v0.9.0

🎉 New features

• Asset filters to pipeline bulk query and update operations directly and efficiently
• Search for terms by their categories and categories by their parent category
• Search and retrieve logged administrative events, such as user logins
• Models for:
  • Azure Event Hub
  • Functions
  • StarredBy details
  • DbtTag
• Adds a unique X-Atlan-Request-Id header to every call that can be used for tracing requests all the way through Atlan's back-end operations

⛑️ Breaking changes

• Replaces the retrieveFull(), retrieveMinimal(), retrieveByGuid() and retrieveByQualifiedName() methods with a single overloaded get() method:
  • From any asset you can now retrieve an instance using get(), for example: Table.get(). The SDK will automatically detect whether you are retrieving the asset by GUID or qualifiedName and act accordingly. To retrieve the minimal version (without relationships) simply pass an optional argument of false to not include the relationships.
  • For dynamic typing, you can use Asset.get() in place of Asset.retrieveFull() and Asset.retrieveMinimal().
• The builder() method of assets was only ever intended for internal use. As such, it has now been renamed _internal() to make this more apparent. Unless explicitly mentioned in developer.atlan.com, you should always use creator(), updater(), or trimToRequired() rather than directly accessing _internal() to construct an asset.
• The SDK now supports running against multiple tenants in a single JVM. To enable this, we've:
  • Added a mandatory AtlanClient parameter when creating an AssetBatch. (You can pass Atlan.getDefaultClient() to retain the existing behavior.)
  • Removed the ability to directly access the Serde.mapper ObjectMapper — instead now use the .readValue() and .writeValueAsString() operations on an AtlanClient. (For example: Atlan.getDefaultClient().readValue().)
  • Related to above, toJson() on any asset now requires a client — if you were using this directly, you can now call .toJson(Atlan.getDefaultClient()) to retain the existing behavior.
  • Removed the ability to directly access the various ...Endpoint classes. These are now only available through an AtlanClient, for example: Atlan.getDefaultClient().assets.save(List.of(asset1, asset2), false) rather than BulkEntityEndpoint.upsert(List.of(asset1, asset2), false).
  • Removed static caches — caches are now client-specific:
    • Anywhere you previously did something like RoleCache.getIdForName() you will now need to use Atlan.getDefaultClient().getRoleCache().getIdForName().
    • Note that this pattern applies to all caches (RoleCache, CustomMetadataCache, AtlanTagCache, EnumCache, UserCache and GroupCache).
• The static getBaseUrlSafe() method has been removed, as the getBaseUrl() method no longer throws a checked exception and can therefore be used directly instead.
• The upsert() operation has been deprecated, replaced by a save() operation that takes the same arguments and behaves in the same way.

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixes an IndexOutOfBoundsException that could occur when trying to retrieve a non-existent API token by its name.
• Fixes an issue with the default page size used for asset filters, so they should now work even if no page size is explicitly set.
• Marks all API-operable objects as serializable.
• Replaces the toString() method that produced JSON output with an object-native toString(). While this may feel less readable, it removes any dependency on tenant-specific information for the serialization process, and should avoid hiding any transformation or translations that would occur through such a serialization.
• Adds a missing PURGED status that can exist in some circumstances on deleted instances of metadata.
• No longer fails if custom metadata that no longer exists is encountered, but instead uses a value of (DELETED) and logs.
• Fixes blocking check on asset deletion.

🥗 QOL improvements

• Asset instance retrieval has been simplified to a single get() method that works for both GUIDs and qualifiedNames.
• The findByName() method that exists for various assets (connections, glossary objects, etc) no longer requires an attributes argument (if none is provided it will default to the minimal set of attributes now).
• With asset filters you can now quickly query assets of a given type, without needing to build or run a separate IndexSearchRequest. For example: Table.all().filter(...).stream().forEach(a -> log.info("Found asset: {}", a));
• All operations that you can invoke against an endpoint now have variants that can take an (optional) RequestOptions object. When provided, these allow you to override default client settings for that individual request — timeouts, max retries, etc.
• Bumps all library dependencies to their latest stable versions.

v0.8.0

🎉 New features

• Adds the ability to manage arbitrary files as assets
• Adds an abstraction layer for easing implementation of real-time event-processing through webhooks, in particular when implemented using AWS Lambda functions
• Index searches have been simplified through their builder() method, defaults for paging, and providing other shortcut factory methods (like of() for the DSL)
• Lineage listings have been simplified through their builder() method and defaults for parameters
• Index search and lineage listing responses now provide Iterable<Asset> and .stream() interfaces to simplify iterating through results. These make looping much simpler while simultaneously more efficient — In both cases the next pages of results are automatically fetched as-needed, lazily, so you can always break-out early without retrieving unnecessary results.

