[MOD] Sequence/Array builders: optimizations

Author: ChristianGruen

basex-core/src/main/java/org/basex/query/expr/List.java

@@ -201,13 +201,20 @@ private Iter iter(final int i) throws QueryException {
 
   @Override
   public Value value(final QueryContext qc) throws QueryException {
-    // special case: concatenate two sequences
-    if(exprs.length == 2) return exprs[0].value(qc).append(exprs[1].value(qc), qc);
-
-    // general case: concatenate all sequences (unknown size: do not create compact data structures)
-    final ValueBuilder vb = new ValueBuilder(qc, size() != -1 ? size() : Integer.MIN_VALUE);
-    for(final Expr expr : exprs) vb.add(expr.value(qc));
-    return vb.value(this);
+    // known result size: use builders with optimized data structures
+    final long size = size();
+    if(size != -1) {
+      final ValueBuilder vb = new ValueBuilder(qc, size);
+      for(final Expr expr : exprs) vb.add(expr.value(qc));
+      return vb.value(this);
+    }
+    // otherwise, append values (uses tree structure)
+    Value value = null;
+    for(final Expr expr : exprs) {
+      final Value v = expr.value(qc);
+      value = value != null ? value.append(v, qc) : v;
+    }
+    return value;
   }
 
   @Override

basex-core/src/main/java/org/basex/query/func/array/ArrayPut.java

@@ -19,7 +19,7 @@ public XQArray item(final QueryContext qc, final InputInfo ii) throws QueryExcep
     final XQArray array = toArray(arg(0), qc);
     final long position = toLong(arg(1), qc);
     final Value member = arg(2).value(qc);
-    return array.put(toPos(array, position, false), member);
+    return array.putMember(toPos(array, position, false), member, qc);
   }
 
   @Override

basex-core/src/main/java/org/basex/query/func/fn/FnInsertSeparator.java

@@ -21,7 +21,7 @@ public final Value value(final QueryContext qc) throws QueryException {
     final Iter input = arg(0).iter(qc);
     final Value separator = arg(1).value(qc);
 
-    final ValueBuilder vb = new ValueBuilder(qc);
+    final ValueBuilder vb = new ValueBuilder(qc, size());
     Item item = input.next();
     if(item != null) {
       vb.add(item);

basex-core/src/main/java/org/basex/query/func/fn/PlanFn.java

@@ -470,7 +470,7 @@ private XQMap record(final ANode node, final Plan plan, final QueryContext qc)
   private XQArray mixed(final ANode node, final ANode parent, final Plan plan,
       final QueryContext qc, final boolean ignoreEmpty) throws QueryException {
 
-    final ArrayBuilder ab = new ArrayBuilder();
+    final ArrayBuilder ab = new ArrayBuilder(qc);
     for(final ANode attr : children(NodeType.ATTRIBUTE, node)) {
       ab.add(new MapBuilder().put(nodeName(attr, node, plan, qc), attr.string()).map());
     }

basex-core/src/main/java/org/basex/query/value/Value.java

@@ -165,7 +165,8 @@ public final Value insert(final long pos, final Value value, final QueryContext
 
   /**
    * Removes an item at the given position.
-   * @param pos deletion position, must be between 0 and {@link #size()} - 1
+   * @param pos deletion position, must be greater than 0 and smaller than {@link #size()} - 1
+   *   (use {@link #subSeq(long, long, QueryContext)} to remove first or last item)
    * @param qc query context
    * @return new sequence
    */

basex-core/src/main/java/org/basex/query/value/ValueBuilder.java

@@ -2,6 +2,7 @@
 
 import org.basex.query.*;
 import org.basex.query.expr.*;
+import org.basex.query.util.list.*;
 import org.basex.query.value.item.*;
 import org.basex.query.value.seq.*;
 import org.basex.query.value.seq.tree.*;
@@ -19,9 +20,9 @@ public final class ValueBuilder {
   /** QueryContext. */
   private final QueryContext qc;
 
