EtlKit

EtlKit is a fully open-source (MIT) ETL library and data integration toolbox for .NET.

This project originated as a fork of the original ETLBox library by Andreas Lennartz. Starting with version 2.0, the original author decided to close the source and commercialize its newer branch. EtlKit continues the open-source 1.x lineage, keeping it up to date with modern .NET and database drivers while adding new features.

The 2.0 rename: EtlBox.Classic → EtlKit

⚠️ Breaking change in 2.0 ⚠️

As of version 2.0, the project formerly known as EtlBox.Classic has been renamed to EtlKit. This avoids versioning and package naming conflicts with the original closed-source ETLBox library.

The rename is comprehensive — NuGet package IDs, .NET root namespaces, the solution file, and DI registration extension methods have all changed. The class names and public API surface of every component are otherwise identical to 1.x; this is purely a rename, not a redesign.

What changed:

• NuGet packages: EtlBox.Classic.* → EtlKit.* (e.g. EtlBox.Classic → EtlKit, EtlBox.Classic.Kafka → EtlKit.Kafka)
• Root namespaces: ALE.ETLBox.* → EtlKit.* (the ALE. prefix is dropped entirely)
• DI extensions: AddEtlBoxCore() → AddEtlKitCore(), AddEtlBoxJson() → AddEtlKitJson(), etc.
• Exception type: EtlBoxException → EtlKitException
• Solution file: ETLBox.sln → EtlKit.sln

Migrating from 1.x to 2.0:

1. Update all <PackageReference> entries from EtlBox.Classic.* to EtlKit.*.
2. Find-and-replace using ALE.ETLBox → using EtlKit across all .cs files.
3. Find-and-replace EtlBoxException → EtlKitException.
4. Find-and-replace AddEtlBox → AddEtlKit in DI registration calls.
5. Update solution references from ETLBox.sln to EtlKit.sln.

See the CHANGELOG for the full mapping table of old → new package IDs and namespaces.

Documentation

• Guides and tutorials — docs/ contains hand-written guides and tutorials, browsable directly in the repo
• API Reference — docfx/ contains the DocFx config that generates the hosted API reference from XML comments

Installation

EtlKit targets .NET Standard 2.0, so it works in any .NET project that supports it (basically all modern .NET versions).

EtlKit is available on NuGet. Add the package to your project via your NuGet package manager.

See individual package descriptions to understand what each one does. For a basic setup you only need EtlKit.

What is EtlKit

A lightweight ETL (extract, transform, load) library and data integration toolbox for .NET. Source and destination components let you read and write data from the most common databases and file types. Transformations allow you to harmonize, filter, aggregate, validate and clean your data.

Create your own tailor-made data flow with your .NET language of choice. EtlKit is written in C# and offers full support for .NET Core.

Why EtlKit

EtlKit is a comprehensive C# class library that is able to manage your whole ETL or ELT. You can use it to create your own dataflow pipelines programmatically in .NET, e.g. with C#. Besides a big set of dataflow components it comes with some control flow tasks that let you easily manage your database or simply execute SQL code without any boilerplate code. It also offers extended logging capabilities based on ILogger to monitor and analyze your ETL job runs.

EtlKit is a fully functional alternative to other ETL tools like Sql Server Integration Services (SSIS). Creating your ETL processes programmatically has some advantages:

Build ETL in .NET: Code your ETL with your favorite .NET language, fitting your team's skills and leveraging a mature toolset.

Runs everywhere: EtlKit runs on Linux, macOS, and Windows. It is written in .NET Standard and successfully tested with the latest versions of .NET Core & .NET.

Run locally: Develop and test your ETL code locally on your desktop using your existing development & debugging tools.

Process In-Memory: EtlKit comes with dataflow components that allow in-memory processing which is much faster than storing data on disk and processing later.

Manage Change: Track your changes with git (or other source controls), code review your ETL logic, and use your existing CI/CD processes.

Data integration: While supporting different databases, flat files and web services, you can use EtlKit as a foundation for your custom-made Data Integration platform.

Made for big data: EtlKit relies on Microsoft's TPL.Dataflow library and was designed to work with big amounts of data.

Data Flow and Control Flow

EtlKit is split into two main components: Data Flow and Control Flow Tasks. The Data Flow part offers the core ETL components. The tasks in the Control Flow allow you to manage your databases with a simple syntax. Both components come with customizable logging functionalities.

