Updated methods to ensure proper cleanup when removing activity execution contexts, including handling parent-child relationships. Improved robustness and clarity by introducing step-by-step removal for multiple contexts.
Replaced inline filtering logic with reusable methods to simplify and unify activity execution context management. Introduced `ClearCompletedActivityExecutionContexts` to remove redundant contexts and migrated filtering logic to `WorkflowExecutionContext`. Improved code clarity and maintainability by removing duplicate methods and streamlining exception throwing.
Replaced `ActivityExecutionRecordExtractor` with a `ChangeTrackingDictionary` for better state mutation tracking in activity execution contexts. Introduced tainting mechanisms to track dirty states and ensure precise logging and persistence of activity execution logs. Updated several components like `ActivityExecutionLogSink` and `ExpressionExecutionContext` to utilize these changes effectively.
The PreviousSubStatus assignment was removed as it was unused and unnecessary. This cleanup simplifies the code and ensures focus only on relevant functionality.
Removed obsolete middleware for persisting bookmarks, execution logs, and variables, integrating their functionality into the commit state handler. Added early exit checks for empty collections in persistence methods. Updated activity invoker logic for better state commit handling during execution.
Reintroduced `GetAncestors` method and added new methods: `GetDescendants`, `GetActiveChildren`, and `GetChildren`. These methods enhance navigation through activity execution contexts by providing easy access to hierarchical relationships like ancestors, descendants, and children.
Introduced `ActivityCommitStateBehavior` and `WorkflowCommitStateOptions` to enable flexible state commit handling in workflows. Integrated commit logic into activity and workflow execution contexts and middleware. This improves control over when state is committed during workflow execution.
Resolved unnecessary whitespace issues and adjusted spacing around conditional statements to improve code readability and maintain consistency. These changes do not alter functionality but enhance the code's clarity and professional formatting.
Updated the method to safely access the `Value` property when retrieving the first record, preventing potential null reference exceptions. This ensures more robust and error-free behavior when querying output records.
Replaced default value assignments with null for clarity and simplicity. Refactored ActivityOutputRegister to optimize record storage and retrieval using grouped dictionary entries instead of flat lists. Adjusted related methods to improve performance and maintain consistency.
Enhanced `ActivityOutputRegister` with dictionary-based lookups for improved performance and added unique key generation methods to efficiently retrieve outputs.
Introduce a `DisableVariableCopying` option to improve performance by preventing workflow variables from being copied into the Jint engine or back into the workflow context. Updated related logic to honor this setting and ensure compatibility with existing behavior.
Replaced `GetActiveChildren` with a direct `Children` property in `Flowchart.cs` to streamline logic. Removed redundant `GetDescendents`, `GetActiveChildren`, and `GetChildren` methods from `ActivityExecutionContextExtensions.cs`. Also updated GitHub workflows to allow performance-related branches.
Replaced usage of `GetDescendents` with `Children` property for better clarity and efficiency when checking running activity instances. This simplifies the logic and aligns with the existing structure of activity context handling.
Optimized activity execution context management by introducing parent-child relationships and improving immutability. Adjusted several APIs to enhance clarity, performance, and maintainability, including the use of `AsReadOnly` collections and removal of redundant code.
Refactored `ExceptionHandlingMiddleware` to improve readability and modularity by splitting responsibilities into smaller methods. Replaced `GetAggregateStatus` with direct use of `source.Status` for significant performance improvement when a large number of activity instances are involved. Updated null-checks for clarity and fixed inconsistent usage of default values.
Replaced `default` with `null` for optional parameters to improve code clarity and maintain consistency. This change ensures better readability and aligns with common coding practices, especially when null is the intended default value.
Updated `VariableMapper` for null assignment consistency and streamlined `VariableModel` instantiation. Improved `TypeJsonConverter` to handle list type aliases more explicitly, replacing ambiguous syntax with clearer format.
Refactored TypeJsonConverter to distinguish and properly handle serialization of arrays and generic collections. Updated integration tests to include cases for round-tripping primitive arrays and collections for improved coverage.
Added support for configuring Hangfire job storage per database provider, including PostgreSql, SQLite, and SQL Server. Introduced new methods for flexible Hangfire setup, while marking older APIs and storage configuration extensions as obsolete. Refactored related configurations for streamlined and centralized job scheduling logic.
Updated the logic to use `MakeArrayType` for array handling instead of generic collection types, ensuring consistency and better alignment with expected type structures. Adjustments were made in both the `VariableDefinitionMapper` and the `TypeJsonConverter`.
Replaced explicit type initializations with concise target-typed `new()` expressions where applicable, improving code readability and aligning with updated C# conventions. Updated .editorconfig to enforce consistent use of `var` for type declarations.
