Skip to content

Latest commit

 

History

History
207 lines (190 loc) · 11.8 KB

CHANGELOG.md

File metadata and controls

207 lines (190 loc) · 11.8 KB

0.9.8

  • Add Kafka adapter to send and receive data via Kafka individually per input/output. See docs.
  • Some frontend bug fixes

0.9.7

  • Log execution performance metrics
  • UI: Add deep links to (multiple) workflows and components
  • Allow to provide parsing options for manual input / direct_provisioning adapter for Pandas-like data. In particular this allows to choose other orients of the json representation, like "split". "split" is relevant for SERIES objects since it allows to enter series data with duplicate entries in the index, which is otherwise not possible.
    • BREAKING_CHANGE: For SERIES objects hetida designer now outputs the wrapped format with orient="split" parsing option to guarantee inclusion of duplicate indices for "direct_provisioning" output wirings.
  • Add url parameters to multiple trafo put and get endpoints for more fine-granular test wiring stripping. Use this for example when transfering trafos between instanced where adapters are not present on the target system
  • Several new components and example workflows
  • Add an --add parameter to hdctl which allows to add files to an existing export directory instead of overwriting the whole directory every time.

0.9.6

  • fix receiving duplicates with inlcude_dependencies param in trafo GET endpoints
  • expand_component_code now also supports adding test wirings to code with non direct_provisioning adapters
  • add instance support for non-sync commands in hdctl

0.9.5

  • Add hdctl sync feature, allowing for comfortable "hybrid development" of component code, i.e. switching frequently between working locally in your IDE and development in the hetida designer UI. Among other things this enables and streamlines several aspects of component/workflow development and operations tasks:
    • Keep your components/workflows versioned in a git repo.
    • Develop unit tests / doctests along with your components.
    • Develop component code in Jupyter.
    • Simplified component/workflow GitOps.
    • See the sync documentation for details.
  • Allow /transformations GET and PUT endpoints to emit / receive components as python code instead of json objects
  • new (possibly code-changing!) parameters to the /api/transformations GET endpoint that when set to true make sure that the component code contains everything (test wiring, documentation as module docstring, COMPONENT_INFO dict containing metadata)
  • BREAKING CHANGE: Importing components from python code which only contain the old @register decorator will no longer work. Component code must contain the new COMPONENT_INFO dictionary from now on instead in order to be importable from code via the various importing means. Importing from json files is not affected.
  • BREAKING CHANGE: All transformations in the open source repo have changed insignificantly: In component code, the @register decorater has been replaced by COMPONENT_INFO dictionaries. In addition, attributes that have already been added to the respective classes in previous releases are now added to the transformation JSONs as well. If you re-import them (note: this is not done automatically when updating the docker image) into an existing hd instance where you have persisted models you may be affected by deserialization problems and need to re-create those persisted models.
  • Bugfixes and improvements on handling default values. Default values can now include metadata.
  • some new base components

0.9.4

  • bug fixes around default value handling
  • BREAKING CHANGE: DRAFT transformation revisions with a released timestamp will no longer be accepted but cause a ValueError. A database migration fixing affected components and workflows is added. However if you export(ed) transformations with an earlier version than 0.9 and afterwards import them into a version >=0.9.4 the transformations may again include the bug. We therefore strongly recommend that you make a backup and/or export both before and after upgrading.

0.9.3

  • improve dashboarding token refresh
  • minor fixes and documentation improvements

0.9.2

  • BREAKING CHANGE: Fixed environment variable name HETIDA_DESIGNER_RUNTIME_ENGINE_URL (from mispelled HETIDA_DESIGNER_RUNTIME_EGINE_URL)
  • Add experimental dashboarding feature

0.9.1

0.9.0

  • add optional inputs with default parameters
  • add structured exception handling ⚠️ BREAKING CHANGE: The attribute error in the response JSON of the execute endpoint is no longer a string but a mapping/dictionary
  • add free text filters for outputs (adapter system)
  • add blob storage adapter config options
  • add general sql adapter with timeseries table support
  • add improved metadata attrs handling - in particular a manual input method
    • BREAKING CHANGE: Consequently the output json format for direct provisioning has changed to include metadata for dataframes, multitsframes and (time)series objects.
  • add python demo adapter sources and output free text filters
  • add additional validations
  • fix workflow validation issues via db migration
  • fix filter for categories in export
  • upgrade to python 3.11
  • upgrade dependencies
  • WARNING/BREAKING CHANGE: One of the introduced bugfixes includes a database migration fixing affected workflows. However if you export(ed) transformations with an earlier version than 0.9 and afterwards import them into a version >=0.9 the transformations may again include the bug. We therefore strongly recommend that you make a backup and/or export both before and after upgrading.