Data Flow overview

EtlKit comes with a set of Data Flow components to construct your own ETL pipeline. You can connect with different sources (e.g. a Csv file), add some transformations to manipulate that data on-the-fly (e.g. calculating a sum or combining two columns) and then store the changed data in a connected destination (e.g. a database table).

To create your own data flow , you basically follow three steps:

• First you define your dataflow components (sources, optionally transformations and destinations)
• link these components together
• tell your source to start reading the data and wait for the destination to finish

Now the source will start reading and post its data into the components connected to its output. As soon as a connected component retrieves any data in its input, the component will start with processing the data and then send it further down the line to its connected components. The dataflow will finish when all data from the source(s) are read, processed by the transformations and received in the destination(s).

Transformations are not always needed - you can directly connect a source to a destination. Normally, each source has one output, each destination one input and each transformation at least one input and one or more outputs.

Of course, all data is processed asynchronously by the components. Each compoment has its own set of buffers, so while the source is still reading data, the transformations can already process it and the destinations can start writing the processed information into their target. So in an optimal flow only the current row needed for processing is stored in memory. Depending on the processing speed of your components, the buffer of each component can store additional rows to optimize throughput.

Data Flow example

It's easy to create your own data flow pipeline. This example data flow will read data from a MySql database, modify a value and then store the modified data in a Sql Server table and a csv file, depending on a filter expression.

Step 1 is to create a source, the transformations and destinations:

var sourceCon = new MySqlConnectionManager("Server=10.37.128.2;Database=EtlKit_ControlFlow;Uid=etlkit;Pwd=etlkitpassword;");
var destCon = new SqlConnectionManager("Data Source=.;Integrated Security=SSPI;Initial Catalog=EtlKit;");

DbSource<MySimpleRow> source = new DbSource<MySimpleRow>(sourceCon, "SourceTable");
RowTransformation<MySimpleRow, MySimpleRow> rowTrans = new RowTransformation<MySimpleRow, MySimpleRow>(
    row => {  
        row.Value += 1;
        return row;
    });
Multicast<MySimpleRow> multicast = new Multicast<MySimpleRow>();
DbDestination<MySimpleRow> sqlDest = new DbDestination<MySimpleRow>(destCon, "DestinationTable");
CsvDestination<MySimpleRow> csvDest = new CsvDestination<MySimpleRow>("example.csv");

Now we link these elements together.

source.LinkTo(trans);
rowTrans.LinkTo(multicast);
multicast.LinkTo(sqlDest, row => row.FilterValue > 0);
multicast.LinkTo(csvDest, row => row.FilterValue < 0);

Finally, start the dataflow at the source and wait for the destinations to rececive all data (and the completion message from the source).

source.Execute();
sqlDest.Wait();
csvDest.Wait();

Data integration

The heart of an ETL framework is it's ability to integrate with other systems. The following table shows which types of sources and destination are supported out-of-the box with the current version of EtlKit. You can always integrate any other system not listed here by using a CustomSource or CustomDestination - though you have to write the integration code yourself.

Source or Destination	Support for	Limitations
Databases	Sql Server, Postgres, SQLite, MySql, Clickhouse	Full support
Queues and streaming	Kafka, RabbitMQ	Kafka — full support, RabbitMQ — destination only
Flat files	Csv, Json, Xml	Full support
Office	Microsoft Access, Excel	Full support for Access, Excel only as source
Cube	Sql Server Analysis Service	Only XMLA statements
Memory	.NET IEnumerable & Collections	Full support
API	REST, OpenAI	Full support
Cloud Services	Tested with Azure	Full support
Any other	integration with custom written code	No limitations

You can choose between different sources and destination components. DbSource and DbDestination will connect to the most used databases (e.g. Sql Server, Postgres, MySql, SQLite). CsvSource, CsvDestination give you support for flat files - based on CSVHelper. ExcelSource allows you to read data from an excel sheet. JsonSource, JsonDestination, XmlSource and XmlDestination let you read and write json from files or web service request. MemorySource, MemoryDestinatiation as well as CustomSource and CustomDestination will give you a lot flexibility to read or write data directly from memory or to create your own custom made source or destination component.

