## Release History

### 4.49.0-beta.1 (Unreleased)

#### Features Added

#### Breaking Changes

#### Bugs Fixed
* Improved partition planning performance for change feed with large number of feed ranges. - See [PR 49086](https://github.com/Azure/azure-sdk-for-java/pull/49086)

#### Other Changes

### 4.48.0 (2026-05-01)

#### Features Added
* Added `additionalHeaders` support to allow setting additional headers (e.g., `x-ms-cosmos-workload-id`) that are sent with every request. - See [PR 48128](https://github.com/Azure/azure-sdk-for-java/pull/48128)
* Added new `CosmosItemsDataSource.readManyByPartitionKeys` Spark function to execute bulk queries by a list of pk-values with better efficiency. Configure null handling via `spark.cosmos.read.readManyByPk.nullHandling` - default `Null` treats a null PK column as JSON null (`addNullValue`), `None` treats it as `PartitionKey.NONE` (`addNoneValue` / `NOT IS_DEFINED`). These route to different physical partitions - picking the wrong mode silently returns zero rows. See [PR 48801](https://github.com/Azure/azure-sdk-for-java/pull/48801)
* Added Spark config `spark.cosmos.read.readManyByPk.maxConcurrentBatchPrefetch` (default `1`) to bound the per-task prefetch parallelism the SDK uses inside `readManyByPartitionKeys`. See [PR 48801](https://github.com/Azure/azure-sdk-for-java/pull/48801)
* Added Spark config `spark.cosmos.read.readManyByPk.maxBatchSize` (default `100`) to set the max. number of partition keys used for a single batch. See [PR 48930](https://github.com/Azure/azure-sdk-for-java/pull/48930)

### 4.47.0 (2026-04-17)

#### Features Added
* Added support for change feed with `startFrom` point-in-time on merged partitions by enabling the `CHANGE_FEED_WITH_START_TIME_POST_MERGE` SDK capability in the azure-cosmos SDK. - See [PR 48752](https://github.com/Azure/azure-sdk-for-java/pull/48752)

#### Bugs Fixed
* Fixed an issue where `readContainerThroughput` was always called even when `targetThroughput` is explicitly configured, requiring unnecessary `throughputSettings/read` permission for AAD principals. - See [PR 48800](https://github.com/Azure/azure-sdk-for-java/pull/48800)
* Fixed JVM `<clinit>` deadlock when multiple threads concurrently trigger Cosmos SDK class loading for the first time. - See [PR 48689](https://github.com/Azure/azure-sdk-for-java/pull/48689)

### 4.46.0 (2026-03-27)

#### Bugs Fixed
* Fixed an issue where creating containers with hierarchical partition keys (multi-hash) through the Spark catalog on the AAD path would fail. - See [PR 48548](https://github.com/Azure/azure-sdk-for-java/pull/48548)

### 4.45.0 (2026-03-13)

#### Features Added
* Added `vectorEmbeddingPolicy` support in Spark catalog `TBLPROPERTIES` for creating vector-search-enabled containers. - See [PR 48349](https://github.com/Azure/azure-sdk-for-java/pull/48349)

### 4.44.2 (2026-03-05)

#### Other Changes
* Changed azure-resourcemanager-cosmos usage to a pinned version which is deployed across all public and non-public clouds - [PR 48268](https://github.com/Azure/azure-sdk-for-java/pull/48268)

### 4.44.1 (2026-03-03)

#### Other Changes
* Reduced noisy warning logs in Gateway mode - [PR 48189](https://github.com/Azure/azure-sdk-for-java/pull/48189)

### 4.44.0 (2026-02-27)

#### Features Added
* Added config entry `spark.cosmos.account.azureEnvironment.management.scope` to allow specifying the Entra ID scope/audience to be used when retrieving tokens to authenticate against the ARM/management endpoint of non-public clouds. - See [PR 48137](https://github.com/Azure/azure-sdk-for-java/pull/48137)

