import unittest import numpy as np from example import (Q, Description, ReversibleNetwork, ChannelQuotient, CatalyticPeeler, IndependentCatalysis, ExactCatalyticLaw, ChannelCensus, seed_cutoff, toy, probability, wilson) class ScientificChecks(unittest.TestCase): def test_canonical_identity_counts_and_gateways(self): self.assertEqual(Description('AA','A').canonical(),Description('A','AA')) self.assertNotEqual(Description('A','B').canonical(),Description('B','A').canonical()) self.assertEqual(Description('AB','AB').canonical(),Description('AB','AB')) for n in range(2,9): q=ChannelQuotient.binary(n);row=ChannelCensus.formula(n) self.assertEqual(len(q.split.channels),row['split']);self.assertEqual(len(q.quotient.channels),row['quotient']) self.assertLessEqual(row['deleted'],row['elementary_loss_bound']);self.assertLessEqual(max(map(len,q.fibres)),2) if n>=4:self.assertEqual((len(q.quotient.gateways()),len(q.quotient.gateways(True))),(34,30)) self.assertEqual(ChannelCensus.formula(12)['deleted'],96) self.assertEqual(ChannelCensus.formula(24)['quotient'],738196506) with self.assertRaises(ValueError):ChannelQuotient(ReversibleNetwork(('A','AA'),('A',),(Description('A','A'),Description('A','A')))) def test_closure_projection_and_witness_lifting(self): q=toy() for mask in range(8): active=[i for i in range(3) if mask>>i&1];projected=set(q.mapping[i] for i in active) self.assertEqual(q.split.closure(active),q.quotient.closure(projected)) for marks in ExactCatalyticLaw(q.split).configurations(): a=CatalyticPeeler(q.split).run(marks);b=CatalyticPeeler(q.quotient).run(q.project_marks(marks)) brute=any(q.split.is_raf([j for j in range(3) if mask>>j&1],marks) for mask in range(1,8)) self.assertEqual(a['has_raf'],brute);self.assertEqual(a['has_raf'],b['has_raf']) if b['has_raf']:self.assertTrue(q.split.is_raf(q.lift_witness(b['usable'],marks),marks)) # Only the noncanonical duplicate is catalysed: fixed representative loses RAF. marks=(0,1,0);b=CatalyticPeeler(q.quotient).run(q.project_marks(marks));self.assertTrue(b['has_raf']) self.assertEqual(q.lift_witness(b['usable'],marks),(1,)) self.assertFalse(q.split.is_raf((0,),marks)) def test_exact_probability_laws_and_domination(self): q=toy();sp=ExactCatalyticLaw(q.split);qt=ExactCatalyticLaw(q.quotient) for p,a,b in [(Q(1,2),Q(4077,4096),Q(239,256)),(Q(1,3),Q(505585,531441),Q(5137,6561))]: self.assertEqual(sp.probability([p]*3),a);self.assertEqual(qt.probability(q.or_parameters(p)),a);self.assertEqual(qt.probability([p]*2),b) for a in (Q(0),Q(1,10),Q(1,2),Q(1)): self.assertTrue(all(p<=a for p in q.or_parameters(a/2))) self.assertTrue(all(p>=a for p in q.or_parameters(a))) with self.assertRaises(ValueError):probability('11/10') with self.assertRaises(ValueError):probability(.1) def test_complete_history_identity_and_exact_column_weights(self): law=ExactCatalyticLaw(toy().quotient);histories=law.history_polynomials() self.assertEqual(histories,law.history_polynomials(True));self.assertEqual(len(histories),6) for p in (Q(0),Q(1,3),Q(1,2),Q(1)): total=Q(0) for history,counts in histories.items(): expected=sum(count*p**k*(1-p)**(law.bits-k) for k,count in counts.items()) self.assertEqual(law.history_probability(history,p),expected);total+=expected self.assertEqual(total,1) def test_reversible_cleavage_food_only_and_unusable_channels(self): network=ReversibleNetwork(('A','B','AB'),('AB',),(Description('A','B'),)) self.assertEqual(network.closure((0,)),7);self.assertTrue(network.is_raf((0,),(4,))) food_only=ReversibleNetwork(('A','AA'),('A','AA'),(Description('A','A'),)) result=CatalyticPeeler(food_only).run((1,));self.assertTrue(result['has_raf']);self.assertEqual(result['closure'],food_only.food) unusable=ReversibleNetwork(('A','AA','AAA','AAAA'),('A',),(Description('AA','AA'),)) result=CatalyticPeeler(unusable).run((1,));self.assertFalse(result['has_raf']);self.assertEqual(result['history'],((0,),)) # Correlated duplicate observations keep p; independent OR would change it. q=toy() for marks in ExactCatalyticLaw(q.quotient).configurations(): duplicated=tuple(marks[j] for j in q.mapping);self.assertEqual(q.project_marks(duplicated),marks) def test_normalization_cutoffs_and_finite_event_scope(self): row=ChannelCensus.formula(12);p=Q(12,row['quotient']) self.assertAlmostEqual(ChannelCensus.openness(p,row['M']),.699150,places=6) self.assertEqual(seed_cutoff('1/10',10)['L'],536670);self.assertEqual(seed_cutoff('1/2',10)['L'],20820) net=ChannelQuotient.binary(4).quotient for j in net.gateways(): H=net.closure((j,));escape=any(len(w)>2 for w in net.generated_words(H)) self.assertEqual(escape,j in net.gateways(True)) inactive=set(range(len(net.channels)))-set(net.gateways(True));self.assertEqual(net.closure(inactive),net.food) self.assertEqual(ChannelCensus.openness(0,100),0);self.assertEqual(ChannelCensus.openness(1,100),1) def test_sampling_reproducibility_boundaries_and_limits(self): net=toy().quotient;a=IndependentCatalysis(np.random.default_rng(123));b=IndependentCatalysis(np.random.default_rng(123)) for _ in range(20):self.assertEqual(a.sample(net,'1/3'),b.sample(net,'1/3')) self.assertEqual(a.sample(net,0),(0,0));self.assertEqual(a.sample(net,1),(15,15)) self.assertEqual(wilson(0,100)[0],0);self.assertEqual(wilson(100,100)[1],1);self.assertIsNone(wilson(0,0)) self.assertGreater(wilson(0,100)[1],0);self.assertLess(wilson(100,100)[0],1) with self.assertRaises(ValueError):ExactCatalyticLaw(ChannelQuotient.binary(3).quotient) with self.assertRaises(ValueError):ChannelQuotient.binary(13) with self.assertRaises(ValueError):CatalyticPeeler(net).run((16,0)) if __name__=='__main__':unittest.main()