Database Systems

_Recursive CTEs and DuckDB in Education_ (Torsten on the _Disseminate_ Podcast)

Thu, 16 Oct 2025 00:00:00 UTC

Jack Waudby invited Torsten to ramble about Recursive CTEs and DuckDB in Education in the DuckDB in Research series of the Disseminate Podcast. We cover USING KEY and also touch on WITH TRAMPOLINE. If you’ve got a spare 60 minutes during your commute, you are invited to tune in.

The _DuckDB Explain Visualizer_ is now available!

Wed, 26 Feb 2025 00:00:00 UTC

We are happy to announce the release of the DuckDB Explain Visualizer! It is a web-based tool that allows you to visualize the query plans of DuckDB. The DuckDB Explain Visualizer is available here: https://db.cs.uni-tuebingen.de/explain.

What is the DuckDB Explain Visualizer?

If you spend any significant amount of time looking at DuckDB query plans, sooner or later you will realize that the textual representation of the query plan has its limitations. It can be hard to read and understand, especially for complex plans. And worse, if the plan is too large, it can be cut off, making it impossible to understand the full plan. Also, there are no visual cues to help you understand the plan, find bottlenecks, or identify optimization opportunities.

Luckily, the DuckDB team has added an alternative way to represent query plans: As a JSON object, which can be visualized in a more user-friendly way. This is similar to the EXPLAIN (FORMAT JSON) feature in PostgreSQL. We leverage the JSON representation to render the query plan in a much more visually appealing fashion, interactive, and with additional information (e.g. slowest operator or the number of rows processed by each operator).

One of the main features of the DuckDB Explain Visualizer is the ability to share query plans with others. Once you have uploaded a query plan, you can share the link, and others can view the query plan without having to upload it again. We believe this feature will be especially useful for teaching, debugging, and collaboration in general.

DuckDB Explain Visualizer

How to use the DuckDB Explain Visualizer?

To use the DuckDB Explain Visualizer, you need to have a DuckDB query plan in JSON format. You can obtain this by running a query prefixed with EXPLAIN (FORMAT JSON). Then, you can paste the JSON output into the DuckDB Explain Visualizer to visualize the query plan. You can also use the ANALYZE option to include runtime statistics in the query plan.

1EXPLAIN (ANALYZE, FORMAT JSON)
2WITH RECURSIVE t(x) AS (
3 SELECT 1
4 UNION ALL
5 SELECT x + 1 FROM t WHERE x < 100
6)
7SELECT * FROM t;

Alternatively, you can use PRAGMA commands to enable profiling and save the output to a file. This is useful for long-running queries or queries that you want to analyze later. The output file can then be uploaded to the DuckDB Explain Visualizer. The following commands enable profiling and save the output to a file:

1PRAGMA enable_profiling = 'json';
2PRAGMA profiling_output = '/path/to/file.json'; -- (optional)
3[your-query-here]

How did we build the DuckDB Explain Visualizer?

The basis for the DuckDB Explain Visualizer is the excellent PostgreSQL Explain Visualizer (PEV) by Alex Tatiyants. This tool has been re-implemented a few years ago as PostgreSQL Explain Visualizer 2 (PEV2). Both PEV and PEV2 are web-based tools that allows users to visualize query execution plans of PostgreSQL. Exactly what we want—and need—for DuckDB! The source code of the DuckDB Explain Visualizer is available on GitHub.

Credits

During Matthis Noël’s bachelor thesis, he adapted PEV2 to work with DuckDB’s execution plans. This involved understanding the structure of DuckDB’s execution plans and adapting the visualization tool to work with them.

_Trampoline-Style Queries for SQL_ will be at CIDR 2025

Wed, 04 Dec 2024 00:00:00 UTC

We love the single-track system-focused CIDR series of conferences and we are happy to be able join the community again for CIDR 2025, once more located in Amsterdam (NL), January 19-22, 2025. Together with our colleagues Thomas Neumann and Altan Birler of TU Munich, we discuss a new foundation for iterative queries: we step away from fixpoints and instead bank on Trampoline-Style queries for SQL, a concept inspired by the programming languages and compilers domain. WITH TRAMPOLINE has already been prototypically implemented in Thomas’ Umbra RDBMS. Trampolines are versatile—they also serve as the backbone for our work in the compilation of imperative-style programs to SQL.

_A Fix for the Fixation on Fixpoints_ accepted at CIDR 2023

Mon, 28 Nov 2022 00:00:00 UTC

Tübingen returns to CIDR, this time for the 2023 edition to be held in Amsterdam (NL), January 8–11, 2023. We will report on our recent explorations into alternative forms of iteration in SQL—generalizing recursive common table expressions—that are easier to write and read as well as more efficient to evaluate than WITH RECURSIVE (which has been established more than 20 years ago but since then has been largely left untouched).