### 4.43.1 (2026-02-25)

#### Bugs Fixed
* Fixed an issue where `TransientIOErrorsRetryingIterator` would trigger extra query during retries and on close. - See [PR 47996](https://github.com/Azure/azure-sdk-for-java/pull/47996)

#### Other Changes
* Added status code history in `BulkWriterNoProgressException` error message. - See [PR 48022](https://github.com/Azure/azure-sdk-for-java/pull/48022)
* Reduced the log noise level for frequent transient errors - for example throttling - in Gateway mode - [PR 48112](https://github.com/Azure/azure-sdk-for-java/pull/48112)

### 4.43.0 (2026-02-10)

#### Features Added
* Added transactional batch support. See [PR 47478](https://github.com/Azure/azure-sdk-for-java/pull/47478) and [PR 47697](https://github.com/Azure/azure-sdk-for-java/pull/47697) and [47803](https://github.com/Azure/azure-sdk-for-java/pull/47803)
* Added support for throughput bucket. - See [47856](https://github.com/Azure/azure-sdk-for-java/pull/47856)

#### Bugs Fixed
* Fixed an issue for micro batch stream query where feed range starts with null or incorrect initial offset. **NOTE:** This issue only happens when a partition split happened during initial offset calculation stage. - See [47742](https://github.com/Azure/azure-sdk-for-java/pull/47742)
* Fixed `java.lang.ClassCastException` during bulk write operations for write strategy `ItemPatch` or `ItemPatchIfExists`. - See [47748](https://github.com/Azure/azure-sdk-for-java/pull/47748)

### 4.42.0 (2025-12-09)

#### Other Changes
* Enabled hostname validation for RNTBD connections to backend - [PR 47111](https://github.com/Azure/azure-sdk-for-java/pull/47111)

### 4.41.0 (2025-10-21)

#### Features Added
* Added support `spark.cosmos.write.strategy` value `ItemPatchIfExists` which allows gracefully ignoring documents/patch-instructions when the document does not exist (anymore). - See [47034](https://github.com/Azure/azure-sdk-for-java/pull/47034)
* Added support to optionally omit info about spark environment and/or machine-info for driver/executors from `UserAgent` header via new config `spark.cosmos.userAgent.format` (allowed values are `SparkEnvAndWorkers` (default value), `OnlySparkEnv` and `NoSparkEnv`.  - See [47047](https://github.com/Azure/azure-sdk-for-java/pull/47047)

### 4.40.0 (2025-09-27)

#### Features Added
* Added support for feed range cache refresh interval config. - See [46759](https://github.com/Azure/azure-sdk-for-java/pull/46759)

#### Other Changes
* Added improvement to reduce partition planning time for large containers. - See [46727](https://github.com/Azure/azure-sdk-for-java/pull/46727)

### 4.39.0 (2025-09-05)

#### Bugs Fixed
* Reverted known issue due to shading log4j (which was introduced in 4.38.1). - See [PR 46546](https://github.com/Azure/azure-sdk-for-java/pull/46546) and [PR 46608](https://github.com/Azure/azure-sdk-for-java/pull/46608)
* Added change feed performance monitoring which is used to improve end lsn calculation in `CosmosPartitionPlanner`. - See [PR 46320](https://github.com/Azure/azure-sdk-for-java/pull/46320)
* Added `spark.cosmos.auth.aad.audience` as a valid configuration option to allow using AAD tokens with custom audiences. - See [PR 46554](https://github.com/Azure/azure-sdk-for-java/pull/46554)

### 4.38.1 (2025-08-22)

**NOTE: This version has a known issue due to shading log4j - Please use more recent versions >= 4.38.2 or 4.38.0 instead**

#### Other Changes
* Added log4j-core to the list of shaded packages to avoid conflicts when customers use log4j in a different version. **NOTE: This change caused known issue - Please use a more recent version instead** - See [PR 45924](https://github.com/Azure/azure-sdk-for-java/pull/46451)