⛑️ Breaking changes

• Search builders now have a required argument (the DSL or query to search), rather than starting blank.
• Lineage list builders now have a required argument (the GUID of the asset to start from), rather than starting blank.
• AssetBatch operations now throw any checked exceptions rather than swallowing and logging them.
• Implements multi-inheritance polymorphism for assets.
  • This means a given asset type can now inherit its full set of attributes and relationships (across multiple supertypes).
  • However, it means we have removed the abstract classes that represented supertypes of assets and replaced these with interfaces. For example, there is no longer a SQL or a Catalog class, but an ISQL and an ICatalog interface.
  • Anywhere you previously were checking an asset using if (a instanceof Catalog) with such an abstract class, you'll now need to do the same check against the interface instead (e.g. if (a instanceof ICatalog)).
  • The benefit of this change is that the asset types in the SDK now accurately reflect multi-inheritance of Atlan itself. So a SnowflakeTag for example will be both instanceof ITag and instanceof ISQL. (With class-based inheritance, we had to choose just one of these to extend.)
  • Relationships are all now specified using their interface rather than their class, so for example getAssignedTerms() will now return IGlossaryTerms rather than GlossaryTerms. As long as you are accessing instance methods for your operations, this should be a transparent change for you in almost all cases, as the interface has all the same (non-static) methods defined on it that exist on the class.

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixes an issue introduced in the last release where the assignedTerms relationship was missing from generated test classes.
• Fixes an issue that required an attributes.csv file to be present for injecting descriptions into generated POJOs — this is now optional as part of the code generation.
• Fixes an issue for TableauDatasource objects, which can now return both calculated fields and datasource fields through the single fields relationship.
• Fixes the level of caching attempted for roles — previously every role was cached, whereas now only the workspace-roles will be cached.

🥗 QOL improvements

• The Iterable<Asset> interface provided by index search and lineage listing responses allow you to:
  • Use a simple for-each loop: for (Asset a : response) { ... }
  • Use a forEach() lambda: response.forEach(a -> { ... });
• The .stream() interface provides the results as a stream, allowing you to:
  • Use chain-able stream operations, such as: response.stream().filter(...).limit(n).forEach(a -> {...}).collect()
• When the retry limit is reached for automated retries, any underlying exceptions or response body should now be logged in addition to the fact that the retry limit was reached.
• Every snapshot build and release now uploads a lambda-layer.zip file that can be uploaded directly as an AWS Lambda layer, providing the full Java SDK (and its dependencies). You can then implement any of your own logic through the abstraction layer and simply upload that minimal code directly, without needing to re-package the entire SDK and its dependencies. (This allows you to upload and manage the SDK layer independently of possibly many event handlers implemented on top of it.)

v0.7.0

🎉 New features

• Adds higher-performance lineage list API
• Adds ability to manage an icon or image for tags (classifications)
• Adds new type definitions for:
  • MicroStrategy objects
• Renames "classification" to "tag" to align with UI changes
• Adds new persona/purpose/policy management

⛑️ Breaking changes

• Renames "classifications" to "tags". Back-end payloads remain unchanged, but any properties or methods that referred to classification will now refer to atlanTag. For example:
  • mappedClassificationName -> mappedAtlanTagName
  • purposeClassifications -> purposeAtlanTags
  • ClassificationCache -> AtlanTagCache
  • Classification -> AtlanTag
  • ClassificationDef -> AtlanTagDef
  • classifications -> atlanTags
  • classificationNames -> atlanTagNames
  • appendClassifications -> appendAtlanTags
  • addClassifications -> addAtlanTags
• Updates persona, purpose and policy management to use new optimized internal changes. As a result, the specific objects and methods needed to manage personas, purposes and policies have changed:
  • Persona and Purpose have moved into com.atlan.model.assets
  • These now directly provide createMetadataPolicy and createDataPolicy (and createGlossaryPolicy in Persona) helpers to manage the policies
  • AuthPolicy is the new base object for all policies (the helpers above now create AuthPolicy objects)
  • Policy management is now more consistent across the two mechanisms, and policies can now be created in bulk for both as well

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🥗 QOL improvements

• Refactors the code generators used to build the SDK, so that they can be reused to create "extended" SDKs. All generated code is now also annotated (javax.annotation.Generated) for easier identification and simplifying common tasks like excluding generated code from coverage tests.

