- Installation
- Guides
- Data Import & Export
- CSV Import
- CSV Export
- Parquet Import
- Parquet Export
- Query Parquet
- HTTP Parquet Import
- S3 Parquet Import
- Meta Queries
- Python
- Install
- Execute SQL
- Jupyter Notebooks
- SQL on Pandas
- Import From Pandas
- Export To Pandas
- SQL on Arrow
- Import From Arrow
- Export To Arrow
- DuckDB with Ibis
- SQL Editors
- Documentation
- Connect
- Data Import
- Client APIs
- Overview
- Python
- R
- Java
- C
- Overview
- Startup
- Configure
- Query
- Data Chunks
- Values
- Types
- Prepared Statements
- Appender
- Table Functions
- Replacement Scans
- API Reference
- C++
- Node.js
- WASM
- ODBC
- CLI
- SQL
- Introduction
- Statements
- Overview
- Select
- Insert
- Delete
- Update
- Create Schema
- Create Table
- Create View
- Create Sequence
- Create Macro
- Drop
- Alter Table
- Copy
- Export
- Query Syntax
- Data Types
- Expressions
- Functions
- Overview
- Numeric Functions
- Text Functions
- Pattern Matching
- Date Functions
- Timestamp Functions
- Time Functions
- Interval Functions
- Date Formats
- Date Parts
- Blob Functions
- Nested Functions
- Utility Functions
- Indexes
- Aggregates
- Window Functions
- Samples
- Information Schema
- Configuration
- Pragmas
- Extensions
- Development
- Why DuckDB
- FAQ
- Code of Conduct
- Live Demo
- Benchmarking
Continuous Benchmarking
Continuous benchmarking is run on every commit to detect performance regressions in DuckDB. All queries run with a standard timeout of 30 seconds to avoid the benchmark suite from taking too much time to run. Benchmarks are ran using the benchmark runner, and are defined in the benchmark subdirectory of the DuckDB source repository. The following sets of benchmarks are run: