Remove redundant semicolons in the cleanup pass.

Remove redundant semicolons at top level, in class and enum bodies, and in statements and blocks during the cleanup pass.

PiperOrigin-RevId: 1004558693

Author: cushon

core/src/main/java/com/google/googlejavaformat/java/CleanupPass.java

@@ -26,7 +26,7 @@ ImmutableList<Replacement> findReplacements(Cleanups cleanups) throws FormatterE
           cleanups.context(), cleanups.unit(), cleanups.contents());
     }
   },
-  /// AST cleanups: missing control-flow braces.
+  /// AST cleanups: missing control-flow braces and redundant semicolons.
   AST_CLEANUPS {
     @Override
     ImmutableList<Replacement> findReplacements(Cleanups cleanups) throws FormatterException {

core/src/main/java/com/google/googlejavaformat/java/CleanupScanner.java

@@ -17,6 +17,10 @@
 import static com.google.common.collect.ImmutableList.toImmutableList;
 import static com.google.googlejavaformat.java.Trees.getEndPosition;
 import static com.google.googlejavaformat.java.Trees.getStartPosition;
+import static com.sun.tools.javac.code.Flags.ENUM;
+import static com.sun.tools.javac.code.Flags.GENERATED_MEMBER;
+import static java.lang.Math.max;
+import static java.lang.Math.min;
 import static java.util.Comparator.comparingInt;
 
 import com.google.common.collect.ImmutableList;
@@ -26,16 +30,24 @@
 import com.google.common.collect.TreeRangeSet;
 import com.google.googlejavaformat.java.JavaInput.Tok;
 import com.sun.source.tree.BlockTree;
+import com.sun.source.tree.CaseTree;
+import com.sun.source.tree.ClassTree;
+import com.sun.source.tree.CompilationUnitTree;
 import com.sun.source.tree.DoWhileLoopTree;
 import com.sun.source.tree.EmptyStatementTree;
 import com.sun.source.tree.EnhancedForLoopTree;
 import com.sun.source.tree.ForLoopTree;
 import com.sun.source.tree.IfTree;
+import com.sun.source.tree.ImportTree;
+import com.sun.source.tree.NewClassTree;
 import com.sun.source.tree.StatementTree;
 import com.sun.source.tree.Tree;
+import com.sun.source.tree.VariableTree;
 import com.sun.source.tree.WhileLoopTree;
 import com.sun.source.util.TreePathScanner;
+import com.sun.tools.javac.tree.JCTree;
 import com.sun.tools.javac.tree.JCTree.JCCompilationUnit;
+import com.sun.tools.javac.tree.TreeInfo;
 import java.util.ArrayList;
 import java.util.Collection;
 import java.util.List;
@@ -44,18 +56,21 @@
 import java.util.TreeMap;
 import org.jspecify.annotations.Nullable;
 
-/// Scans the compilation unit for non-whitespace cleanup opportunities, inserting missing
-/// control-flow braces.
+/// Scans the compilation unit for non-whitespace cleanup opportunities: inserting missing
+/// control-flow braces and removing redundant semicolons.
 final class CleanupScanner extends TreePathScanner<@Nullable Void, @Nullable Void> {
 
+  // Replace with Flags.IMPLICIT_CLASS once JDK 25 is the minimum supported version
+  private static final int IMPLICIT_CLASS = 1 << 19;
+
   static ImmutableList<Replacement> scan(
       JCCompilationUnit unit, String contents, Collection<Range<Integer>> ranges)
       throws FormatterException {
     RangeSet<Integer> targetRanges = TreeRangeSet.create(ranges);
     ImmutableList<Tok> toks = JavaInput.buildToks(contents, ImmutableSet.of());
     RangeSet<Integer> activeDeletions = TreeRangeSet.create();
     NavigableMap<Integer, StringBuilder> activeInsertions = new TreeMap<>();
-    new CleanupScanner(unit, toks, targetRanges, activeDeletions, activeInsertions)
+    new CleanupScanner(contents, unit, toks, targetRanges, activeDeletions, activeInsertions)
         .scan(unit, null);
     if (activeDeletions.isEmpty() && activeInsertions.isEmpty()) {
       return ImmutableList.of();
@@ -108,18 +123,21 @@ private static void verifyCommentsPreserved(
     }
   }
 
+  private final String contents;
   private final JCCompilationUnit unit;
   private final ImmutableList<Tok> toks;
   private final RangeSet<Integer> targetRanges;
   private final RangeSet<Integer> activeDeletions;
   private final NavigableMap<Integer, StringBuilder> activeInsertions;
 
   CleanupScanner(
+      String contents,
       JCCompilationUnit unit,
       ImmutableList<Tok> toks,
       RangeSet<Integer> targetRanges,
       RangeSet<Integer> activeDeletions,
       NavigableMap<Integer, StringBuilder> activeInsertions) {
+    this.contents = contents;
     this.unit = unit;
     this.toks = toks;
     this.targetRanges = targetRanges;
@@ -135,6 +153,154 @@ private void insert(int pos, String text) {
     activeInsertions.computeIfAbsent(pos, k -> new StringBuilder()).append(text);
   }
 
