xiangfu0 commented on code in PR #19998:
URL: https://github.com/apache/pinot/pull/19998#discussion_r4346516848
##########
pinot-server/src/main/java/org/apache/pinot/server/api/resources/TablesResource.java:
##########
@@ -396,11 +397,15 @@ public String getCompressionStats(
static int addColumnContributions(int currentCount, ImmutableSegment segment,
@Nullable Set<String> columnFilter) {
int additionalCount = 0;
+ SegmentMetadata segmentMetadata = segment.getSegmentMetadata();
if (columnFilter == null) {
- additionalCount =
segment.getSegmentMetadata().getColumnMetadataMap().size();
+ additionalCount = segmentMetadata instanceof SegmentMetadataImpl
Review Comment:
Minor: this counts `getColumnNames().size()` (physical + virtual) but
`SegmentCompressionStatsReader` only emits rows for physical columns. So tables
with many virtual columns over-count here, and the `MAX_COLUMN_CONTRIBUTIONS`
cap could reject a request whose real output would actually fit. Using the
physical column count would match what's emitted.
##########
pinot-segment-local/src/main/java/org/apache/pinot/segment/local/indexsegment/immutable/ImmutableSegmentImpl.java:
##########
@@ -333,12 +331,12 @@ public DataSource getDataSource(String column, Schema
schema) {
@Override
public Set<String> getColumnNames() {
- return _segmentMetadata.getSchema().getColumnNames();
+ return _indexContainerMap.keySet();
Review Comment:
This now returns `_indexContainerMap.keySet()` instead of the schema's
column names, which changes behavior for OPEN_STRUCT columns. The parent column
lives only in `_dataSources` (synthesized above) and isn't in the container
map, while the materialized children (`col$key`, `col$__sparse__`) are. So
`SELECT *` would now expand to the internal child columns and drop the logical
parent. Could we keep using the schema column set here, or at least exclude
materialized children?
##########
pinot-segment-spi/src/main/java/org/apache/pinot/segment/spi/index/metadata/SegmentMetadataImpl.java:
##########
@@ -483,21 +574,84 @@ public String getEndOffset() {
@Override
public TreeMap<String, ColumnMetadata> getColumnMetadataMap() {
- return _columnMetadataMap;
+ if (_mappedStore == null) {
+ return _columnMetadataMap;
+ }
+ // Compatibility API for metadata/admin callers. The serving path uses
column names and direct lookup instead.
+ TreeMap<String, ColumnMetadata> result = new TreeMap<>();
+ for (String column : getPhysicalColumnNames()) {
+ result.put(column, getColumnMetadataFor(column));
+ }
+ result.putAll(_virtualColumnMetadata);
+ return result;
+ }
+
+ /// Returns the physical column names without materializing a metadata map
or Schema on mapped segments.
+ public Set<String> getPhysicalColumnNames() {
+ return _mappedStore != null ? _physicalColumnNames :
_schema.getPhysicalColumnNames();
+ }
+
+ /// Indicates whether the immutable segment is reading its metadata from the
binary mapping.
+ public boolean isMapped() {
+ return _mappedStore != null;
+ }
+
+ @Nullable
+ @Override
+ public ColumnMetadata getColumnMetadataFor(String column) {
+ if (_mappedStore == null) {
+ return _columnMetadataMap != null ? _columnMetadataMap.get(column) :
null;
+ }
+ ColumnMetadata virtual = _virtualColumnMetadata.get(column);
+ if (virtual != null) {
+ return virtual;
+ }
+ if (_removedColumns.contains(column)) {
+ return null;
+ }
+ int ordinal = _mappedStore.ordinal(column);
+ return ordinal >= 0 ? new MappedColumnMetadata(_mappedStore, ordinal) :
null;
Review Comment:
This allocates a fresh `MappedColumnMetadata` on every call, so the
per-instance caches inside it (bounds, partitions, partitionFunction) never get
reused. Hot callers like the upsert/dedup managers and the constructor loops in
`ImmutableSegmentImpl` would re-decode min/max and partition sets from the
buffer each time. Worth caching the per-ordinal view so those memoizations
actually stick.