Replaces static resolver state with an instance-level resolver in `CrossScopedReferenceHandler` to enhance flexibility and thread-safety. Refactors `JsonWorkflowStateSerializer` to use `GetOptions` for improved API consistency and clarity.
Replaced `PerCallReferenceHandlerWrapper` with `CrossScopedReferenceHandler` and optimized the resolver initialization using `AsyncLocal`. This simplifies the code and ensures a more efficient handling of reference resolution during serialization.
The `ApplyOptions` method now correctly calls the base implementation before applying custom configurations. This ensures that any base behavior is preserved, preventing potential issues with serialization options.
Replaced direct use of `CrossScopedReferenceHandler` with `PerCallReferenceHandlerWrapper` for improved abstraction and reusability. Updated `ApplyOptions` to leverage the new wrapper, simplifying reference resolver management. Commented out redundant code to streamline the implementation.
- **Improves performance:** The converters and static properties are pre-configured when the class is initialized. Only the `ReferenceHandler` is refreshed per call, avoiding repeated configuration of common options like converters.
- **Thread-safe:** Since `_cachedOptions` is immutable, it can safely be reused across threads.
**Tradeoffs:** Small performance cost during cloning, but still much faster than fully recreating options each time.
Added try-catch block to log failures when reading variables from storage, ensuring robust error handling. Also refined type checking for `ExpandoObject` deserialization in `ObjectConverter` to prevent invalid operations.
Added logic to cancel child activities during activity cancellation. Removed redundant bookmark removal logic in the `Fork` activity for better clarity and efficiency.
Refactored `StorageDriverContext` to include `Variable` and updated its usage across relevant methods to ensure context accuracy. Enhanced `ObjectConverter` to support conditional deserialization of JSON objects. Adjusted `WorkflowInstanceStorageDriver` to utilize improved variable resolution with new conversion options.
Updated the log message to use the full variable type name instead of the base type. This provides more detailed context for debugging failed variable parsing issues.
Updated the exception message to use the full type name of the variable, providing clearer details for debugging type compatibility issues during variable value parsing.
Added `TryParseValue` method to handle parse errors gracefully and prevent crashes. Updated `VariablePersistenceManager` to log warnings when variable parsing fails, providing better debugging support and resilience during workflow execution.
Commented out state filtering logic to address issues with log sanitization. The raw value is now directly stored in the activity state until a redesigned solution is implemented.
Remove the "Deprecated" label from the display name for clarity while keeping the `Obsolete` attribute to indicate planned deprecation. This ensures better alignment between the display name and actual usage guidance.
This was broken by both #5871 and #5682:
1. When a List<Guid> was serialized, it was recognized as a primitive collection and thus plainly written to the JSON without any type information: ["d4d8404c-4357-47ff-a343-649a116539f5"]
2. When this JSON was deserialized, due to lack of type info, it was deserialized as List<object>, containing strings. This is already not good.
3. When this List<object> gets serialized again (e.g. due to multiple workflow suspends causing WorkflowState serialization), this time it fails the primitive collection recognition, because object is not a primitive type. It now gets serialized as {"_items": ["d4d8404c-4357-47ff-a343-649a116539f5"], "_type": "Object[]"}
4. When that JSON gets deserialized, it tries to ReadType() but ReadType() fails to parse Object[] since it lacks the logic from TypeJsonConverter to throw away the [] before looking up Object in the WellKnownTypeRegistry, so it returns null as a type. Without a type but being faced with a json object { ... } it now deserializes into an ExpandoObject
5. Any further serialization / deserializations will now cause the expando object to get nested deeper and deeper every time.
Moved WorkflowStorageDriver and WorkflowInstanceStorageDriver to a new namespace, VariableStorageDrivers, to improve organization. Updated namespaces in related services and ensured backward compatibility with comments. Minor fix made to TypeJsonConverter for consistency and additional configuration adjustments to support the new structure.
* Add log persistence configuration and strategy evaluation
Introduced `LogPersistenceConfiguration` to support dynamic strategy and expression-based log persistence evaluations. Updated `Default default values.
* Add LogPersistenceConfiguration and enum support
Introduced LogPersistenceConfiguration class and LogPersistenceEvaluationMode enum in both Elsa.Api.Client and Elsa.Workflows.Runtime. Updated JavaScript services to handle enums correctly and register LogPersistenceMode.
* Rename log persistence key for consistency
Updated the log persistence key from `LogPersistenceStrategyKey` to `LogPersistenceConfigKey` to maintain consistency across the codebase. This change ensures that property access aligns with the updated naming conventions used in the application's configuration.
* Reuse JSON serializer options
* Update log persistence config structure in comments
This commit revises the JSON example in code comments to reflect the updated structure of the log persistence configuration. The changes include updated evaluation modes and strategy types for default, inputs, and outputs sections. These modifications aim to enhance clarity and provide accurate documentation of the expected configuration format.