(Note: this is really just a re-issue of the v0.7.0 release that fixes a JavaDoc error that caused publishing to Maven Central to fail.)

v0.6.0

🎉 New features

• New connector types (icons):
  • Azure Data Lake
  • Delta Lake
  • Mini SQL (mSQL)
  • Apache Iceberg
  • Apache Impala
  • Apache Spark SQL
  • MariaDB
  • Firebolt
  • Cloudera Data Warehouse
  • Starburst Galaxy
  • Redis
  • GraphQL
• Adds new type definitions for:
  • Redash objects
  • Monte Carlo objects
• Adds new crawlers for:
  • Microsoft SQL Server
  • Sigma
• Adds initial experimental exposure of Playbooks

⛑️ Breaking changes

• Deprecates the top-level meanings relationship, as it is not consistently present on responses. Instead use the assignedTerms relationship.
• Removes the parameter for custom metadata handling from the upsert() method. The behavior when true was very invasive without being obvious, so instead we have replaced it with two separate methods:
  • upsertMergingCM() will merge the provided custom metadata in a request with any existing custom metadata already on the asset. This is net-new functionality.
  • upsertReplacingCM() will overwrite all custom metadata on the asset. (Any custom metadata not provided in the request will result in that custom metadata being removed from the asset.) This is the same behavior as the original upsert()method when sending the replaceCustomMetadata parameter with a value of true.
  • The same distinction in handling is now available when using an AssetBatch as well. (The default behavior when unspecified remains as before: all custom metadata in a batch is ignored, by default.)
• Removes AbstractProcess and AbstractColumnProcess — instead, use LineageProcess and ColumnProcess directly.
• Simplifies the asset filtering provided to the DbtCrawler to be a string rather than series of complex nested objects.

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixes nested loops causing excessive retries on connection creation
• Fixes serde issues of complex embedded structures (internal maps and lists of maps)
• Fixes the outer meanings objects, which were previously missed out by deserialization
• Fixes an issue where connections were improperly created due to missing empty lists for admin roles/users/groups
• Fixes the replaceCustomMetadata() method, so it should now properly replace only the properties within the specified custom metadata set on an asset
• Fixes handling of deleted objects (like classifications and custom metadata) for audit entries, which will be unable to resolve the human-readable names of these objects once they've been deleted.
• Fixes deserialization of Atlan structs (complex nested objects). Previously these would always be deserialized as empty.

🥗 QOL improvements

• Adds an idempotent appendClassifications() method, which will only add classifications that are not already associated with an asset. (The existing options would throw errors if any provided classifications in a request overlapped with existing classifications on an asset, so were not idempotent.)
• Adds helpers for creating custom metadata with custom logos (whether emojis or entirely custom images)

v0.5.1

🎉 New features

• New connector types (icons):
  • AlloyDB
  • Aurora
  • Azure Analysis Services
  • Azure Cosmos DB
  • Clari
  • ClickHouse
  • CockroachDB
  • Firebird
  • Greenplum
  • Hex
  • IBM Informix
  • Kafka
  • Marketo
  • MonetDB
  • MongoDB
  • MuleSoft
  • Oracle TimesTen
  • Percona Server
  • QuickSight
  • Rockset
  • SAP IQ
  • SAP MaxDB
  • SAP SQL
  • SingleStore
  • SQLite
  • Teradata
  • ThoughtSpot
  • YugabyteDB
• Adds crawler for dbt Core

🐞 Bug fixes

• Adds metadata annotations to all crawlers
• Fixes integration tests for inclusion in CI/CD-driven daily regression testing
• Fixes bundled JRE for container image, to include all necessary base Java dependencies for logging

🥗 QOL improvements

• Adds an AtlanAsyncMutator interface to expose a block() method, to force waiting until asynchronous operations are completed. Currently implemented for both connection creation and asset deletion.
• Adds a more flexible addClassifications() method that allows controlling propagation settings of all provided classifications.

