Sample reduction

Description: Reduces a configuration suite to a (greedily) minimal subset that preserves exactly the t-wise coverage the full suite already achieves.

Application: Shrinks an existing test suite of configurations without losing any of the t-wise feature interactions it already covers, reducing testing cost.

Example: A suite of 20 random configurations can often be reduced to a handful that keeps exactly the same pairwise coverage.

---

Code Examples

Python easy to use facade usage

from flamapy.interfaces.python.flamapy_feature_model import FLAMAFeatureModel
# Load the feature model
fm = FLAMAFeatureModel("path/to/feature/model")
# Reduce a configuration suite while keeping its pairwise (t=2) coverage
sample = fm.sampling(size=20)
result = fm.sample_reduction(sample, t=2)
print(result)

Python flamapy framework ADVANCED usage

from flamapy.core.discover import DiscoverMetamodels
# Initialize the discover metamodel
dm = DiscoverMetamodels()
# Get the fm metamodel representation using the transformation
# required to get to the fm metamodel
feature_model = dm.use_transformation_t2m("path/to/feature/model", 'fm')
# Manually call a M2M transformation to PySAT
sat_model = dm.use_transformation_m2m(feature_model, "pysat")
# Build the configuration suite to analyse, e.g. with a random sample
sampling = dm.get_operation(sat_model, 'PySATSampling')
sampling.set_sample_size(20)
sample = sampling.execute(sat_model).get_result()
# Get the operation
operation = dm.get_operation(sat_model, 'PySATSampleReduction')
# Set the configurations to reduce and the interaction strength
operation.set_configurations(sample)
operation.set_t(2)
# Execute the operation
operation.execute(sat_model)
# Get and print the result
result = operation.get_result()
print(result)

---

Facade signature: sample_reduction(self, configurations: Any, t: int = 2)

Inputs:

name	type	required	default
configurations	Any	yes	—
t	int	no	2

Default backend: sat

Implemented by:

plugin	exact	scale limit
pysat_metamodel	yes	—