rdblue commented on code in PR #16100: URL: https://github.com/apache/iceberg/pull/16100#discussion_r3192092709
########## core/src/main/java/org/apache/iceberg/TrackedFileAdapters.java: ########## @@ -0,0 +1,600 @@ +/* + * 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.iceberg; + +import java.nio.ByteBuffer; +import java.util.Collections; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.function.Function; +import org.apache.iceberg.relocated.com.google.common.base.Preconditions; +import org.apache.iceberg.relocated.com.google.common.collect.Maps; +import org.apache.iceberg.types.Conversions; +import org.apache.iceberg.types.Type; + +/** + * Adapts {@link TrackedFile} entries to the {@link DataFile} and {@link DeleteFile} APIs. + * + * <p>V4 colocates deletion vectors with data file entries in {@link TrackedFile}. Rather than + * extending {@link DataFile} with deletion vector fields, DVs are extracted as separate {@link + * DeleteFile} objects via {@link #asDVDeleteFile(TrackedFile, Map)}. This matches the v3 convention + * where DVs are tracked as {@link DeleteFile} entries in delete manifests and keeps the existing + * {@link FileScanTask} contract ({@code file()} + {@code deletes()}) unchanged. + */ +class TrackedFileAdapters { + + private TrackedFileAdapters() {} + + static DataFile asDataFile(TrackedFile file, Map<Integer, PartitionSpec> specsById) { + Preconditions.checkState( + file.contentType() == FileContent.DATA, + "Cannot convert tracked file to DataFile: content type is %s, not DATA", + file.contentType()); + return new TrackedDataFile(file, resolveSpec(file, specsById)); + } + + static DeleteFile asDVDeleteFile(TrackedFile file, Map<Integer, PartitionSpec> specsById) { + Preconditions.checkState( + file.contentType() == FileContent.DATA, + "Cannot extract DV from tracked file: content type is %s, not DATA", + file.contentType()); + Preconditions.checkState( + file.deletionVector() != null, "Cannot extract DV from tracked file: no deletion vector"); + return new TrackedDVDeleteFile(file, resolveSpec(file, specsById)); + } + + static DeleteFile asEqualityDeleteFile(TrackedFile file, Map<Integer, PartitionSpec> specsById) { + Preconditions.checkState( + file.contentType() == FileContent.EQUALITY_DELETES, + "Cannot convert tracked file to DeleteFile: content type is %s, not EQUALITY_DELETES", + file.contentType()); + return new TrackedDeleteFile(file, resolveSpec(file, specsById)); + } + + private static PartitionSpec resolveSpec( + TrackedFile file, Map<Integer, PartitionSpec> specsById) { + Integer specId = file.specId(); + if (specId != null) { + PartitionSpec spec = specsById.get(specId); + Preconditions.checkArgument( + spec != null, "Cannot find partition spec for spec ID: %s", specId); + return spec; + } + + return PartitionSpec.unpartitioned(); + } + + // TODO: TrackedFile will likely get an explicit partition tuple field (using a union partition + // schema), replacing this transform-based derivation. Once that lands, this method should be + // removed and the adapter should read the tuple directly. + // + // This derives partition values by applying the partition transform to the lower bound of the + // source column stats. This is correct because each data file belongs to exactly one partition, + // so lower == upper for partition source columns. For non-identity transforms (bucket, truncate), + // the transform of the lower bound produces the correct partition value under this invariant. + @SuppressWarnings({"unchecked", "rawtypes"}) + static StructLike extractPartition(TrackedFile file, PartitionSpec spec) { + if (spec.isUnpartitioned()) { + return BaseFile.EMPTY_PARTITION_DATA; + } + + ContentStats stats = file.contentStats(); + if (stats == null) { + return new PartitionData(spec.partitionType()); + } + + PartitionData partition = new PartitionData(spec.partitionType()); + + for (int i = 0; i < spec.fields().size(); i += 1) { + PartitionField field = spec.fields().get(i); + + if (field.transform().isVoid()) { + partition.set(i, null); + continue; + } + + FieldStats<?