v0.5.0

🎉 New features

• Models for new asset types:
  • Kafka
  • Qlik
  • QuickSight
• New connector types:
  • Sigma
  • Synapse
  • Airflow
  • OpenLineage
  • Dataflow
  • Qlik Sense
• Models for events that are generated by Atlan
• Badges and badge conditions complete with creator(), factory methods, and related enumerations
• New AssetBatch class to abstract and encode processing batches of assets at the same time
• (Experimental) Initial objects and endpoints for requests handling
• (Experimental) Builds a shadow jar file that includes Numaflow, for potential event-based integration

⛑️ Breaking changes

• Renames the sqlAsset property to primarySqlAsset in DbtSource — to allow setting the new multi-valued sqlAssets property using the singular builder pattern
• Renames the MEANINGS searchable field enum to ASSIGNED_TERMS
• Aligns to latest ADLS model (removes attributes adlsAccountCreationTime, adlsContainerLastModifiedTime, adlsObjectCreationTime, adlsObjectLastModifiedTime
• S3 objects now require bucketQualifiedName and bucketName for creation
• Removes the updater() method for personas and purposes — these should instead always be retrieved first, updated, then sent back to Atlan as an update
• Makes name a mandatory parameter for the updateCertificate() and updateAnnouncement() operations on a glossary
• Enforces immutability of objects (must use the builder pattern):
  • Removes all setter methods (set...()) from the models.
  • Objects can only be built and modified now using the builder()s (and related creator() and updater() methods that return builders)
  • If you were previously modifying objects in-place, you now need to build an immutable copy of them to update them: asset.setDescription("abc123"); -> asset = asset.toBuilder().description("abc123").build();
  • Note that collection-based fields will only append using this pattern, to replace you first need to chain the equivalent clear...() method: term.toBuilder().clearCategories().category(GlossaryCategory.refByGuid("...")).build(); would replace all categories defined on the term with the single category provided.
• Adds the ability to limit a dbt crawler to a specific connection (an extra parameter in the fully-parameterized crawler method)

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixed an issue where an aggregation that had no content would cause an IOException during deserialization — now instead simply returns the empty aggregation result
• Fixes custom metadata creation, which now requires a name and description in all payloads (even if empty strings)
• Fixes an issue where the contents of a README were left unencoded during serialization
• Removes unused connectionName property when creating a lineage process
• Pass-through internal server error details (500 response codes)
• Fixes a problem where if a persona or purpose was not supplied with a default blank description for creation, a knock-on issue occurred when trying to add policies to those objects
• Invalidates and refreshes caches any time a type definition is created, updated, or deleted via the SDK
• Drops the optional entityGuid when setting classification assignments
• Fixes an issue where a trailing / on the base URL would have caused non-obvious failures (returning HTML instead of API responses)
• Changes the debugging output for internal server errors (>500 response codes), to differentiate them from permission errors (403 response codes) since both will retry
• Fixes an issue related to the deserialization of popularity details within event payloads
• Fixes an issue where a null classification assignment could cause an IOException by trying to resolve a classification with the name null

🥗 QOL improvements

• Adds an isArchived() helper for AttributeDefs, to make it easier to determine whether they are active or soft-deleted
• Adds an archive() helper for AttributeDefs, to abstract away the steps needed to archive (soft-delete) an attribute definition
• Adds a factory method for creating classification assignments without requiring entity details (Classification.of())
• Reduces logging level for API retries from INFO to DEBUG
• Adds excludeMeanings and excludeClassifications options for further optimizing searches that do not require these details in the results
• Adds generateQualifiedName() helpers to objects that have creators() and updaters(), to encode creating the appropriate qualifiedNames for these objects

v0.4.0

🎉 New features

• Adds the ability to provide your own unique ID for lineage processes, for more consistent updates (see below)

⛑️ Breaking changes

• LineageProcess creator method signature has changed, to:
  • Simplify its inputs — rather than requiring a connector type, connection name and connection qualifiedName you can now provide only the connection qualifiedName and the rest will be determined from this automatically.
  • Allow a unique ID to be provided — previously a unique ID was always generated for you, as a hash of the inputs and outputs of the process. By providing your own ID you can now directly match the unique ID in the orchestrating system. This will allow you to update the same lineage process over time, even if its inputs and outputs change, without creating new nodes in lineage. (Previously any change to the inputs or outputs of a process would have resulted in a different generated hash ID, and thus a different lineage process.) This remains an optional input, though, so if you send the argument as null a hashed ID will still be generated (as before).

