All Downloads are FREE. Search and download functionalities are using the official Maven repository.

io.trino.operator.join.DefaultPagesHash Maven / Gradle / Ivy

/*
 * Licensed 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 io.trino.operator.join;

import io.airlift.units.DataSize;
import io.trino.operator.HashArraySizeSupplier;
import io.trino.operator.PagesHashStrategy;
import io.trino.spi.Page;
import io.trino.spi.PageBuilder;
import io.trino.spi.block.Block;
import it.unimi.dsi.fastutil.longs.LongArrayList;
import it.unimi.dsi.fastutil.objects.ObjectArrayList;

import java.util.Arrays;
import java.util.List;

import static io.airlift.slice.SizeOf.instanceSize;
import static io.airlift.slice.SizeOf.sizeOf;
import static io.airlift.slice.SizeOf.sizeOfByteArray;
import static io.airlift.slice.SizeOf.sizeOfIntArray;
import static io.airlift.units.DataSize.Unit.KILOBYTE;
import static io.trino.operator.SyntheticAddress.decodePosition;
import static io.trino.operator.SyntheticAddress.decodeSliceIndex;
import static io.trino.operator.join.PagesHash.getHashPosition;
import static java.lang.Math.toIntExact;
import static java.util.Objects.requireNonNull;

/**
 * The PagesHash object that handles all cases - single/multi channel joins
 * with any types.
 * This implementation assumes arrays used in the hash are always a power of 2
 */
public final class DefaultPagesHash
        implements PagesHash
{
    private static final int INSTANCE_SIZE = instanceSize(DefaultPagesHash.class);
    private static final DataSize CACHE_SIZE = DataSize.of(128, KILOBYTE);
    private final LongArrayList addresses;
    private final PagesHashStrategy pagesHashStrategy;

    private final int mask;
    private final int[] keys;
    private final long size;

    // Native array of hashes for faster collisions resolution compared
    // to accessing values in blocks. We use bytes to reduce memory foot print
    // and there is no performance gain from storing full hashes
    private final byte[] positionToHashes;

    public DefaultPagesHash(
            LongArrayList addresses,
            PagesHashStrategy pagesHashStrategy,
            PositionLinks.FactoryBuilder positionLinks,
            HashArraySizeSupplier hashArraySizeSupplier)
    {
        this.addresses = requireNonNull(addresses, "addresses is null");
        this.pagesHashStrategy = requireNonNull(pagesHashStrategy, "pagesHashStrategy is null");

        // reserve memory for the arrays
        int hashSize = hashArraySizeSupplier.getHashArraySize(addresses.size());

        mask = hashSize - 1;
        keys = new int[hashSize];
        Arrays.fill(keys, -1);

        positionToHashes = new byte[addresses.size()];

        // We will process addresses in batches, to save memory on array of hashes and improve memory locality.
        int positionsInStep = Math.min(addresses.size() + 1, (int) CACHE_SIZE.toBytes() / Integer.SIZE);
        long[] positionToFullHashes = new long[positionsInStep];

        for (int step = 0; step * positionsInStep <= addresses.size(); step++) {
            int stepBeginPosition = step * positionsInStep;
            int stepEndPosition = Math.min((step + 1) * positionsInStep, addresses.size());
            int stepSize = stepEndPosition - stepBeginPosition;

            // First extract all hashes from blocks to native array.
            // Somehow having this as a separate loop is much faster compared
            // to extracting hashes on the fly in the loop below.
            extractHashes(positionToFullHashes, stepBeginPosition, stepSize);

            // index pages
            indexPages(positionLinks, positionToFullHashes, stepBeginPosition, stepSize);
        }

        size = sizeOf(addresses.elements()) + pagesHashStrategy.getSizeInBytes() +
                sizeOf(keys) + sizeOf(positionToHashes);
    }

    private void extractHashes(long[] positionToFullHashes, int stepBeginPosition, int stepSize)
    {
        for (int batchIndex = 0; batchIndex < stepSize; batchIndex++) {
            int addressIndex = batchIndex + stepBeginPosition;
            long hash = readHashPosition(addressIndex);
            positionToFullHashes[batchIndex] = hash;
            positionToHashes[addressIndex] = (byte) hash;
        }
    }

