Add functionality to compare pipelines (#95)

* Add functionality to compare pipelines
* Add tests to verify functionalities

molpipeline/utils/comparison.py

@@ -0,0 +1,101 @@
+"""Functions for comparing pipelines."""
+
+from typing import Any, TypeVar
+
+from molpipeline import Pipeline
+from molpipeline.utils.json_operations import recursive_to_json
+
+_T = TypeVar("_T", list[Any], tuple[Any, ...], set[Any], dict[Any, Any], Any)
+
+
+def remove_irrelevant_params(params: _T) -> _T:
+    """Remove irrelevant parameters from a dictionary.
+
+    Parameters
+    ----------
+    params : TypeVar
+        Parameters to remove irrelevant parameters from.
+
+    Returns
+    -------
+    TypeVar
+        Parameters without irrelevant parameters.
+    """
+    if isinstance(params, list):
+        return [remove_irrelevant_params(val) for val in params]
+    if isinstance(params, tuple):
+        return tuple(remove_irrelevant_params(val) for val in params)
+    if isinstance(params, set):
+        return {remove_irrelevant_params(val) for val in params}
+
+    irrelevant_params = ["n_jobs", "uuid", "error_filter_id"]
+    if isinstance(params, dict):
+        params_new = {}
+        for key, value in params.items():
+            if key.split("__")[-1] in irrelevant_params:
+                continue
+            params_new[key] = remove_irrelevant_params(value)
+        return params_new
+    return params
+
+
+def compare_recursive(  # pylint: disable=too-many-return-statements
+    value_a: Any, value_b: Any
+) -> bool:
+    """Compare two values recursively.
+
+    Parameters
+    ----------
+    value_a : Any
+        First value to compare.
+    value_b : Any
+        Second value to compare.
+
+    Returns
+    -------
+    bool
+        True if the values are the same, False otherwise.
+    """
+    if value_a.__class__ != value_b.__class__:
+        return False
+
+    if isinstance(value_a, dict):
+        if set(value_a.keys()) != set(value_b.keys()):
+            return False
+        for key in value_a:
+            if not compare_recursive(value_a[key], value_b[key]):
+                return False
+        return True
+
+    if isinstance(value_a, (list, tuple)):
+        if len(value_a) != len(value_b):
+            return False
+        for val_a, val_b in zip(value_a, value_b):
+            if not compare_recursive(val_a, val_b):
+                return False
+        return True
+    return value_a == value_b
+
+
+def check_pipelines_equivalent(pipeline_a: Pipeline, pipeline_b: Pipeline) -> bool:
+    """Check if two pipelines are the same.
+
+    Parameters
+    ----------
+    pipeline_a : Pipeline
+        Pipeline to compare.
+    pipeline_b : Pipeline
+        Pipeline to compare.
+
+    Returns
+    -------
+    bool
+        True if the pipelines are the same, False otherwise.
+    """
+    if not isinstance(pipeline_a, Pipeline) or not isinstance(pipeline_b, Pipeline):
+        raise ValueError("Both inputs should be of type Pipeline.")
+    pipeline_json_a = recursive_to_json(pipeline_a)
+    pipeline_json_a = remove_irrelevant_params(pipeline_json_a)
+    pipeline_json_b = recursive_to_json(pipeline_b)
+    pipeline_json_b = remove_irrelevant_params(pipeline_json_b)
+    return compare_recursive(pipeline_json_a, pipeline_json_b)

tests/test_utils/test_comparison.py

@@ -0,0 +1,35 @@
+"""Test the comparison functions."""
+
+from unittest import TestCase
+
+from molpipeline.utils.comparison import check_pipelines_equivalent
+from tests.utils.default_models import get_morgan_physchem_rf_pipeline
+
+
+class TestComparison(TestCase):
+    """Test if functional equivalent pipelines are detected as such."""
+
+    def test_are_equal(self) -> None:
+        """Test if two equivalent pipelines are detected as such."""
+
+        pipeline_a = get_morgan_physchem_rf_pipeline()
+        pipeline_b = get_morgan_physchem_rf_pipeline()
+        self.assertTrue(check_pipelines_equivalent(pipeline_a, pipeline_b))
+
+    def test_are_not_equal(self) -> None:
+        """Test if two different pipelines are detected as such."""
+        # Test changed parameters
+        pipeline_a = get_morgan_physchem_rf_pipeline()
+        pipeline_b = get_morgan_physchem_rf_pipeline()
+        pipeline_b.set_params(mol2fp__morgan__n_bits=1024)
+        self.assertFalse(check_pipelines_equivalent(pipeline_a, pipeline_b))
+
+        # Test changed steps
+        pipeline_b = get_morgan_physchem_rf_pipeline()
+        last_step = pipeline_b.steps[-1]
+        pipeline_b.steps = pipeline_b.steps[:-1]
+        self.assertFalse(check_pipelines_equivalent(pipeline_a, pipeline_b))
+
+        # Test if adding the step back makes the pipelines equivalent
+        pipeline_b.steps.append(last_step)
+        self.assertTrue(check_pipelines_equivalent(pipeline_a, pipeline_b))

tests/utils/default_models.py

@@ -0,0 +1,48 @@
+"""This module contains the default models used for testing molpipeline functions and classes."""
+
+from sklearn.ensemble import RandomForestClassifier
+
+from molpipeline import Pipeline
+from molpipeline.any2mol import SmilesToMol
+from molpipeline.error_handling import ErrorFilter, FilterReinserter
+from molpipeline.mol2any import (
+    MolToConcatenatedVector,
+    MolToMorganFP,
+    MolToRDKitPhysChem,
+)
+from molpipeline.post_prediction import PostPredictionWrapper
+
+
+def get_morgan_physchem_rf_pipeline() -> Pipeline:
+    """Get a pipeline combining Morgan fingerprints and physicochemical properties with a RandomForestClassifier.
+
+    Returns
+    -------
+    Pipeline
+        A pipeline combining Morgan fingerprints and physicochemical properties with a RandomForestClassifier.
+    """
+    error_filter = ErrorFilter(filter_everything=True)
+    pipeline = Pipeline(
+        [
+            ("smi2mol", SmilesToMol()),
+            (
+                "mol2fp",
+                MolToConcatenatedVector(
+                    [
+                        ("morgan", MolToMorganFP(n_bits=2048)),
+                        ("physchem", MolToRDKitPhysChem()),
+                    ]
+                ),
+            ),
+            ("error_filter", error_filter),
+            ("rf", RandomForestClassifier()),
+            (
+                "filter_reinserter",
+                PostPredictionWrapper(
+                    FilterReinserter.from_error_filter(error_filter, None)
+                ),
+            ),
+        ],
+        n_jobs=1,
+    )
+    return pipeline