To create your first transformation, and to see how Transformations are an integral part of the Keboola Connection workflow, go to our Getting Started tutorial.
Transformations allow you to manipulate data in your project. They are the tasks you want to perform (Marketing data preaggregation, Tableau denormalizer, Integrity checker or Join marketing channels and sales, etc.).
In Keboola Connection all Transformations operate on a copy of the data from Storage in a safe and isolated environment. This improves safety, repeatability and traceability of all data operations. The process of isolating data is called mapping. The mapping process copies input data from permanent project Storage into a staging area which represents an ephemeral storage for the purpose of running the transformation script. The transformation script is executed by a transformation backend. To ease the development of the transformation script, all transformations are automatically versioned, and you can use a workspace which provides a copy of the environment and can be run interactively.
The schema below shows a high-level overview of transformations:
A transformation is represented by a Transformation Script (SQL, Julia, OpenRefine, Python, R) which you can use to manipulate your data. To ensure safety of the data in Storage, a transformation operates in a completely separate staging storage created for each transformation. When a transformation job runs, it takes the required data from the project Storage and copies them to a temporary staging storage. This process is called Input Mapping. The transformation job then runs the transformation script operating on the staging storage. When the transformation is finished, the Output Mapping process moves the transformation results back to the designated tables in Storage. The Input and Output mappings ensure complete safety of the transformation processes – the transformation always operates in an isolated workspace.
Input and Output Mapping — separates the source data from your transformation. Mapping creates a secure staging area with data copied from the Storage tables specified in the input mappings. Database table names and CSV file names in transformations are completely unrelated to names of tables in Storage. This means, for example, that you can rename tables in storage and it won’t break any of your transformations.
There are a number of staging options that influence the transformation script code, too. Typically, you will create an SQL transformation that works with data in a Snowflake database, or a Python script that works with CSV files on a “local” disk (local from the perspective of the script). But it is possible to have a Python script that works with data in a Synapse database or with data on an Azure Blob Storage (ABS). The transformations are very flexible, though all the combinations might not be available in all projects at all times.
The Transformation Script is a code that defines what happens with the data while taking the tables from Input Mapping, modifying them and producing the tables referenced in Output Mapping.
A backend is the engine running the transformation script. It is either a database server (Amazon Redshift, Snowflake, Microsoft Synapse on Azure Stack), or a language interpreter (Julia, OpenRefine, Python, R, Spark).
How to decide which backend is appropriate for each task? A rule of thumb is that SQL performs better for joining tables, filtering data, grouping and simple aggregations. Script languages are more suitable for processing one line at a time, raw data processing or custom analytical tasks. You’re free to use the tools that suit you best.
Each transformation within can use a different backend to perform the task with the most suitable tool and programming language. As some tasks are difficult to solve in SQL, feel free to step in with Python and finish the work with SQL again.
The following are the currently available backends:
Note: Switching between SQL backends requires updating the code into the corresponding SQL dialect.
Transformation scripts are stored in configurations and as such they share the common properties and functions. Particularly important is that all changes are automatically versioned and you can always rollback a configuration or copy a version to a new transformation.
You can easily develop transformations using the Workspace. It allows you to run and play with your arbitrary transformation scripts on the copies of your tables without affecting data in your Storage, or your transformations. You can convert a workspace to a transformation and vice-versa.
For Redshift and Synapse, you’ll get a separate database to which the data from input mapping can be loaded. You’ll obtain database credentials which you can use with a database client of your choice. For Snowflake you can do the same. In addition, we provide access to the Snowflake web interface. Therefore you can develop transformations without the necessity to download and install a database client.
For Julia, Python and R transformations, you can also use a workspace represented by an isolated JupyterLab instance.
The current and most-used transformations are now marked as legacy, they receive important updates (e.g., Python, R, Julia updates) but won’t receive any new features. Legacy Transformations will be available together with New Transformations for a long period of time to ensure smooth migration (which will start once all features from Legacy Transformations are available in New Transformations and there is a migration path). New transformations are not available in all projects at the moment, therefore it’s perfectly okay to use the Legacy Transformations for day-to-day work.
Some features of legacy transformations are not yet available in new transformations and some features will never be available in new transformations. The following table shows the feature status as of October 2020. New transformations are generally available on Azure-based stacks.
|Feature||Legacy Transformations||New Transformations|
|OpenRefine Transformations||✓||Not available|
|Snowflake Transformations Query timeout||900s (configurable per project via Support)||7200s (configurable per transformation)|
|Synapse Transformations||Not available||✓|
|Oracle Transformations||Not available||✓|
|Max number||One per user and type||Unconstrained|
|Lifecycle||Terminates after 5 days, can be extended manually||Sleeps after 1 hour of inactivity, can be resumed|
|Load data||Only when starting||✓|
|Unload data||Not available||✓|
|IDE||Jupyer Notebook or RStudio||JupyterLab|
|MLflow support||Not available||Preview|
|Spark support||Not available||Preview|
|Other features||Transformations buckets||✓||Not available|
|Shared code||Not available||✓|
|Code patterns||Not available||✓|
In the transformation overview, if you are seeing workspaces and shared code, you’re using new transformations:
In the transformation detail, if you’re seeing variables and shared code, you’re using new transformations:
In the transformation detail, if you’re seeing buckets and sandboxes, you’re using legacy transformations:
In the transformation detail, if you’re seeing dependent transformations, phases and sandbox you’re using legacy transformations:
New transformations behave like any other component. This means that they use the standard API to manipulate and run configurations and that it is possible to create your own transformation components.
Apart from that, new transformations come with Workspaces (previously named Sandboxes) which have many new features – such as loading and unloading data with a running workspace, resuming the workspace, Spark and MLflow support.
The transformation script can be organized into pieces that we call code, and these can be further organized into blocks. This allows you to somehow structure lengthy scripts. The structure provides no executional isolation — all code pieces execute sequentially in the same context in a top-down left-to-right direction. You can assign custom names to individual blocks and code pieces. You can reorder both the code elements and blocks by dragging them:
The above setup therefore executes: Company, Contact, Employee, Opportunity & Auxiliary, Activities, Opportunity & Contact as if it were a single script (SQL in this case). Splitting the code into scripts is in no way required, you can put your whole script in a single block and code.
Legacy Transformations are the current version of transformations available in most projects in the US and EU stacks. Each transformation bucket can contain any number of individual transformations. It should represent a logical set of operations you want to perform together.
Though marked as legacy, they are perfectly okay to be used for daily work, since they still receive updates, just no new features. Legacy Transformations are grouped into folders called Transformation buckets.
If you want to maintain the best forward compatibility, we recommend not to make complicated structures of phases and dependencies.
Phases in older projects allow multiple transformation steps to be run within a single workspace. Multiple steps with the same input mapping (sharing data) might save a bit of processing time, but ultimately, everything is less clear and isolated. To save time, run multiple orchestration tasks in parallel instead.
Dependencies allow you to chain transformation steps in older projects. A given transformation is executed after all required steps have been executed.