    private void indexPages(PositionLinks.FactoryBuilder positionLinks, long[] positionToFullHashes, int stepBeginPosition, int stepSize)
    {
        for (int position = 0; position < stepSize; position++) {
            int realPosition = position + stepBeginPosition;
            if (isPositionNull(realPosition)) {
                continue;
            }

            long hash = positionToFullHashes[position];
            int pos = getHashPosition(hash, mask);

            insertValue(positionLinks, realPosition, (byte) hash, pos);
        }
    }

    private void insertValue(PositionLinks.FactoryBuilder positionLinks, int realPosition, byte hash, int pos)
    {
        // look for an empty slot or a slot containing this key
        while (keys[pos] != -1) {
            int currentKey = keys[pos];
            if (hash == positionToHashes[currentKey] && positionEqualsPositionIgnoreNulls(currentKey, realPosition)) {
                // found a slot for this key
                // link the new key position to the current key position
                realPosition = positionLinks.link(realPosition, currentKey);

                // key[pos] updated outside of this loop
                break;
            }
            // increment position and mask to handler wrap around
            pos = (pos + 1) & mask;
        }

        keys[pos] = realPosition;
    }

    @Override
    public int getPositionCount()
    {
        return addresses.size();
    }

    @Override
    public long getInMemorySizeInBytes()
    {
        return INSTANCE_SIZE + size;
    }

    @Override
    public int getAddressIndex(int position, Page hashChannelsPage)
    {
        return getAddressIndex(position, hashChannelsPage, pagesHashStrategy.hashRow(position, hashChannelsPage));
    }

    @Override
    public int getAddressIndex(int rightPosition, Page hashChannelsPage, long rawHash)
    {
        int pos = getHashPosition(rawHash, mask);

        while (keys[pos] != -1) {
            if (positionEqualsCurrentRowIgnoreNulls(keys[pos], (byte) rawHash, rightPosition, hashChannelsPage)) {
                return keys[pos];
            }
            // increment position and mask to handler wrap around
            pos = (pos + 1) & mask;
        }
        return -1;
    }

    @Override
    public int[] getAddressIndex(int[] positions, Page hashChannelsPage)
    {
        if (positions.length == 0) {
            return new int[0];
        }
        long[] hashes = new long[positions[positions.length - 1] + 1];
        for (int i = 0; i < positions.length; i++) {
            hashes[positions[i]] = pagesHashStrategy.hashRow(positions[i], hashChannelsPage);
        }

        return getAddressIndex(positions, hashChannelsPage, hashes);
    }

    @Override
    public int[] getAddressIndex(int[] positions, Page hashChannelsPage, long[] rawHashes)
    {
        int positionCount = positions.length;
        int[] hashPositions = calculateHashPositions(positions, rawHashes, positionCount);

        int[] found = new int[positionCount];
        int foundCount = 0;
        int[] result = new int[positionCount];
        Arrays.fill(result, -1);
        int[] foundKeys = new int[positionCount];

        // Search for positions in the hash array. This is the most CPU-consuming part as
        // it relies on random memory accesses
        findPositions(positionCount, hashPositions, foundKeys);
        // Found positions are put into `found` array
        for (int i = 0; i < positionCount; i++) {
            if (foundKeys[i] != -1) {
                found[foundCount++] = i;
            }
        }

        // At this step we determine if the found keys were indeed the proper ones or it is a hash collision.
        // The result array is updated for the found ones, while the collisions land into `remaining` array.
        int remainingCount = checkFoundPositions(positions, hashChannelsPage, rawHashes, found, foundCount, result, foundKeys);
        int[] remaining = found; // Rename for readability

        // At this point for any reasoable load factor of a hash array (< .75), there is no more than
        // 10 - 15% of positions left. We search for them in a sequential order and update the result array.
        findRemainingPositions(positions, hashChannelsPage, rawHashes, hashPositions, result, remainingCount, remaining);

        return result;
    }

    private void findRemainingPositions(int[] positions, Page hashChannelsPage, long[] rawHashes, int[] hashPositions, int[] result, int remainingCount, int[] remaining)
    {
        for (int i = 0; i < remainingCount; i++) {
            int index = remaining[i];
            int position = (hashPositions[index] + 1) & mask; // hashPositions[index] position has already been checked