Transformations

EtlKit has 3 types of transformations: Non-blocking, partially blocking and blocking transformations. Non-blocking transformations will only store the row that is currently processed in memory (plus some more in the buffer to optimize throughput and performance). Partially blocking transformations will load some data in the memory before they process data row-by-row. Blocking transformations will wait until all data has arrived at the component before it starts processing all records subsequently.

The following table is an overview of the most common transformations in EtlKit:

Non-blocking	Partially blocking	Blocking
RowTransformation	LookupTransformation	BlockTransformation
RowBatchTransformation
Aggregation	CrossJoin	Sort
MergeJoin
Multicast
RowDuplication
RowMultiplication
RowFiltration
ExpressionRowFiltration
JsonTransformation
ScriptedRowTransformation
AIBatchTransformation

Designed for big data

EtlKit was designed for performance and is able to deal with big amounts of data. All destinations do support Bulk or Batch operations. By default, every component comes with an input and/or output buffer. You can design your data flow that only batches or your data is stored in memory, which are kept in different buffers for every component to increase throughput. All operations can be execute asynchrounously, so that your processing will run only within separate threads.

Data transformations

Data transformations take input data from source, perform an external operation (via DB, API, or queue) and return produced result into a destination. Namely data transformations are:

Transformation	Input	Processing	Output
SqlQueryTransformation	Parameters to a liquid-based SQL query template	Execute SQL query	SQL query results (0..N for each input row)
SqlCommmandTransformation	Parameters to a liquid-based SQL query template	Execute SQL non-query statement	Input object (or this can be customised with Transform delegate) — 1 for each input row
KafkaTransformation	Parameters to liquid-based message and key templates	Produce messages to Kafka topic (optional key)	Input object or null
RabbitMqTransformation	Parameters to a liquid-based string message template	Publish messages to RabbitMQ channel	Input object or null
JsonTransformation	Json object	Execute JSON path transformation for each field	Output object where each field is the result of Json path evaluation
RestTransformation	Parameters to a liquid-based request URL and body templates	Execute HTTP request	ExpandoObject with response code, raw body and Json parsed body in fields
ScriptRowTransformation	Object as Globals to C# script templates for each result field	Execute C# code	Result object with one field per script
AIBatchTransformation	Parameters to a liquid-based AI prompt template	Execute request to configured OpenAI HTTP endpoint	ExpandoObject with response code, raw body and Json parsed body in fields

Control Flow - overview

Control Flow Tasks gives you control over your database: They allow you to create or delete databases, tables, procedures, schemas or other objects in your database. With these tasks you also can truncate your tables, count rows or execute any sql you like. Anything you can script in sql can be done here - but with only one line of easy-to-read C# code. This improves the readability of your code a lot, and gives you more time to focus on your business logic.

Code tells - here is some example code, without writing the whole "boilerplate" code by ADO.NET.

var conn = new SqlConnectionManager("Server=10.37.128.2;Database=EtlKit_ControlFlow;Uid=etlkit;Pwd=etlkitpassword;");
//Execute some Sql
SqlTask.ExecuteNonQuery(conn, "Do some sql",$@"EXEC myProc");
//Count rows
int count = RowCountTask.Count(conn, "demo.table1").Value;
//Create a table (works on all databases)
CreateTableTask.Create(conn, "Table1", new List<TableColumn>() {
    new TableColumn(name:"key",dataType:"INT",allowNulls:false,isPrimaryKey:true, isIdentity:true),
    new TableColumn(name:"value", dataType:"NVARCHAR(100)",allowNulls:true)
});

Logging

EtlKit uses Microsoft.Extensions.Logging which is a de-facto standard for modern .NET. EtlKit also has its own logging-to-database capabilities, implemented in the EtlKit.Logging.Database package, which has its own configuration and uses NLog internally at the time being.

Contribution

Clone the repository:

git clone https://github.com/etlkit/etlkit.git

Then, open the downloaded solution file EtlKit.sln with Visual Studio 2022 or JetBrains Rider. Now you can build the solution, and use it as a reference in other projects.

Running tests

See TEST_SETUP.md for instructions.

Going further

EtlKit is open source. Feel free to make changes or to fix bugs. Every participation in this open source project is appreciated.

To dig deeper into it, have a look at the test projects. There is a test for (almost) everything that you can do with EtlKit.