFeatureDomain):
+            return None
+        return _domain
+
+    @property
+    def is_constrained(self):
+        """Return contraint"""
+        return self.isConstrained()
+
 
 @_jcustomizer.JImplementationFor("org.kie.kogito.explainability.model.Value")
 # pylint: disable=no-member
@@ -222,11 +239,6 @@ def output(self) -> PredictionOutput:
         """Return input"""
         return self.getOutput()
 
-    @property
-    def constraints(self):
-        """Return constraints"""
-        return self.getConstraints()
-
     @property
     def data_distribution(self):
         """Return data distribution"""
@@ -266,16 +278,22 @@ def output(name, dtype, value=None, score=1.0) -> _Output:
     return _Output(name, _type, Value(value), score)
 
 
-def feature(name: str, dtype: str, value=None) -> Feature:
+def feature(name: str, dtype: str, value=None, domain=None) -> Feature:
     """Helper method to build features"""
+
     if dtype == "categorical":
-        _feature = FeatureFactory.newCategoricalFeature(name, value)
+        _factory = FeatureFactory.newCategoricalFeature
     elif dtype == "number":
-        _feature = FeatureFactory.newNumericalFeature(name, value)
+        _factory = FeatureFactory.newNumericalFeature
     elif dtype == "bool":
-        _feature = FeatureFactory.newBooleanFeature(name, value)
+        _factory = FeatureFactory.newBooleanFeature
     else:
-        _feature = FeatureFactory.newObjectFeature(name, value)
+        _factory = FeatureFactory.newObjectFeature
+
+    if domain:
+        _feature = _factory(name, value, feature_domain(domain))
+    else:
+        _feature = _factory(name, value)
     return _feature
 
 
@@ -291,8 +309,6 @@ def simple_prediction(
 def counterfactual_prediction(
     input_features: List[Feature],
     outputs: List[Output],
-    domains: List[Optional[Tuple]],
-    constraints: Optional[List[bool]] = None,
     data_distribution: Optional[DataDistribution] = None,
     uuid: Optional[_uuid.UUID] = None,
     timeout: Optional[float] = None,
@@ -302,19 +318,10 @@ def counterfactual_prediction(
         uuid = _uuid.uuid4()
     if timeout:
         timeout = Long(timeout)
-    if not constraints:
-        constraints = [False] * len(input_features)
-
-    # build the feature domains from the Python tuples
-    java_domains = _jclass.JClass("java.util.Arrays").asList(
-        [feature_domain(domain) for domain in domains]
-    )
 
     return CounterfactualPrediction(
         PredictionInput(input_features),
         PredictionOutput(outputs),
-        PredictionFeatureDomain(java_domains),
-        constraints,
         data_distribution,
         uuid,
         timeout,