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixed an issue where an aggregation that had no content would cause an IOException during deserialization — now instead simply returns the empty aggregation result

🥗 QOL improvements

• Adds helper method to reverse-engineer connector type from an asset's qualifiedName

v0.3.0

🎉 New features

• Adds simplified set of helpers to construct searches through the QueryFactory
• Adds validation of connection admins during both creation and update of a connection
• Adds validation of users and groups specified in purpose policies

⛑️ Breaking changes

• QueryFactory methods have changed to match the naming convention used for the simplification and the fact that searchable fields are now enumerated rather than simple strings:
  • active() -> beActive()
  • archived() -> beArchived()
  • withLineage() -> haveLineage()
  • withType() -> beOfType()
  • withSuperType() -> haveSuperType()
  • withCertificate() has been removed. Instead use have(KeywordFields.CERTIFICATE_STATUS).eq()
  • withExactName() has been removed. Instead use have(KeywordFields.NAME).eq()
  • whereQualifiedNameStartsWith() has been removed. Instead use have(KeywordFields.QUALIFIED_NAME).startingWith()
  • withAnyValueFor() has been removed. Instead use have(<field>).present()
  • withAtLeastOneClassification() -> beClassifiedByAtLeastOneOf(), which now takes a collection of human-readable classification names rather than internal hashed-string classification IDs
  • withAtLeastOneTerm() -> beDefinedByAtLeastOneOf()

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Fixed an issue where all attributes of type PopularityInsights were incorrectly (de)serialized
• Fixed an issue where a typedef with certain characters in its name (like spaces) could not be deleted

🥗 QOL improvements

• Enumerates all searchable fields to avoid typos or choosing the wrong indexed field for a particular type of search
• Simplified helpers in QueryFactory for building complex queries, aggregations, and sorting
• Simplified helpers in QueryFactory for retrieving results of metrics-based aggregations
• It should no longer be possible to create a connection that you are locked out of due to making a typo in the details of any roles, groups or users you set as connection admins
• It should no longer be possible to create a policy on a purpose that applies to non-existent users or groups

v0.2.0

🎉 New features

• Adds management of user membership in groups
• Adds recently released metadata types: ADLS, Sigma
• Adds Tableau crawler package
• Adds Athena crawler package
• Adds findByName helpers for retrieving glossary objects by their human-readable names
• Adds management of purposes

⛑️ Breaking changes

• Fixes the generation of qualifiedName for process assets, which is not backwards-compatible. (For any existing processes created via the SDK, you'll end up with duplicates due to the new qualifiedName. To avoid this, if possible, remove those existing processes and re-run using this updated SDK to generate the correct qualifiedNames, and then going forward these will be used and there will be no duplication.)

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Enumerates errors even when they are passed-through directly, to avoid an internal fatal exception related to formatting of error messages

🥗 QOL improvements

• Adds creator methods for policies to ensure minimal required information is always present
• Extends the base error class to capture details from differing error response formats

v0.1.0

🎉 New features

• Adds Snowflake Miner package
• Adds enumerations (options) for custom metadata
• Adds the purge deletion type, which can be applied to both active and archived assets
• Adds helpers to re-run existing workflows
• Adds latest model revisions, including usage / popularity metadata

⛑️ Breaking changes

• Flattens Entity and Asset — now all operations return Assets directly (anywhere you previously used Entity or EntityMutationResponse now needs to be replaced with their Asset-equivalent)
• Caches now throw exceptions (NotFoundException) when lookups fail, rather than simply returning null

(Note: we're reserving the right to make breaking changes to the SDK prior to a 1.0.0 release, without incrementing the major version. Once we hit 1.0.0, we'll adhere to https://semver.org)

🐞 Bug fixes

• Corrects include/exclude filters for crawlers

🥗 QOL improvements

• With the Entity -> Asset flattening, you no longer need to cast objects to retrieve things like qualifiedName, name, certificateStatus, and so on
• Enumerates all error messages with a unique code, and simplifies exception creation (default status code per exception)