### 4.38.0 (2025-07-31)

#### Features Added
* Added telemetry support by adding OTEL span attribute naming schemes, introducing Azure Monitor integration, and sampled diagnostics. - See [PR 45924](https://github.com/Azure/azure-sdk-for-java/pull/45924)

#### Other Changes
* Added compatibility with CosmosDB Fabric Native Accounts using the `FabricAccountDataResolver` for authentication. - See [PR 45890](https://github.com/Azure/azure-sdk-for-java/pull/45890)

### 4.37.2 (2025-05-14)

#### Features Added
* Added option to use the connector in non-public Azure clouds. - See [PR 45310](https://github.com/Azure/azure-sdk-for-java/pull/45310)

#### Bugs Fixed
* Fixed an issue during bulk write operations that could result in failing the Spark job in `BulkWriter.flushAndClose` too eagerly in certain cases. - See [PR 44992](https://github.com/Azure/azure-sdk-for-java/pull/44992)
* Fixed hang issue in `CosmosPagedIterable#handle` by preventing race conditions in underlying subscription of `Flux<FeedResponse>`. - [PR 45290](https://github.com/Azure/azure-sdk-for-java/pull/45290)

### 4.37.1 (2025-03-04)

#### Features Added
* Added config option `spark.cosmos.read.responseContinuationTokenLimitInKb` to reduce query continuation token size. - See [PR 44480](https://github.com/Azure/azure-sdk-for-java/pull/44480)

### 4.37.0 (2025-02-20)

#### Other Changes
* Updated netty dependency

### 4.36.1 (2025-02-08)

#### Bugs Fixed
* Fixed an issue in change feed where under certain rare race conditions records could be skipped and excessive requests are prefetched. - See [PR 43788](https://github.com/Azure/azure-sdk-for-java/pull/43788)

### 4.36.0 (2025-01-14)
> [!IMPORTANT]
> We strongly recommend our customers to use version 4.36.0 and above especially if using all versions and deletes change feed.

#### Features Added
* Added the udfs `GetFeedRangesForContainer` and `GetOverlappingFeedRange` to ease mapping of cosmos partition key to databricks table partition key. - See [PR 43092](https://github.com/Azure/azure-sdk-for-java/pull/43092)

#### Bugs Fixed
* Added null checking for previous images for deletes in full fidelity change feed. - See [PR 43483](https://github.com/Azure/azure-sdk-for-java/pull/43483)

#### Other Changes
* Added options to fine-tune settings for bulk operations. - [PR 43509](https://github.com/Azure/azure-sdk-for-java/pull/43509)

### 4.35.0 (2024-11-27)

#### Bugs Fixed
* Fixed an issue when using `ChangeFeed` causing some cosmos partitions to not be fully processed in some cases. - See [PR 42553](https://github.com/Azure/azure-sdk-for-java/pull/42553)

### 4.34.0 (2024-10-10)
#### Bugs Fixed
* Fixed an issue to avoid transient `IllegalArgumentException` due to duplicate json properties for the `uniqueKeyPolicy` property in `DocumentCollection`. - See [PR 41608](https://github.com/Azure/azure-sdk-for-java/pull/41608) and [PR 42244](https://github.com/Azure/azure-sdk-for-java/pull/42244)

### 4.33.1 (2024-08-23)

#### Bugs Fixed
* Fixed an issue to avoid transient `IllegalArgumentException` due to duplicate json properties for the `uniqueKeyPolicy` property. - See [PR 41608](https://github.com/Azure/azure-sdk-for-java/pull/41608)

#### Other Changes
* Added retries on a new `BulkWriter` instance when first attempt to commit times out for bulk write jobs. - See [PR 41553](https://github.com/Azure/azure-sdk-for-java/pull/41553)

