v4.2.0

🎉 New features

• Adds latest typedefs by @cmgrote in #1191
• Adds creator for Procedure by @ErnestoLoma in #1194
• Optimizes cache refreshing to minimize unnecessary cache thrashing by @cmgrote in #1198

🐞 Bug fixes

• Fixes table partition creators by @cmgrote in #1191
• Adds more meaningful error message on invalid enum values by @cmgrote in #1187
• Fixes attempts to send empty payloads when no tag is found by @cmgrote in #1188
• Fixes deadlock on data domain caching by @cmgrote in #1195
• Fixes creation details for multi-valued custom metadata attributes by @cmgrote in #1197

Full Changelog: v4.1.1...v4.2.0

v4.1.1

🎉 New features

• Adds dynamic wait based on rate-limit response header by @cmgrote in #1166
• Adds SnowflakeDynamicTable as possible parentType for columns by @ErnestoLoma in #1132
• Adds get operation to retrieve selective relationships and attributes by @cmgrote in #1170
• Adds support for Anaplan assets by @prateekrai-atlan in #1177
• Adds four new AWS connectors by @rupeshatlan in #1179

🐞 Bug fixes

• Modifies workflow findByType to use regular expression to avoid false positive results by @ErnestoLoma in #1183

📦 Packages

• Fixes typedef toolkit to generate appropriate AttributeDefs on RelationshipDefs by @cmgrote in #1174
• Allows OpenAPI spec loader to support loading multiple JSON files by @pavanmanishd in #1113
• Adds interchangeable Excel and CSV options for export packages by @cmgrote in #1181
• Changes Python package rendering to support Open Telemetry logging by @ErnestoLoma in #1182

New Contributors

• @pavanmanishd made their first contribution in #1113

Full Changelog: v4.1.0...v4.1.1

v4.1.0

🎉 New features

• Adds logging of underlying causes for back-end errors by @cmgrote in #1098
• Adds new connector BIGID by @rupeshatlan in #1112

📦 Packages

• Adds OpenTelemetry logging for all Kotlin-based packages by @cmgrote in #1114
• Adds an uncaught exception handler so any unexpected errors are caught and logged by @cmgrote
• Fixes cyclical relationship detection in asset import by @cmgrote in #1107
• Fixes persistent connection cache population by @cmgrote in #1110
• Improves logging on any CSV decoding errors by @cmgrote in #1109
• Updates packages to use latest version of image by @ErnestoLoma in #1131

🥗 QOL improvements

• Parallelizes package tests by @cmgrote in #1117

New Contributors

• @rupeshatlan made their first contribution in #1112

Full Changelog: v4.0.3...v4.1.0

v4.0.3

📦 Packages

• Fixes bug where URL-based ingestion of OpenAPI spec would fail by @cmgrote in #1097
• Changes level to DEBUG for logger_root in Python packages by @ErnestoLoma in #1094

Full Changelog: v4.0.2...v4.0.3

v4.0.2

🐞 Bug fixes

• Fixes potential NPE in user listing iteration by @cmgrote in #1087
• Fixes bug that attempted to process non-existent input files by @cmgrote in #1095

Full Changelog: v4.0.1...v4.0.2

v4.0.1

This is essentially a rerelease of 4.0.0, so full release notes included here for simplicity:

⛑️ Breaking changes

• Removes a number of static objects, in particular the "default" Atlan client that was accessible through Atlan.getDefaultClient() by @cmgrote in #1072. Every operation that interacts with Atlan (ultimately sending it an API request) must now be explicitly passed an AtlanClient. This change was made for two primary reasons:
  1. It makes explicit which operations will interact with Atlan (those that require an AtlanClient to be passed) vs those that only operate on information locally in-memory (those that do not require an AtlanClient to be passed).
  2. It removes concurrency limitations and risks with shared static resources across the entire JVM (including all threads therein). This applies not only to the default client but also to some of the caches that were being automatically managed.
• Packages developed using this release of the package toolkit will now need to make use of a broader PackageContext object that similarly encapsulates the client and any extra caches specific to packages.

