maedhroz commented on code in PR #2267: URL: https://github.com/apache/cassandra/pull/2267#discussion_r1185290014
########## src/java/org/apache/cassandra/io/tries/IncrementalDeepTrieWriterPageAware.java: ########## @@ -0,0 +1,405 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.cassandra.io.tries; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Iterator; +import java.util.List; + +import javax.annotation.concurrent.NotThreadSafe; + +import org.apache.cassandra.io.util.DataOutputPlus; + +/** + * This class is a variant of {@link IncrementalTrieWriterPageAware} which is able to build even very deep + * tries. While the parent class uses recursion for clarity, it may end up with stack overflow for tries with + * very long keys. This implementation can switch processing from stack to heap at a certain depth (provided + * as a constructor param). + * <p> + * This class intentionally repeats code present in the parent class, both in the in-stack and on-heap versions + * of each of the three implemented recursive operations. Removing this repetition can cause higher stack usage + * and thus stack overflow failures. + */ +@NotThreadSafe +public class IncrementalDeepTrieWriterPageAware<VALUE> extends IncrementalTrieWriterPageAware<VALUE> +{ + private final int maxRecursionDepth; + + public IncrementalDeepTrieWriterPageAware(TrieSerializer<VALUE, ? super DataOutputPlus> trieSerializer, DataOutputPlus dest, int maxRecursionDepth) + { + super(trieSerializer, dest); + this.maxRecursionDepth = maxRecursionDepth; + } + + public IncrementalDeepTrieWriterPageAware(TrieSerializer<VALUE, ? super DataOutputPlus> trieSerializer, DataOutputPlus dest) + { + this(trieSerializer, dest, 64); + } + + /** + * Simple framework for executing recursion using on-heap linked trace to avoid stack overruns. + */ + static abstract class Recursion<NODE> + { + final Recursion<NODE> parent; + final NODE node; + final Iterator<NODE> childIterator; + + Recursion(NODE node, Iterator<NODE> childIterator, Recursion<NODE> parent) + { + this.parent = parent; + this.node = node; + this.childIterator = childIterator; + } + + /** + * Make a child Recursion object for the given node and initialize it as necessary to continue processing + * with it. + * <p> + * May return null if the recursion does not need to continue inside the child branch. + */ + abstract Recursion<NODE> makeChild(NODE child); + + /** + * Complete the processing this Recursion object. + * <p> + * Note: this method is not called for the nodes for which makeChild() returns null. + */ + abstract void complete() throws IOException; + + /** + * Complete processing of the given child (possibly retrieve data to apply to any accumulation performed + * in this Recursion object). + * <p> + * This is called when processing a child completes, including when recursion inside the child branch + * is skipped by makeChild() returning null. + */ + void completeChild(NODE child) + {} + + /** + * Recursive process, in depth-first order, the branch rooted at this recursion node. + * <p> + * Returns this. + */ + Recursion<NODE> process() throws IOException + { + Recursion<NODE> curr = this; + + while (true) + { + if (curr.childIterator.hasNext()) + { + NODE child = curr.childIterator.next(); + Recursion<NODE> childRec = curr.makeChild(child); + if (childRec != null) + curr = childRec; + else + curr.completeChild(child); + } + else + { + curr.complete(); + Recursion<NODE> currParent = curr.parent; + if (currParent == null) + return curr; + currParent.completeChild(curr.node); + curr = currParent; + } + } + } + } + + @Override + protected int recalcTotalSize(Node<VALUE> node, long nodePosition) throws IOException + { + return recalcTotalSizeRecursiveOnStack(node, nodePosition, 0); + } + + private int recalcTotalSizeRecursiveOnStack(Node<VALUE> node, long nodePosition, int depth) throws IOException + { + if (node.hasOutOfPageInBranch) + { + int sz = 0; + for (Node<VALUE> child : node.children) + { + if (depth < maxRecursionDepth) + sz += recalcTotalSizeRecursiveOnStack(child, nodePosition + sz, depth + 1); + else + sz += recalcTotalSizeRecursiveOnHeap(child, nodePosition + sz); + } + node.branchSize = sz; + } + + // The sizing below will use the branch size calculated above. Since that can change on out-of-page in branch, + // we need to recalculate the size if either flag is set. + if (node.hasOutOfPageChildren || node.hasOutOfPageInBranch) + node.nodeSize = serializer.sizeofNode(node, nodePosition + node.branchSize); + + return node.branchSize + node.nodeSize; + } + + private