-  /** Sequence builder, only instantiated if there are at least two items. */
-  private SeqBuilder sequence;
-  /** Capacity ({@link Integer#MIN_VALUE}: create no compact data structures). */
+  /** Builder, only instantiated if there are at least two items. */
+  private SeqBuilder builder;
+  /** Capacity ({@link Long#MIN_VALUE}: create no compact data structures). */
   private final long capacity;
   /** The first added value is cached. */
   private Value single;
@@ -37,7 +38,7 @@ public ValueBuilder(final QueryContext qc) {
   /**
    * Constructor.
    * @param qc query context (required for interrupting running queries)
-   * @param capacity initial capacity ({@link Integer#MIN_VALUE}: create no compact data structures)
+   * @param capacity initial capacity ({@link Long#MIN_VALUE}: create no compact data structures)
    */
   public ValueBuilder(final QueryContext qc, final long capacity) {
     this.qc = qc;
@@ -52,24 +53,23 @@ public ValueBuilder(final QueryContext qc, final long capacity) {
   public ValueBuilder add(final Value value) {
     if(!value.isEmpty()) {
       qc.checkStop();
-      final Value sngl = single;
-      if(sngl != null) {
-        single = null;
-        if(capacity != Integer.MIN_VALUE) {
-          sequence = isStr(sngl) && isStr(value) ? new StrSeqBuilder() :
-                     isAtm(sngl) && isAtm(value) ? new AtmSeqBuilder() :
-                     isInt(sngl) && isInt(value) ? new IntSeqBuilder() :
-                     isDbl(sngl) && isDbl(value) ? new DblSeqBuilder() :
-                     isBln(sngl) && isBln(value) ? new BlnSeqBuilder() : null;
+      if(builder == null) {
+        final Value sngl = single;
+        if(sngl == null) {
+          single = value;
+          return this;
         }
-        if(sequence == null) sequence = new TreeSeqBuilder();
+        single = null;
+        builder = capacity == Long.MIN_VALUE ? new TreeSeqBuilder() :
+          isStr(sngl) && isStr(value) ? new StrSeqBuilder() :
+          isAtm(sngl) && isAtm(value) ? new AtmSeqBuilder() :
+          isInt(sngl) && isInt(value) ? new IntSeqBuilder() :
+          isDbl(sngl) && isDbl(value) ? new DblSeqBuilder() :
+          isBln(sngl) && isBln(value) ? new BlnSeqBuilder() :
+          new ItemSeqBuilder();
         add(sngl);
       }
-      if(sequence != null) {
-        sequence = sequence.add(value, qc);
-      } else {
-        single = value;
-      }
+      builder = builder.add(value, qc);
     }
     return this;
   }
@@ -90,9 +90,9 @@ public Value value() {
    */
   public Value value(final Type type) {
     try {
-      return sequence != null ? sequence.value(type) : single != null ? single : Empty.VALUE;
+      return builder != null ? builder.value(type) : single != null ? single : Empty.VALUE;
     } finally {
-      sequence = null;
+      builder = null;
       single = null;
     }
   }
@@ -154,12 +154,12 @@ static boolean isBln(final Value value) {
 
   @Override
   public String toString() {
-    return Util.className(this) + '[' + Util.className(sequence != null ? sequence :
+    return Util.className(this) + '[' + Util.className(builder != null ? builder :
       single != null ? single : Empty.VALUE) + ']';
   }
 
   /** String sequence builder. */
-  final class StrSeqBuilder implements SeqBuilder {
+  final class StrSeqBuilder extends SeqBuilder {
     /** Values. */
     private final TokenList values = new TokenList(capacity);
 
@@ -179,7 +179,7 @@ public Value value(final Type type) {
   }
 
   /** Untyped atomic sequence builder. */
-  final class AtmSeqBuilder implements SeqBuilder {
+  final class AtmSeqBuilder extends SeqBuilder {
     /** Values. */
     private final TokenList values = new TokenList(capacity);
 
@@ -199,7 +199,7 @@ public Value value(final Type type) {
   }
 
   /** Integer sequence builder. */
-  final class IntSeqBuilder implements SeqBuilder {
+  final class IntSeqBuilder extends SeqBuilder {
     /** Values. */
     private final IntList values = new IntList(capacity);
 
@@ -222,7 +222,7 @@ public Value value(final Type type) {
   }
 
   /** Double sequence builder. */
-  final class DblSeqBuilder implements SeqBuilder {
+  final class DblSeqBuilder extends SeqBuilder {
     /** Values. */
     private final DoubleList values = new DoubleList(capacity);
 
@@ -242,7 +242,7 @@ public Value value(final Type type) {
   }
 
   /** Boolean sequence builder. */
-  final class BlnSeqBuilder implements SeqBuilder {
+  final class BlnSeqBuilder extends SeqBuilder {
     /** Values. */
     private final BoolList values = new BoolList(capacity);
 
@@ -260,4 +260,21 @@ public Value value(final Type type) {
       return BlnSeq.get(values.finish());
     }
   }
+
+  /** Item sequence builder. */
+  final class ItemSeqBuilder extends SeqBuilder {
+    /** Items. */
+    private final ItemList items = new ItemList(capacity);
+
+    @Override
+    public SeqBuilder add(final Item item) {
+      items.add(item);
+      return this;
+    }
+
+    @Override
+    public Value value(final Type type) {
+      return items.value(type);
+    }
+  }
 }

basex-core/src/main/java/org/basex/query/value/array/ArrBuilder.java

@@ -4,23 +4,23 @@
 import org.basex.query.value.type.*;
 
 /**
- * An interface for creating sequences.
+ * Abstract array builder.
  *
  * @author BaseX Team, BSD License
  * @author Christian Gruen
  */
-public interface ArrBuilder {
+public abstract class ArrBuilder {
   /**
-   * Adds a member to the end of the sequence.
-   * @param value value to add
+   * Appends a member.
+   * @param value value to append
    * @return reference to this builder for convenience
    */
-  ArrBuilder add(Value value);
+  abstract ArrBuilder add(Value value);
 
