Merge branch '8-certain-git-histories-can-lead-to-a-stack-overflow-error' into next-release

Author: AlexanderSchultheiss

src/main/java/org/variantsync/vevos/simulation/variability/SPLCommit.java

@@ -138,7 +138,7 @@ public Optional<SPLCommit[]> parents() {
         return Optional.ofNullable(parents);
     }
 
-    public void setParents(final SPLCommit[] parents) {
+    public void setParents(final SPLCommit... parents) {
         this.parents = parents;
     }
 

src/main/java/org/variantsync/vevos/simulation/variability/sequenceextraction/LongestNonOverlappingSequences.java

@@ -83,29 +83,52 @@ public List<NonEmptyList<SPLCommit>> extract(final Collection<SPLCommit> commits
             }
         }
 
+        // Sort the sequences once more, after filtering them
+        commitSequences.sort((o1, o2) -> Integer.compare(o2.size(), o1.size()));
         return commitSequences;
     }
 
-    // Recursively build sequences
     private static Set<LinkedList<SPLCommit>> retrieveSequencesForStart(final Map<SPLCommit, Set<SPLCommit>> parentChildMap, final SPLCommit start) {
-        if (!parentChildMap.containsKey(start)) {
-            // Create a new sequence that contains the start commit
-            final Set<LinkedList<SPLCommit>> sequenceSet = new HashSet<>();
+        // Set for holding all retrieved sequences
+        Set<LinkedList<SPLCommit>> sequenceSet = new HashSet<>();
+        {
+            // We start with the 'start' commit
             final LinkedList<SPLCommit> sequence = new LinkedList<>();
             sequence.add(start);
             sequenceSet.add(sequence);
-            return sequenceSet;
-        } else {
-            // Collect the sequences of the children and prepend the commit to each of them as parent
-            final Set<LinkedList<SPLCommit>> sequences = new HashSet<>();
-            for (final SPLCommit child : parentChildMap.get(start)) {
-                final Set<LinkedList<SPLCommit>> childSequences = retrieveSequencesForStart(parentChildMap, child);
-                for (final LinkedList<SPLCommit> childSequence : childSequences) {
-                    childSequence.addFirst(start);
-                    sequences.add(childSequence);
+        }
+
+        // We continue to build sequences, going from parent to descendents, until the ends of all possible sequences have been reached
+        boolean incomplete = true;
+        while(incomplete) {
+            Set<LinkedList<SPLCommit>> updatedSet = new HashSet<>();
+            // Set to false at the start of the iteration, it is set to true if at least one sequence has been extended
+            incomplete = false;
+            // For all found sequences, check whether their last commit has more children that can be added to the sequence
+            // if multiple children exist, new sequences are created
+            for (LinkedList<SPLCommit> sequence : sequenceSet) {
+                final SPLCommit parent = sequence.getLast();
+                if (parentChildMap.containsKey(parent)) {
+                    final Set<SPLCommit> children = parentChildMap.get(parent);
+                    for (SPLCommit child : children) {
+                        // There is at least one child, we are not done yet
+                        incomplete = true;
+                        // Create a new sequence from the old sequence for each child
+                        final LinkedList<SPLCommit> anotherSequence = new LinkedList<>(sequence);
+                        anotherSequence.add(child);
+                        // Add the new sequence to the new set
+                        updatedSet.add(anotherSequence);
+                    }
+                } else {
+                    // There are no children, simply add the old sequence to the updated set
+                    updatedSet.add(sequence);
                 }
             }
-            return sequences;
+            // Update the sequence set
+            sequenceSet = updatedSet;
         }
+
+        return sequenceSet;
     }
 }
+

src/test/java/org/variantsync/vevos/simulation/variability/sequenceextraction/LongestNonOverlappingSequencesTest.java

@@ -0,0 +1,94 @@
+package org.variantsync.vevos.simulation.variability.sequenceextraction;
+
+import org.junit.Test;
+import org.variantsync.functjonal.list.NonEmptyList;
+import org.variantsync.vevos.simulation.variability.SPLCommit;
+
+import java.util.Arrays;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+public class LongestNonOverlappingSequencesTest {
+
+    @Test
+    public void exampleExtraction() {
+        // For example, if the commits comprise three partially overlapping sequences ([A-B-C-D-E], [X-Y-Z], [A-B-F-G]),
+        // the function will return the sequences ([A-B-C-D-E], [X-Y-Z], [F-G]).
+        final SPLCommit a = new SPLCommit("A");
+        final SPLCommit b = new SPLCommit("B");
+        final SPLCommit c = new SPLCommit("C");
+        final SPLCommit d = new SPLCommit("D");
+        final SPLCommit e = new SPLCommit("E");
+        final SPLCommit f = new SPLCommit("F");
+        final SPLCommit g = new SPLCommit("G");
+        final SPLCommit x = new SPLCommit("X");
+        final SPLCommit y = new SPLCommit("Y");
+        final SPLCommit z = new SPLCommit("Z");
+        // A has no parent
+        a.setParents();
+        // A-B
+        b.setParents(a);
+        // B-C
+        c.setParents(b);
+        // C-D
+        d.setParents(c);
+        // D-E
+        e.setParents(d);
+        // B-F
+        f.setParents(b);
+        // F-G
+        g.setParents(f);
+        // X has no parent
+        x.setParents();
+        // X-Y
+        y.setParents(x);
+        // Y-Z
+        z.setParents(y);
+
+        List<SPLCommit> exampleCommits = Arrays.asList(a, b, c, d, e, f, g, x, y, z);
+        NonEmptyList<SPLCommit> expectedOne = new NonEmptyList<>(Arrays.asList(a,b,c,d,e));
+        NonEmptyList<SPLCommit> expectedTwo = new NonEmptyList<>(Arrays.asList(x, y, z));
+        NonEmptyList<SPLCommit> expectedThree = new NonEmptyList<>(Arrays.asList(f, g));
+
+        LongestNonOverlappingSequences algorithm = new LongestNonOverlappingSequences();
+        List<NonEmptyList<SPLCommit>> result = algorithm.extract(exampleCommits);
+
+        assert result.size() == 3;
+        assert sequencesAreEqual(result.get(0), expectedOne);
+        assert sequencesAreEqual(result.get(1), expectedTwo);
+        assert sequencesAreEqual(result.get(2), expectedThree);
+    }
+
+    @Test
+    public void stackOverflowPrevented() {
+        int size = 10_000;
+        Set<SPLCommit> commits = new HashSet<>();
+        SPLCommit previousCommit = new SPLCommit("0");
+        previousCommit.setParents();
+        commits.add(previousCommit);
+        for (int i = 1; i < size; i++) {
+            SPLCommit commit = new SPLCommit(String.valueOf(i));
+            commit.setParents(previousCommit);
+            commits.add(commit);
+            previousCommit = commit;
+        }
+
+        LongestNonOverlappingSequences algo = new LongestNonOverlappingSequences();
+        var result = algo.extract(commits);
+        assert result.size() == 1;
+        assert result.get(0).size() == size;
+    }
+
+    private boolean sequencesAreEqual(final NonEmptyList<SPLCommit> a, final NonEmptyList<SPLCommit> b) {
+        if (a.size() != b.size()) {
+            return false;
+        }
+        for (int i = 0; i < a.size(); i++) {
+            if (!a.get(i).id().equals(b.get(i).id())) {
+                return false;
+            }
+        }
+        return true;
+    }
+}