            while (keys[position] != -1) {
                if (positionEqualsCurrentRowIgnoreNulls(keys[position], (byte) rawHashes[positions[index]], positions[index], hashChannelsPage)) {
                    result[index] = keys[position];
                    break;
                }
                // increment position and mask to handler wrap around
                position = (position + 1) & mask;
            }
        }
    }

    private int checkFoundPositions(
            int[] positions,
            Page hashChannelsPage,
            long[] rawHashes,
            int[] found,
            int foundCount,
            int[] result,
            int[] foundKeys)
    {
        int[] remaining = found; // Rename for readability
        int remainingCount = 0;
        for (int i = 0; i < foundCount; i++) {
            int index = found[i];
            if (positionEqualsCurrentRowIgnoreNulls(foundKeys[index], (byte) rawHashes[positions[index]], positions[index], hashChannelsPage)) {
                result[index] = foundKeys[index];
            }
            else {
                remaining[remainingCount++] = index;
            }
        }
        return remainingCount;
    }

    private void findPositions(int positionCount, int[] hashPositions, int[] foundKeys)
    {
        for (int i = 0; i < positionCount; i++) {
            foundKeys[i] = keys[hashPositions[i]];
        }
    }

    private int[] calculateHashPositions(int[] positions, long[] rawHashes, int positionCount)
    {
        int[] hashPositions = new int[positionCount];

        for (int i = 0; i < positionCount; i++) {
            hashPositions[i] = getHashPosition(rawHashes[positions[i]], mask);
        }
        return hashPositions;
    }

    @Override
    public void appendTo(long position, PageBuilder pageBuilder, int outputChannelOffset)
    {
        long pageAddress = addresses.getLong(toIntExact(position));
        int blockIndex = decodeSliceIndex(pageAddress);
        int blockPosition = decodePosition(pageAddress);

        pagesHashStrategy.appendTo(blockIndex, blockPosition, pageBuilder, outputChannelOffset);
    }

    private boolean isPositionNull(int position)
    {
        long pageAddress = addresses.getLong(position);
        int blockIndex = decodeSliceIndex(pageAddress);
        int blockPosition = decodePosition(pageAddress);

        return pagesHashStrategy.isPositionNull(blockIndex, blockPosition);
    }

    private long readHashPosition(int position)
    {
        long pageAddress = addresses.getLong(position);
        int blockIndex = decodeSliceIndex(pageAddress);
        int blockPosition = decodePosition(pageAddress);

        return pagesHashStrategy.hashPosition(blockIndex, blockPosition);
    }

    private boolean positionEqualsCurrentRowIgnoreNulls(int leftPosition, byte rawHash, int rightPosition, Page rightPage)
    {
        if (positionToHashes[leftPosition] != rawHash) {
            return false;
        }

        long pageAddress = addresses.getLong(leftPosition);
        int blockIndex = decodeSliceIndex(pageAddress);
        int blockPosition = decodePosition(pageAddress);

        return pagesHashStrategy.positionEqualsRowIgnoreNulls(blockIndex, blockPosition, rightPosition, rightPage);
    }

    private boolean positionEqualsPositionIgnoreNulls(int leftPosition, int rightPosition)
    {
        long leftPageAddress = addresses.getLong(leftPosition);
        int leftBlockIndex = decodeSliceIndex(leftPageAddress);
        int leftBlockPosition = decodePosition(leftPageAddress);

        long rightPageAddress = addresses.getLong(rightPosition);
        int rightBlockIndex = decodeSliceIndex(rightPageAddress);
        int rightBlockPosition = decodePosition(rightPageAddress);

        return pagesHashStrategy.positionEqualsPositionIgnoreNulls(leftBlockIndex, leftBlockPosition, rightBlockIndex, rightBlockPosition);
    }

    public static long getEstimatedRetainedSizeInBytes(
            int positionCount,
            HashArraySizeSupplier hashArraySizeSupplier,
            LongArrayList addresses,
            List> channels,
            long blocksSizeInBytes)
    {
        return sizeOf(addresses.elements()) +
                (channels.size() > 0 ? sizeOf(channels.get(0).elements()) * channels.size() : 0) +
                blocksSizeInBytes +
                sizeOfIntArray(hashArraySizeSupplier.getHashArraySize(positionCount)) +
                sizeOfByteArray(positionCount);
    }
}




© 2015 - 2024 Weber Informatics LLC | Privacy Policy