🧪 Experimental

• Adds limited ability to configure an SSO provider by @cmgrote in #1069

🐞 Bug fixes

• Fixes validations when creating a README by @cmgrote in #1059

📦 Packages

• Adds multi-pass handling of cyclical relationships to asset import package by @cmgrote in #1061
• Fixes Python outputs generated by package toolkit by @ErnestoLoma in #1068
• Adds the ability in the package toolkit to set default values into the configuration objects by defining a fallback for any inputs in the package.pkl by @cmgrote in #1072
• Fixes impersonation within packages by @cmgrote in #1084
• Adds further type-checking on input CSVs by @cmgrote in #1086
• Fixes potential NPE in user iteration listing by @cmgrote in #1087

🥗 QOL improvements

• Migrates bulk streaming of search to use search_after by @cmgrote in #1060

Full Changelog: v3.1.2...v4.0.1

AtlanClient changes

As mentioned in the breaking changes, this release introduces a complete revamp of the way the AtlanClient is managed and passed around to operations that actually call Atlan APIs behind-the-scenes.

Resource-isolated

Previously the client itself and some caches were managed statically across the entire JVM. This could cause non-obvious and very difficult reproduce issues when heavy concurrency exists. With this release, all caches are managed within the AtlanClient directly, and there is no longer a JVM-wide static client. Instead, you can create a new client and have it "live" only for as long as you need it:

try (AtlanClient client = new AtlanClient()) {
    // Do anything you need against Atlan within this block
}
// Once you exit the block, the client and all its internal caches will be removed

Must be explicitly passed

Furthermore, any operation that actually interacts with Atlan directly (ultimately via its APIs behind-the-scenes) must now be explicitly passed such a client:

GlossaryTerm term = GlossaryTerm.findByName(client, "A", "B");
term.trimToRequired().userDescription("Hi there!").build().save(client);

This means there is an extensive set of changes to many of the methods across the SDK — if you're planning to adopt this release (and higher), rather than reading a huge list of such operations it'll probably be simplest to update your dependency to this latest release in your codebase and then look for any compilation breakages. The explicit AtlanClient is almost always expected as the new first parameter in any methods that now require it.

Package toolkit

Introduction of PackageContext

In the same way we have resource-isolated the client and its caches for the SDK, we have resource-isolated the context of a running package and its further caches in this release. The full context needed for running a package is now bundled into an object called a PackageContext, which can be created from a package's configuration using Utils.initializeContext() and passing it te configuration. Where you previously may have just used this pattern:

@JvmStatic
fun main(args: Array<String>) {
    val config = Utils.setPackageOps<OpenAPISpecLoaderCfg>()
    // Receive inputs from the config to decide what to do in your logic
}

You should now use this pattern:

@JvmStatic
fun main(args: Array<String>) {
    Utils.initializeContext<OpenAPISpecLoaderCfg>().use { ctx ->
        // Receive inputs from the config to decide what to do in your logic
        val url = ctx.config.specUrl // config items will be available nested under ctx.config
        val client = ctx.client // the AtlanClient will be nested under ctx.client
        val termCache = ctx.termCache // a number of lazy-load-managed caches are also available in the context
    }
}

v4.0.0

⛑️ Breaking changes

• Removes a number of static objects, in particular the "default" Atlan client that was accessible through Atlan.getDefaultClient() by @cmgrote in #1072. Every operation that interacts with Atlan (ultimately sending it an API request) must now be explicitly passed an AtlanClient. This change was made for two primary reasons:
  1. It makes explicit which operations will interact with Atlan (those that require an AtlanClient to be passed) vs those that only operate on information locally in-memory (those that do not require an AtlanClient to be passed).
  2. It removes concurrency limitations and risks with shared static resources across the entire JVM (including all threads therein). This applies not only to the default client but also to some of the caches that were being automatically managed.
• Packages developed using this release of the package toolkit will now need to make use of a broader PackageContext object that similarly encapsulates the client and any extra caches specific to packages.