##########
pinot-segment-spi/src/main/java/org/apache/pinot/segment/spi/index/metadata/MappedColumnMetadata.java:
##########
@@ -0,0 +1,364 @@
+/**
+ * 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.pinot.segment.spi.index.metadata;
+
+import com.fasterxml.jackson.core.type.TypeReference;
+import java.io.ByteArrayInputStream;
+import java.io.DataInputStream;
+import java.io.IOException;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import javax.annotation.Nullable;
+import org.apache.pinot.segment.spi.ColumnMetadata;
+import org.apache.pinot.segment.spi.compression.ChunkCompressionType;
+import org.apache.pinot.segment.spi.index.IndexService;
+import org.apache.pinot.segment.spi.index.IndexType;
+import org.apache.pinot.segment.spi.partition.PartitionFunction;
+import org.apache.pinot.segment.spi.partition.PartitionFunctionFactory;
+import org.apache.pinot.spi.config.table.FieldConfig.EncodingType;
+import org.apache.pinot.spi.data.FieldSpec;
+import org.apache.pinot.spi.data.FieldSpec.DataType;
+import org.apache.pinot.spi.utils.JsonUtils;
+
+/// A small view of one fixed record in [MappedSegmentMetadataStore]. No
per-column values are copied into this view.
+/// The segment owns the mapping; callers must not use the view after the
segment directory is closed.
+public final class MappedColumnMetadata implements ColumnMetadata {
+ private final MappedSegmentMetadataStore _store;
+ private final int _ordinal;
+ @Nullable
+ private volatile PartitionFunction _partitionFunction;
+ @Nullable
+ private volatile Set<Integer> _partitions;
+ @Nullable
+ private volatile Bounds _bounds;
+
+ private static final class Bounds {
+ @Nullable
+ private final Comparable<?> _minimum;
+ @Nullable
+ private final Comparable<?> _maximum;
+
+ private Bounds(@Nullable Comparable<?> minimum, @Nullable Comparable<?>
maximum) {
+ _minimum = minimum;
+ _maximum = maximum;
+ }
+ }
+
+ MappedColumnMetadata(MappedSegmentMetadataStore store, int ordinal) {
+ _store = store;
+ _ordinal = ordinal;
+ }
+
+ @Override
+ public FieldSpec getFieldSpec() {
+ // Field specs are shared across mapped segments; callers must treat the
returned spec as read-only.
+ return _store.fieldSpec(_ordinal);
+ }
+
+ @Override
+ public int getTotalDocs() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.TOTAL_DOCS);
+ }
+
+ @Override
+ public int getCardinality() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.CARDINALITY);
+ }
+
+ @Override
+ public boolean hasDictionary() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.HAS_DICTIONARY);
+ }
+
+ @Nullable
+ @Override
+ public EncodingType getForwardIndexEncoding() {
+ switch (_store.getInt(_ordinal, MappedSegmentMetadataStore.ENCODING)) {
+ case 0:
+ return null;
+ case 1:
+ return EncodingType.RAW;
+ case 2:
+ return EncodingType.DICTIONARY;
+ default:
+ throw new IllegalStateException("Invalid forward-index encoding");
+ }
+ }
+
+ @Override
+ public boolean isSorted() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.SORTED);
+ }
+
+ @Override
+ public boolean isNonNull() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.NON_NULL);
+ }
+
+ @Nullable
+ @Override
+ public Comparable<?> getMinValue() {
+ return bounds()._minimum;
+ }
+
+ @Nullable
+ @Override
+ public Comparable<?> getMaxValue() {
+ return bounds()._maximum;
+ }
+
+ private Bounds bounds() {
+ Bounds bounds = _bounds;
+ if (bounds == null) {
+ synchronized (this) {
+ bounds = _bounds;
+ if (bounds == null) {
+ bounds = new Bounds(_store.value(_ordinal, true),
_store.value(_ordinal, false));
+ _bounds = bounds;
+ }
+ }
+ }
+ return bounds;
+ }
+
+ @Override
+ public boolean isMinMaxValueInvalid() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.INVALID_MIN_MAX);
+ }
+
+ @Override
+ public int getLengthOfShortestElement() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.SHORTEST);
+ }
+
+ @Override
+ public int getLengthOfLongestElement() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.LONGEST);
+ }
+
+ @Override
+ public boolean isAscii() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.ASCII);
+ }
+
+ @Override
+ public int getBitsPerElement() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.BITS);
+ }
+
+ @Override
+ public int getTotalNumberOfEntries() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.TOTAL_ENTRIES);
+ }
+
+ @Override
+ public int getMaxNumberOfMultiValues() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.MAX_MV);
+ }
+
+ @Override
+ public int getMaxRowLengthInBytes() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.MAX_ROW);