int recalcTotalSizeRecursiveOnHeap(Node<VALUE> node, long nodePosition) throws IOException + { + if (node.hasOutOfPageInBranch) + new RecalcTotalSizeRecursion(node, null, nodePosition).process(); + + if (node.hasOutOfPageChildren || node.hasOutOfPageInBranch) + node.nodeSize = serializer.sizeofNode(node, nodePosition + node.branchSize); + + return node.branchSize + node.nodeSize; + } + + class RecalcTotalSizeRecursion extends Recursion<Node<VALUE>> + { + final long nodePosition; + int sz; + + RecalcTotalSizeRecursion(Node<VALUE> node, Recursion<Node<VALUE>> parent, long nodePosition) + { + super(node, node.children.iterator(), parent); + sz = 0; + this.nodePosition = nodePosition; + } + + @Override + Recursion<Node<VALUE>> makeChild(Node<VALUE> child) + { + if (child.hasOutOfPageInBranch) + return new RecalcTotalSizeRecursion(child, this, nodePosition + sz); + else + return null; + } + + @Override + void complete() + { + node.branchSize = sz; + } + + @Override + void completeChild(Node<VALUE> child) + { + // This will be called for nodes that were recursively processed as well as the ones that weren't. + + // The sizing below will use the branch size calculated above. Since that can change on out-of-page in branch, + // we need to recalculate the size if either flag is set. + if (child.hasOutOfPageChildren || child.hasOutOfPageInBranch) + { + long childPosition = this.nodePosition + sz; + child.nodeSize = serializer.sizeofNode(child, childPosition + child.branchSize); + } + + sz += child.branchSize + child.nodeSize; + } + } + + @Override + protected long write(Node<VALUE> node) throws IOException + { + return writeRecursiveOnStack(node, 0); + } + + private long writeRecursiveOnStack(Node<VALUE> node, int depth) throws IOException + { + long nodePosition = dest.position(); + for (Node<VALUE> child : node.children) + if (child.filePos == -1) + { + if (depth < maxRecursionDepth) + child.filePos = writeRecursiveOnStack(child, depth + 1); + else + child.filePos = writeRecursiveOnHeap(child); + } + + nodePosition += node.branchSize; + assert dest.position() == nodePosition + : "Expected node position to be " + nodePosition + " but got " + dest.position() + " after writing children.\n" + dumpNode(node, dest.position()); + + serializer.write(dest, node, nodePosition); + + assert dest.position() == nodePosition + node.nodeSize + || dest.paddedPosition() == dest.position() // For PartitionIndexTest.testPointerGrowth where position may jump on page boundaries. + : "Expected node position to be " + (nodePosition + node.nodeSize) + " but got " + dest.position() + " after writing node, nodeSize " + node.nodeSize + ".\n" + dumpNode(node, nodePosition); + return nodePosition; + } + + private long writeRecursiveOnHeap(Node<VALUE> node) throws IOException + { + return new WriteRecursion(node, null).process().node.filePos; + } + + class WriteRecursion extends Recursion<Node<VALUE>> + { + long nodePosition; + + WriteRecursion(Node<VALUE> node, Recursion<Node<VALUE>> parent) + { + super(node, node.children.iterator(), parent); + nodePosition = dest.position(); + } + + @Override + Recursion<Node<VALUE>> makeChild(Node<VALUE> child) + { + if (child.filePos == -1) + return new WriteRecursion(child, this); + else + return null; + } + + @Override + void complete() throws IOException + { + nodePosition = nodePosition + node.branchSize; + assert dest.position() == nodePosition + : "Expected node position to be " + nodePosition + " but got " + dest.position() + " after writing children.\n" + dumpNode(node, dest.position()); + + serializer.write(dest, node, nodePosition); + + assert dest.position() == nodePosition + node.nodeSize + || dest.paddedPosition() == dest.position() // For PartitionIndexTest.testPointerGrowth where position may jump on page boundaries. + : "Expected node position to be " + (nodePosition + node.nodeSize) + " but got " + dest.position() + " after writing node, nodeSize " + node.nodeSize + ".\n" + dumpNode(node, nodePosition); + + node.filePos = nodePosition; + } + } + + @Override + protected long writePartial(Node<VALUE> node, DataOutputPlus dest, long baseOffset) throws IOException + { + return writePartialRecursiveOnStack(node, dest, baseOffset, 0); + } + + private long writePartialRecursiveOnStack(Node<VALUE> node, DataOutputPlus dest, long baseOffset, int depth) throws IOException Review Comment: nit: Seems like the duplication is intended, so maybe throw a `@SuppressWarnings("DuplicatedCode")` on there. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