🧪 Experimental

• Adds limited ability to configure an SSO provider by @cmgrote in #1069

🐞 Bug fixes

• Fixes validations when creating a README by @cmgrote in #1059

📦 Packages

• Adds multi-pass handling of cyclical relationships to asset import package by @cmgrote in #1061
• Fixes Python outputs generated by package toolkit by @ErnestoLoma in #1068
• Adds the ability in the package toolkit to set default values into the configuration objects by defining a fallback for any inputs in the package.pkl by @cmgrote in #1072

🥗 QOL improvements

• Migrates bulk streaming of search to use search_after by @cmgrote in #1060

Full Changelog: v3.1.2...v4.0.0

AtlanClient changes

As mentioned in the breaking changes, this release introduces a complete revamp of the way the AtlanClient is managed and passed around to operations that actually call Atlan APIs behind-the-scenes.

Resource-isolated

Previously the client itself and some caches were managed statically across the entire JVM. This could cause non-obvious and very difficult reproduce issues when heavy concurrency exists. With this release, all caches are managed within the AtlanClient directly, and there is no longer a JVM-wide static client. Instead, you can create a new client and have it "live" only for as long as you need it:

try (AtlanClient client = new AtlanClient()) {
    // Do anything you need against Atlan within this block
}
// Once you exit the block, the client and all its internal caches will be removed

Must be explicitly passed

Furthermore, any operation that actually interacts with Atlan directly (ultimately via its APIs behind-the-scenes) must now be explicitly passed such a client:

GlossaryTerm term = GlossaryTerm.findByName(client, "A", "B");
term.trimToRequired().userDescription("Hi there!").build().save(client);

This means there is an extensive set of changes to many of the methods across the SDK — if you're planning to adopt this release (and higher), rather than reading a huge list of such operations it'll probably be simplest to update your dependency to this latest release in your codebase and then look for any compilation breakages. The explicit AtlanClient is almost always expected as the new first parameter in any methods that now require it.

Package toolkit

Introduction of PackageContext

In the same way we have resource-isolated the client and its caches for the SDK, we have resource-isolated the context of a running package and its further caches in this release. The full context needed for running a package is now bundled into an object called a PackageContext, which can be created from a package's configuration using Utils.initializeContext() and passing it te configuration. Where you previously may have just used this pattern:

@JvmStatic
fun main(args: Array<String>) {
    val config = Utils.setPackageOps<OpenAPISpecLoaderCfg>()
    // Receive inputs from the config to decide what to do in your logic
}

You should now use this pattern:

@JvmStatic
fun main(args: Array<String>) {
    Utils.initializeContext<OpenAPISpecLoaderCfg>().use { ctx ->
        // Receive inputs from the config to decide what to do in your logic
        val url = ctx.config.specUrl // config items will be available nested under ctx.config
        val client = ctx.client // the AtlanClient will be nested under ctx.client
        val termCache = ctx.termCache // a number of lazy-load-managed caches are also available in the context
    }
}

v3.1.2

🐞 Bug fixes

• Fixes bug that caused full delta calculation in relational and cube assets builders to miss net-new assets by @cmgrote in #1055

Full Changelog: v3.1.1...v3.1.2

v3.1.1

🎉 New features

• Adds immediateNeighbors to FluentLineage to simplify lineage retrieval by @pavanmanish-atlan in #1048

🐞 Bug fixes

• Fixes bug with automated token exchange on expired API token retries by @cmgrote in #1043

📦 Packages

• Corrects downloads for GCS and ADLS deltas by @cmgrote in #1041
• Adds full delta calculation logic for relational and cube assets builders by @cmgrote in #1044
• Prevents renaming of connections as part of the enrichment migrator by @ErnestoLoma in #1049

New Contributors

• @pavanmanish-atlan made their first contribution in #1048

Full Changelog: v3.0.1...v3.1.0

v3.0.1

🎉 New features

• Adds username and extra (non-sensitive) fields to credential response by @cmgrote in #1026

🧪 Experimental

• Adds application typedefs and further data model typedef enhancements by @cmgrote in #1035

🐞 Bug fixes

• Corrects typo that caused CategoryCache to be loaded incorrectly by @cmgrote in #1030
• Avoids page overruns in edge cases where total results is very nearly the limit before query rewriting needs to happen by @cmgrote in #1033

📦 Packages

• Adds case insensitivity and table view agnosticism options to enrichment migrator by @ErnestoLoma in #1032
• Fixes bug where a purpose with no tag could cause admin export to fail by @cmgrote in #1036

Full Changelog: v3.0.0...v3.0.1