0.8.9

  • free text filters for inputs (adapter system)
  • add blob storage configuration options
  • improved blob storage adapter performance
  • type-specific blob storage serialization (keras models)
  • new data type: MultiTSFrames (collections of multiple timeseries with non-simultaneous timestamps)
  • api improvements / fixes
  • frontend adaption to new transformations endpoints.
  • NOTE: The endpoints deprecated since 0.7.* may be actually removed in one of the next releases
  • new workflows / components: When upgrading it is recommended to re-deploy the base components/workflows

0.8.8

  • add automatic bucket creation to blob storage adapter

0.8.7

  • add S3 compatible built-in blob storage adapter
  • add pickle persistence for ANY type to local file adapter
  • fix docker-compose reverse proxy config

0.8.6

  • (frontend) add configurable user info text on main tab
  • fixes, in particular around validation
  • fixes around updating components
  • add maintenance backend endpoints (deactivated by default)
  • add hdctl command line bash script for maintenance / devops tasks via those endpoints

0.8.5

  • fix NaN value serialization (should now be serialized to null json values)
  • fix broken pure plot execution config
  • more bug fixes
  • improvements for example workflows

0.8.4

  • fix outgoing auth config parsing

0.8.3

  • some new timeseries related base components and example workflows
  • improved export / import supporting clean-up operations and filters
  • more flexible authentication for outgoing requests
  • bug fixes

0.8.2

  • add some time measurement to successful execution output
  • update base images

0.8.1

  • improve contextualized logging
  • add async (web hook / callback) execution web endpoint
  • minor improvements and fixes

0.8.0

  • authentication via OpenID Connect. See documentation. When upgrading, you may have to explicitely deactivate auth via HD_USE_AUTH=false environment variables for runtime and backend service in your setup, since it is activated per default for both these services. See the changed default docker-compose setup!
  • automatic deployment of base components. Documented here.
  • usability improvements wiring / selection dialog
  • adapter system: directly attach any metadata as attributes to dataframes / series. See for example here
  • additional playwright end2end tests
  • bug fixes, refactorings

0.7.6

  • fix db config management bug
  • add more logging

0.7.5

  • fix db config password secret handling

0.7.4

  • fix DB user config environment variable name
  • small documentation fixes

0.7.3

  • Allow special characters in adapter source/sink ids
  • small bug fixes
  • documentation restructuring and improvements

0.7.2

  • Bugfixes concerning migration and importing
  • documentation fixes and improvements

0.7.1

  • Bug fixes!
  • you may now write async components by making your main function async
  • improve execution logging: now contains operator names and level/hierarchy
  • new endpoint to run latest revision of a revision group
  • restore the Kafka execution consumer and add improvements, in particular allow many configuration options to be set
  • Kafka execution also allows to execute latest revision of a revision group
  • add user interface end-to-end test suite (using playwright)

0.7.0

  • Complete Rewrite of the backend service in Python (formerly Java). This includes a lot of bug fixes.

⚠️ IMPORTANT: Upgrading from 0.6.* to 0.7 requires manual migration steps!

  • update frontend dependencies
  • export/import of components/workflows feature (see docs)

0.6.19

0.6.18

  • (security fix) upgrade java dependencies (see Issue #9)
  • upgrade Python dependencies
  • preparations for export / import feature
  • add docker build and push script replacing Travis build

0.6.17

  • (security fix) upgrade log4j to 2.16.0

0.6.16

Important: It is strongly recommended to upgrade designer installations to this version or higher due to the critical log4j security vulnerability known as "Log4Shell" (0-day Remote Code Execution)!

  • update log4j dependency (important security fix!)
  • fix workflow deployment
  • minor documentation updates

0.6.15

  • fix Issue #6
  • add documentation for postgres backup
  • add documentation for using R via rpy2

0.6.14

  • add output information to /workflows endpoint

0.6.13

  • minor fixes and improvements

0.6.12

  • update some dependencies
  • add ortools to default runtime dependencies

0.6.11

  • improve default timeout settings and add some documentation
  • add component export/import facilities from/to only a Python code file
  • extend component code generation to include information enabling export/import from just the component code.

0.6.10

  • remove buggy demo workflows

0.6.9

  • security updates dependencies
  • minor fixes and improvements

0.6.8

  • minor fixes and improvements

0.6.7

  • upgrade Python dependencies

0.6.6

  • switch/adapt to unprivileged docker images
  • add some more default Python dependencies to runtime
  • add info endpoints for liveness probes
  • minor documentation fix

0.6.5

  • add documentation for workflow execution via web endpoint

0.6.4

  • add built-in local file adapter to runtime

0.6.1, 0.6.2, 0.6.3

  • fix adapter documentation
  • fix travis build process (reduce log output to handle maximum log size limitations)
  • fix Issue #4

0.6.0

  • introducing the hetida designer adapter system