+  private void deleteRedundantSemicolon(int semiPos) {
+    int start = semiPos;
+    while (start > 0 && isHorizontalWhitespace(contents.charAt(start - 1))) {
+      start--;
+    }
+    int lineBegin = start;
+    while (lineBegin > 0 && !isAtLineStart(contents, lineBegin)) {
+      lineBegin--;
+    }
+    int end = semiPos + 1;
+    while (end < contents.length() && isHorizontalWhitespace(contents.charAt(end))) {
+      end++;
+    }
+    boolean nothingBeforeOnLine =
+        start == lineBegin || activeDeletions.encloses(Range.closedOpen(lineBegin, start));
+    if (nothingBeforeOnLine && end < contents.length()) {
+      if (contents.charAt(end) == '\r') {
+        end++;
+        if (end < contents.length() && contents.charAt(end) == '\n') {
+          end++;
+        }
+      } else if (contents.charAt(end) == '\n') {
+        end++;
+      }
+    }
+    activeDeletions.add(Range.closedOpen(start, end));
+  }
+
+  private static boolean isHorizontalWhitespace(char c) {
+    return c == ' ' || c == '\t';
+  }
+
+  private static boolean isAtLineStart(String contents, int pos) {
+    return pos == 0 || contents.charAt(pos - 1) == '\n' || contents.charAt(pos - 1) == '\r';
+  }
+
+  @Override
+  public @Nullable Void visitCompilationUnit(CompilationUnitTree node, @Nullable Void unused) {
+    checkTopLevelSemicolons(node);
+    return super.visitCompilationUnit(node, null);
+  }
+
+  // Top-level semicolons between or after type declarations are represented as EmptyStatementTree
+  // nodes in CompilationUnitTree.getTypeDecls(), but javac omits top-level semicolons from the AST
+  // in some cases:
+  // - In compilation units with no type declarations (e.g. only package and/or import declarations
+  //   followed by extra semicolons), JavacParser drops trailing semicolons at EOF without adding
+  //   them to getTypeDecls().
+  // - In compact source files (implicitly declared classes), JavacParser.constructImplicitClass
+  //   strips all JCSkip (EmptyStatementTree) nodes when constructing the synthetic ClassDef.
+  // Scan token gaps between top-level declarations so all redundant top-level semicolons are
+  // handled uniformly.
+  private void checkTopLevelSemicolons(CompilationUnitTree node) {
+    RangeSet<Integer> validSpans = TreeRangeSet.create();
+    if (node.getPackage() != null) {
+      addNodeSpan(validSpans, node.getPackage());
+    }
+    if (node.getModule() != null) {
+      addNodeSpan(validSpans, node.getModule());
+    }
+    for (ImportTree imp : node.getImports()) {
+      addNodeSpan(validSpans, imp);
+    }
+    for (Tree decl : node.getTypeDecls()) {
+      switch (decl) {
+        case EmptyStatementTree unused -> {}
+        case ClassTree classTree
+            when (TreeInfo.flags((JCTree) classTree) & IMPLICIT_CLASS) == IMPLICIT_CLASS -> {
+          for (Tree member : classTree.getMembers()) {
+            if ((TreeInfo.flags((JCTree) member) & GENERATED_MEMBER) == 0) {
+              addNodeSpan(validSpans, member);
+            }
+          }
+        }
+        default -> addNodeSpan(validSpans, decl);
+      }
+    }
+    deleteSemicolonsInGaps(0, contents.length(), validSpans);
+  }
+
+  private void addNodeSpan(RangeSet<Integer> spans, Tree tree) {
+    spans.add(Range.closedOpen(getStartPosition(tree), getEndPosition(tree, unit)));
+  }
+
+  @Override
+  public @Nullable Void visitClass(ClassTree tree, @Nullable Void unused) {
+    checkClassSemicolons(tree);
+    return super.visitClass(tree, null);
+  }
+
+  private void checkClassSemicolons(ClassTree tree) {
+    if ((TreeInfo.flags((JCTree) tree) & IMPLICIT_CLASS) == IMPLICIT_CLASS
+        || tree.getMembers() == null) {
+      return;
+    }
+    int classStart = getStartPosition(tree);
+    int classEnd = getEndPosition(tree, unit);
+    RangeSet<Integer> memberSpans = TreeRangeSet.create();
+    // Record the spans of constructor arguments in `new MyClass(arg1, arg2) {...}` and enum
+    // constants with class bodies `FOO(arg1, arg2) {...}` (where javac sets the anonymous
+    // ClassTree's start position to the enum constant identifier before the argument list), so
+    // semicolons inside argument expressions (such as lambdas) do not show up as gaps in the
+    // class body.
+    if (getCurrentPath().getParentPath().getLeaf() instanceof NewClassTree newClass) {
+      for (Tree arg : newClass.getArguments()) {
+        addNodeSpan(memberSpans, arg);
+      }
+    }
+    int lastEnumConstantEnd = classStart;
+    int firstNonEnumMemberStart = classEnd;
+    for (Tree member : tree.getMembers()) {
+      long flags = TreeInfo.flags((JCTree) member);
+      if ((flags & GENERATED_MEMBER) != 0) {
+        continue;
+      }
+      int mStart = getStartPosition(member);
+      int mEnd = getEndPosition(member, unit);
+      memberSpans.add(Range.closedOpen(mStart, mEnd));
+      if (tree.getKind() == Tree.Kind.ENUM) {
+        if (member instanceof VariableTree && (flags & ENUM) != 0) {
+          lastEnumConstantEnd = max(lastEnumConstantEnd, mEnd);
+        } else {
+          firstNonEnumMemberStart = min(firstNonEnumMemberStart, mStart);
+        }
+      }
+    }
+    if (tree.getKind() == Tree.Kind.ENUM && !memberSpans.isEmpty()) {
+      Tok enumSemi = findTokenAtOrAfter(lastEnumConstantEnd, firstNonEnumMemberStart, ";");
+      if (enumSemi != null) {
+        memberSpans.add(
+            Range.closedOpen(enumSemi.getPosition(), enumSemi.getPosition() + enumSemi.length()));
+      }
+    }
+    deleteSemicolonsInGaps(classStart, classEnd, memberSpans);
+  }
+
+  private void deleteSemicolonsInGaps(int start, int end, RangeSet<Integer> validSpans) {
+    for (Range<Integer> gap :
+        validSpans.complement().subRangeSet(Range.closedOpen(start, end)).asRanges()) {
+      for (Tok semi : semicolonTokensInRange(gap.lowerEndpoint(), gap.upperEndpoint())) {
+        int pos = semi.getPosition();
+        if (inRange(pos, pos + 1)) {
+          deleteRedundantSemicolon(pos);
+        }
+      }
+    }
+  }
+
   @Override
   public @Nullable Void visitIf(IfTree tree, @Nullable Void unused) {
     super.visitIf(tree, null);
@@ -214,6 +380,20 @@ private int closingBraceInsertionPos(StatementTree stmt) {
     return stmtEnd;
   }
 
