TRAC 0.2.0 beta 1

This first beta of the 0.2 series is to provide early sight of the Python model runtime. It includes initial versions of the core elements of the executor framework and will run the Pandas example model included in the repository doc folder. PySpark support is not included and there are gaps around validation and type handling. Documentation is limited.

• TRAC model API for Python
• TRAC runtime executor framework for Python models
• Support for non-partitioned datasets as Pandas data frames
• Support for dev-mode, which allows running models using concise, developer-friendly config
• Metadata updates:
  • Added StorageDefinition to describe physical storage of data items (physical and logical attributes evolve separately)
  • Removed physical storage attributes from DataDefinition, replaced with a reference to StorageDefinition
  • Changed DataDefinition to describe data items using part/snap/delta structure (only a single root partition is used)

TRAC 0.1.1

TRAC 0.1.1 adds a number of core API and metadata enhancements to the metadata service. The intention is to provide core metadata features and structure as early as possible in the development timeline, to make them available for building other components and reduce disruption from late metadata changes. This release mainly addresses the tagging layer of the metadata model, object definitions will be addressed as the platform capabilities enabled by those definitions are added.

Version 0.1.1 is not metadata-compatible with version 0.1.0.

• Tag update API - Use a series of TagUpdate operations to add, update and remove tag attributes (replaces the "PUT resource" style API in version 0.1.0)
• Universal definition of time - Timestamps are assigned to every object and tag, "as-of" criteria are available for every read and search operation
• Add tag selectors - Selectors can be used to address any object/tag using explicit version numbers, as-of times or the latest version
• IDs and headers - ID/header information is moved into tags, so objects can be created cleanly before their ID is known
• Add MAP type to the available metadata value types

TRAC 0.1.0

First release of Accenture TRAC, intended for reference and experimentation.

• First draft of TRAC metadata structures (.proto files)
• First draft of service API for the metadata service (.proto files)
• Metadata service implementation using gRPC
• Metadata DAL implementation using JDBC, initial support for 6 SQL dialects:
  MySQL, MariaDB, PostgreSQL, Microsoft SQL Server, Oracle, H2
• Platform Gateway - Proxies the metadata API, exposes REST / json over HTTP/1.1
• Shared ConfigManager component, loading is delegated to config loader plugins
• Config loader implementation for loading from local files
• Logging and error reporting (standard errors reported over gRPC and REST)
• Deploy tool for managing schema and tenants in the metadata database
• Initial documentation included in README.md, .proto files and doc/examples
• Unit and integration testing using GitHub actions
• Licensed under the Apache license, version 2.0