> fieldStats = stats.statsFor(field.sourceId()); + if (fieldStats == null || fieldStats.lowerBound() == null) { + partition.set(i, null); + continue; + } + + Type sourceType = spec.schema().findType(field.sourceId()); + Function boundTransform = field.transform().bind(sourceType); + partition.set(i, boundTransform.apply(fieldStats.lowerBound())); + } + + return partition; + } + + static Map<Integer, Long> valueCounts(ContentStats stats) { + if (stats == null) { + return null; + } + + Map<Integer, Long> result = Maps.newHashMap(); + for (FieldStats<?> fs : stats.fieldStats()) { + if (fs != null && fs.valueCount() != null) { + result.put(fs.fieldId(), fs.valueCount()); + } + } + + return result.isEmpty() ? null : Collections.unmodifiableMap(result); + } + + static Map<Integer, Long> nullValueCounts(ContentStats stats) { + if (stats == null) { + return null; + } + + Map<Integer, Long> result = Maps.newHashMap(); + for (FieldStats<?> fs : stats.fieldStats()) { + if (fs != null && fs.nullValueCount() != null) { + result.put(fs.fieldId(), fs.nullValueCount()); + } + } + + return result.isEmpty() ? null : Collections.unmodifiableMap(result); + } + + static Map<Integer, Long> nanValueCounts(ContentStats stats) { + if (stats == null) { + return null; + } + + Map<Integer, Long> result = Maps.newHashMap(); + for (FieldStats<?> fs : stats.fieldStats()) { + if (fs != null && fs.nanValueCount() != null) { + result.put(fs.fieldId(), fs.nanValueCount()); + } + } + + return result.isEmpty() ? null : Collections.unmodifiableMap(result); + } + + static Map<Integer, ByteBuffer> lowerBounds(ContentStats stats) { + if (stats == null) { + return null; + } + + Map<Integer, ByteBuffer> result = Maps.newHashMap(); + for (FieldStats<?> fs : stats.fieldStats()) { + if (fs != null && fs.lowerBound() != null && fs.type() != null) { + result.put(fs.fieldId(), Conversions.toByteBuffer(fs.type(), fs.lowerBound())); + } + } + + return result.isEmpty() ? null : Collections.unmodifiableMap(result); + } + + static Map<Integer, ByteBuffer> upperBounds(ContentStats stats) { + if (stats == null) { + return null; + } + + Map<Integer, ByteBuffer> result = Maps.newHashMap(); + for (FieldStats<?> fs : stats.fieldStats()) { + if (fs != null && fs.upperBound() != null && fs.type() != null) { + result.put(fs.fieldId(), Conversions.toByteBuffer(fs.type(), fs.upperBound())); + } + } + + return result.isEmpty() ? null : Collections.unmodifiableMap(result); + } + + /** + * Shared base for adapters that delegate to a {@link TrackedFile} for content file fields. + * + * <p>Subclasses provide {@code content()}, {@code firstRowId()}, {@code equalityFieldIds()}, and + * the copy methods. + */ + private abstract static class AbstractTrackedContentFile<F extends ContentFile<F>> + implements ContentFile<F> { + private final TrackedFile file; + private final Tracking tracking; + private final PartitionSpec spec; + + private AbstractTrackedContentFile(TrackedFile file, PartitionSpec spec) { + this.file = file; + this.tracking = file.tracking(); + this.spec = spec; + } + + TrackedFile file() { + return file; + } + + Tracking tracking() { + return tracking; + } + + PartitionSpec spec() { + return spec; + } + + @Override + public Long pos() { + return tracking != null ? tracking.manifestPos() : null; + } + + @Override + public int specId() { + return spec.specId(); Review Comment: If the spec ID is set on `file` then it should be returned. The last version was this: ```java return file.specId() != null ? file.specId() : 0; ``` The problem wasn't that it was returning `file.specId()`. When that spec ID is set, it is canonical and is the ID that was used to look up `spec`. The problem was that it was guessing ID `0` for the unpartitioned spec, which is not correct. The updated version should be this: ```java return file.specId() != null ? file.specId() : spec.specId(); ``` -- 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]