+  @Override
+  public @Nullable Void visitEmptyStatement(EmptyStatementTree tree, @Nullable Void unused) {
+    // Include CaseTree for empty statements in colon-style switch statement groups
+    // (e.g. `switch (x) { case FOO: ; case BAR: }`).
+    Tree parent = getCurrentPath().getParentPath().getLeaf();
+    if (parent instanceof BlockTree || parent instanceof CaseTree) {
+      int pos = getStartPosition(tree);
+      if (inRange(pos, pos + 1)) {
+        deleteRedundantSemicolon(pos);
+      }
+    }
+    return null;
+  }
+
   // Hand-rolled lower-bound binary search to avoid allocating a dummy Tok or wrapper list for
   // Collections.binarySearch.
   private int tokIndexAtOrAfter(int pos) {
@@ -239,4 +419,31 @@ private int tokIndexAtOrAfter(int pos) {
     }
     return null;
   }
+
+  private @Nullable Tok findTokenAtOrAfter(int startPos, int endPos, String text) {
+    for (int i = tokIndexAtOrAfter(startPos); i < toks.size(); i++) {
+      Tok tok = toks.get(i);
+      if (tok.getPosition() >= endPos) {
+        break;
+      }
+      if (tok.isToken() && tok.getText().equals(text)) {
+        return tok;
+      }
+    }
+    return null;
+  }
+
+  private List<Tok> semicolonTokensInRange(int startPos, int endPos) {
+    List<Tok> result = new ArrayList<>();
+    for (int i = tokIndexAtOrAfter(startPos); i < toks.size(); i++) {
+      Tok tok = toks.get(i);
+      if (tok.getPosition() >= endPos) {
+        break;
+      }
+      if (tok.isToken() && tok.getText().equals(";")) {
+        result.add(tok);
+      }
+    }
+    return result;
+  }
 }

core/src/main/java/com/google/googlejavaformat/java/UsageException.java

@@ -57,7 +57,7 @@ Maximum line length (default is 100).
   --skip-reordering-modifiers
     Do not reorder modifiers into the JLS-recommended order.
   --cleanup[=(true|false)]
-    Enables non-whitespace cleanups, such as inserting missing braces (default is true).
+    Enables non-whitespace cleanups, such as inserting missing braces and removing redundant semicolons (default is true).
   --dry-run, -n
     Prints the paths of the files whose contents would change if the formatter were run normally.
   --set-exit-if-changed