### 4.33.0 (2024-06-22)

#### Features Added
* Added a service trait `CosmosClientBuilderInterceptor` to allow intercepting and customizing the CosmosClient creation. - See [PR 40714](https://github.com/Azure/azure-sdk-for-java/pull/40714)

#### Bugs Fixed
* Fixed a race condition resulting in not always re-enqueueing retries for bulk writes. - See [PR 40714](https://github.com/Azure/azure-sdk-for-java/pull/40714)

### 4.32.1 (2024-06-07)

#### Other Changes
* Added retries when retrieving new pages for query or readMany operations are timing out to avoid unbounded awaits. - See [PR 40506](https://github.com/Azure/azure-sdk-for-java/pull/40506)
* Ensured that no statistics are reported when custom queries via `spark.cosmos.read.customQuery` are used. - See [PR 40506](https://github.com/Azure/azure-sdk-for-java/pull/40506)

### 4.32.0 (2024-05-24)

#### Features Added
* Added config option `spark.cosmos.auth.aad.clientCertPemBase64` to allow using SPN (ServicePrincipal name) authentication with certificate instead of client secret. - See [PR 40325](https://github.com/Azure/azure-sdk-for-java/pull/40325)
* Added config option `spark.cosmos.accountDataResolverServiceName` to allow specifying which `AccountDataResolver` trait implementation to use if there are multiple on the class path. - See [PR 40325](https://github.com/Azure/azure-sdk-for-java/pull/40325)

#### Bugs Fixed
* Fixed an issue where `SHOW DATABASES IN` only return one database even though multiple databases exist. - See [PR 40277](https://github.com/Azure/azure-sdk-for-java/pull/40277)
* Fixed an issue where `SHOW TABLES FROM` only return one container even though multiple containers exist. - See [PR 40277](https://github.com/Azure/azure-sdk-for-java/pull/40277)
* Fixed UserAgent encoding when the suffix contains non-ASCII characters. - See[PR 40293](https://github.com/Azure/azure-sdk-for-java/pull/40293)

#### Other Changes
* Added robustness improvement to avoid client-side parsing errors `java.lang.IllegalArgumentException: Unable to parse JSON` when Gateway returns duplicate `unqiueKeyPolicy` in IndexPolicy (invalid json) - See[PR 40306](https://github.com/Azure/azure-sdk-for-java/pull/40306)

### 4.31.0 (2024-05-20)

#### Features Added
* Added capability in azure-cosmos-spark to allow the spark environment to support access tokens via AccountDataResolver. - See [PR 40079](https://github.com/Azure/azure-sdk-for-java/pull/40079)

### 4.30.0 (2024-04-27)

#### Features Added
* Added capability to use (and enforce) native netty transport. The native transport is more efficient - esepcially when the number of TCP connections being used is high. - See [PR 39834](https://github.com/Azure/azure-sdk-for-java/pull/39834)
* Added ManagedIdentity authentication support for azure-cosmos-spark in Databricks. - See [PR 39870](https://github.com/Azure/azure-sdk-for-java/pull/39870)

### 4.29.0 (2024-04-16)

#### Features Added
* Spark 3.5 support: - See [PR 39395](https://github.com/Azure/azure-sdk-for-java/pull/39395).

#### Bugs Fixed
* Fixed an issue causing failures when using change feed in batch mode with a batch location and `ChangeFeedBatch.planInputPartitions` is called multiple times (for example because physcial query plan gets retrieved) and some changes have been made in the monitored container between those calls). - See [PR 39635](https://github.com/Azure/azure-sdk-for-java/pull/39635)
* Made `AccountDataResolver` trait public again. - See [PR 39736](https://github.com/Azure/azure-sdk-for-java/pull/39736)

#### Other Changes
* Optimized the partitioning strategy implementation details to avoid unnecessarily high RU usage. - See [PR 39438](https://github.com/Azure/azure-sdk-for-java/pull/39438)
  
### NOTE: See CHANGELOG.md in 3.1, 3.2, 3.3 and 3.4 projects for changes prior to 4.29.0