   /**
    * Creates an array containing the values of this builder.
    * @param type type of all members in this sequence
    * @return resulting sequence
    */
-  XQArray array(ArrayType type);
+  abstract XQArray array(ArrayType type);
 }

basex-core/src/main/java/org/basex/query/value/array/ArrayBuilder.java

@@ -12,39 +12,32 @@
  * @author Christian Gruen
  */
 public final class ArrayBuilder {
-  /** QueryContext (can be {@code null}). */
+  /** QueryContext. */
   private final QueryContext qc;
 
-  /** Sequence builder, only instantiated if there are at least two members. */
-  private ArrBuilder array;
-  /** Capacity ({@link Integer#MIN_VALUE}: create no compact data structures). */
+  /** Builder, only instantiated if there are at least two items. */
+  private ArrBuilder builder;
+  /** Capacity ({@link Long#MIN_VALUE}: create no compact data structures). */
   private final long capacity;
   /** The first added value is cached. */
   private Value single;
 
   /**
    * Constructor.
-   */
-  public ArrayBuilder() {
-    this(null);
-  }
-
-  /**
-   * Constructor.
-   * @param qc query context (can be {@code null}, required for interrupting running queries)
+   * @param qc query context (required for interrupting running queries)
    */
   public ArrayBuilder(final QueryContext qc) {
     this(qc, -1);
   }
 
   /**
    * Constructor.
-   * @param qc query context (can be {@code null}, required for interrupting running queries)
-   * @param capacity initial capacity ({@link Integer#MIN_VALUE}: create no compact data structures)
+   * @param qc query context (required for interrupting running queries)
+   * @param capacity initial capacity ({@link Long#MIN_VALUE}: create no compact data structures)
    */
   public ArrayBuilder(final QueryContext qc, final long capacity) {
     this.qc = qc;
-    this.capacity = qc != null ? capacity : Integer.MIN_VALUE;
+    this.capacity = capacity;
   }
 
   /**
@@ -53,18 +46,19 @@ public ArrayBuilder(final QueryContext qc, final long capacity) {
    * @return reference to this builder for convenience
    */
   public ArrayBuilder add(final Value value) {
-    final Value sngl = single;
-    if(sngl != null) {
+    if(builder == null) {
+      qc.checkStop();
+      final Value sngl = single;
+      if(sngl == null) {
+        single = value;
+        return this;
+      }
       single = null;
-      array = capacity != Integer.MIN_VALUE && sngl.size() == 1 && value.size() == 1 ?
-        new ItemArrayBuilder(qc, capacity) : new TreeArrayBuilder();
+      builder = capacity == Long.MIN_VALUE || sngl.size() != 1 || value.size() != 1 ?
+        new TreeArrayBuilder() : new ItemArrayBuilder();
       add(sngl);
     }
-    if(array != null) {
-      array = array.add(value);
-    } else {
-      single = value;
-    }
+    builder = builder.add(value);
     return this;
   }
 
@@ -84,10 +78,10 @@ public XQArray array() {
    */
   public XQArray array(final ArrayType type) {
     try {
-      return array != null ? array.array(type) : single != null ? XQArray.get(single) :
+      return builder != null ? builder.array(type) : single != null ? XQArray.get(single) :
         XQArray.empty();
     } finally {
-      array = null;
+      builder = null;
       single = null;
     }
   }
@@ -100,4 +94,26 @@ public XQArray array(final ArrayType type) {
   public XQArray array(final Expr expr) {
     return expr != null ? array((ArrayType) expr.seqType().type) : array();
   }
+
+  /** Item array builder. */
+  final class ItemArrayBuilder extends ArrBuilder {
+    /** Value builder. */
+    private final ValueBuilder vb = new ValueBuilder(qc, capacity);
+
+    @Override
+    public ArrBuilder add(final Value value) {
+      if(value.size() == 1) {
+        vb.add(value);
+        return this;
+      }
+      final TreeArrayBuilder ab = new TreeArrayBuilder();
+      for(final Value member : vb.value()) ab.add(member);
+      return ab.add(value);
+    }
+
+    @Override
+    public XQArray array(final ArrayType type) {
+      return new ItemArray(vb.value(type.valueType().type));
+    }
+  }
 }

basex-core/src/main/java/org/basex/query/value/array/BigArray.java

@@ -68,7 +68,8 @@ public long structSize() {
   }
 
   @Override
-  public XQArray put(final long pos, final Value value) {
+  public XQArray putMember(final long pos, final Value value, final QueryContext qc) {
+    qc.checkStop();
     final Type tp = union(value);
     long p = pos;
     final int ll = left.length;

basex-core/src/main/java/org/basex/query/value/array/EmptyArray.java

@@ -38,7 +38,7 @@ public long structSize() {
   }
 
   @Override
-  public XQArray put(final long pos, final Value value) {
+  public XQArray putMember(final long pos, final Value value, final QueryContext qc) {
     throw Util.notExpected();
   }