Skip to content

[IcebergIO] Cache destination Row keys in AssignDestinationsAndPartitions #39155

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
atognolas wants to merge 1 commit into
base: master
Choose a base branch
from
Open

[IcebergIO] Cache destination Row keys in AssignDestinationsAndPartitions #39155

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 9 additions & 2 deletions ...iceberg/src/main/java/org/apache/beam/sdk/io/iceberg/AssignDestinationsAndPartitions.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,7 @@ static class AssignDoFn extends DoFn<Row, KV<Row, Row>> {
private transient @MonotonicNonNull Map<String, PartitionKey> partitionKeys;
private transient @MonotonicNonNull Map<String, BeamRowWrapper> wrappers;
private transient @MonotonicNonNull Map<String, Instant> lastRefreshTimes;
private transient @MonotonicNonNull Map<String, Row> keyCache;

@gemini-code-assist gemini-code-assist Bot Jun 29, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

Using a single flat map with a concatenated string key (tableIdentifier + "|" + partitionPath) requires allocating a new String object for every single input element, even on cache hits. Since this DoFn processes millions of rows, this overhead can be avoided by using a nested map structure: Map<String, Map<String, Row>>.

Suggested change
private transient @MonotonicNonNull Map<String, Row> keyCache;
private transient @MonotonicNonNull Map<String, Map<String, Row>> keyCache;


private final DynamicDestinations dynamicDestinations;
private final IcebergCatalogConfig catalogConfig;
Expand All @@ -98,6 +99,7 @@ public void setup() {
this.wrappers = new HashMap<>();
this.partitionKeys = new HashMap<>();
this.lastRefreshTimes = new HashMap<>();
this.keyCache = new HashMap<>();
}

@ProcessElement
Expand Down Expand Up @@ -175,8 +177,13 @@ public void processElement(

String partitionPath = partitionKey.toPath();

Row destAndPartition =
Row.withSchema(OUTPUT_SCHEMA).addValues(tableIdentifier, partitionPath).build();
String cacheKey = tableIdentifier + "|" + partitionPath;
Row destAndPartition = checkStateNotNull(keyCache).get(cacheKey);
if (destAndPartition == null) {
destAndPartition =
Row.withSchema(OUTPUT_SCHEMA).addValues(tableIdentifier, partitionPath).build();
keyCache.put(cacheKey, destAndPartition);
}
Comment on lines +180 to +186

@gemini-code-assist gemini-code-assist Bot Jun 29, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

Instead of concatenating tableIdentifier and partitionPath into a new String on every element, use a nested map lookup. This avoids any object allocation on cache hits, which is the common case.

      Map<String, Map<String, Row>> cache = checkStateNotNull(keyCache);
      Map<String, Row> tableCache = cache.get(tableIdentifier);
      if (tableCache == null) {
        tableCache = new HashMap<>();
        cache.put(tableIdentifier, tableCache);
      }
      Row destAndPartition = tableCache.get(partitionPath);
      if (destAndPartition == null) {
        destAndPartition =
            Row.withSchema(OUTPUT_SCHEMA).addValues(tableIdentifier, partitionPath).build();
        tableCache.put(partitionPath, destAndPartition);
      }


out.output(KV.of(destAndPartition, data));
}
Expand Down
Loading