+ }
+
+ @Nullable
+ @Override
+ public PartitionFunction getPartitionFunction() {
+ PartitionFunction cached = _partitionFunction;
+ if (cached != null) {
+ return cached;
+ }
+ byte[] bytes = _store.blob(_ordinal,
MappedSegmentMetadataStore.PARTITION_FUNCTION);
+ if (bytes == null) {
+ return null;
+ }
+ try (DataInputStream input = new DataInputStream(new
ByteArrayInputStream(bytes))) {
+ String name = MappedSegmentMetadataStore.readString(input);
+ int numPartitions = input.readInt();
+ int configSize = input.readInt();
+ Map<String, String> config = configSize != 0 ? new HashMap<>(configSize)
: null;
+ for (int i = 0; i < configSize; i++) {
+ config.put(MappedSegmentMetadataStore.readString(input),
MappedSegmentMetadataStore.readString(input));
+ }
+ PartitionFunction function =
PartitionFunctionFactory.getPartitionFunction(name, numPartitions, config);
+ _partitionFunction = function;
+ return function;
+ } catch (IOException e) {
+ throw new IllegalStateException("Cannot decode partition function for "
+ getColumnName(), e);
+ }
+ }
+
+ @Nullable
+ @Override
+ public Set<Integer> getPartitions() {
+ Set<Integer> cached = _partitions;
+ if (cached != null) {
+ return cached;
+ }
+ byte[] bytes = _store.blob(_ordinal,
MappedSegmentMetadataStore.PARTITIONS);
+ if (bytes == null) {
+ return null;
+ }
+ try (DataInputStream input = new DataInputStream(new
ByteArrayInputStream(bytes))) {
+ int count = input.readInt();
+ Set<Integer> partitions = new HashSet<>(count);
+ for (int i = 0; i < count; i++) {
+ partitions.add(input.readInt());
+ }
+ _partitions = partitions;
+ return partitions;
+ } catch (IOException e) {
+ throw new IllegalStateException("Cannot decode partitions for " +
getColumnName(), e);
+ }
+ }
+
+ @Override
+ public boolean isAutoGenerated() {
+ return _store.flag(_ordinal, MappedSegmentMetadataStore.AUTO_GENERATED);
+ }
+
+ @Override
+ public boolean isMaterializedChild() {
+ return getParentColumn() != null;
+ }
+
+ @Nullable
+ @Override
+ public String getParentColumn() {
+ return _store.string(_ordinal, MappedSegmentMetadataStore.PARENT);
+ }
+
+ @Nullable
+ @Override
+ public List<String> getSparseKeys() {
+ return readJson(MappedSegmentMetadataStore.SPARSE_KEYS, new
TypeReference<List<String>>() { });
+ }
+
+ @Nullable
+ @Override
+ public Map<String, Integer> getSparseMultiValueKeys() {
+ return readJson(MappedSegmentMetadataStore.SPARSE_MULTI_VALUE_KEYS,
+ new TypeReference<Map<String, Integer>>() { });
+ }
+
+ @Nullable
+ @Override
+ public Map<String, DataType> getSparseKeyTypes() {
+ return readJson(MappedSegmentMetadataStore.SPARSE_KEY_TYPES,
+ new TypeReference<Map<String, DataType>>() { });
+ }
+
+ @Nullable
+ private <T> T readJson(int field, TypeReference<T> type) {
+ String json = _store.string(_ordinal, field);
+ if (json == null) {
+ return null;
+ }
+ try {
+ return JsonUtils.stringToObject(json, type);
+ } catch (IOException e) {
+ throw new IllegalStateException("Cannot decode OPEN_STRUCT metadata for
" + getColumnName(), e);
+ }
+ }
+
+ @Override
+ public int getNumIndexes() {
+ return _store.getInt(_ordinal, MappedSegmentMetadataStore.INDEX_COUNT);
+ }
+
+ @Override
+ public short getIndexType(int position) {
+ return
IndexService.getInstance().getNumericId(IndexService.getInstance().get(indexAt(position)._type));
+ }
+
+ @Override
+ public long getIndexSize(int position) {
+ return indexAt(position)._size;
+ }
+
+ @Override
+ public long getIndexSizeFor(IndexType type) {
+ for (int i = 0; i < getNumIndexes(); i++) {
+ IndexSize entry = indexAt(i);
+ if (entry._type.equals(type.getId())) {
+ return entry._size;
+ }
+ }
+ return UNAVAILABLE;
+ }
+
+ /// Kept for the segment-metadata REST JSON representation.
+ @SuppressWarnings("unused")
+ public Map<IndexType<?, ?, ?>, Long> getIndexSizeMap() {
+ Map<IndexType<?, ?, ?>, Long> sizes = new HashMap<>();
+ for (int i = 0; i < getNumIndexes(); i++) {
+ IndexSize entry = indexAt(i);
+ sizes.put(IndexService.getInstance().get(entry._type), entry._size);
+ }
+ return sizes;
+ }
+
+ private IndexSize indexAt(int position) {
Review Comment:
`indexAt(position)` re-opens the blob and skips `position` entries every
call, so `getIndexSizeFor`/`getIndexSizeMap` looping `0..n` become O(n^2) in
the number of indexes. For wide columns in the compression-stats/index-size
endpoints that adds up. Decoding the list once into an array and indexing into
it would make this linear.
--
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]
