https://github.com/jangorecki/rollbench

rolling statistics benchmark
https://github.com/jangorecki/rollbench

benchmark data-table pandas polars rolling-windows

Last synced: 2 months ago
JSON representation

rolling statistics benchmark

• Host: GitHub
• URL: https://github.com/jangorecki/rollbench
• Owner: jangorecki
• Created: 2023-10-04T13:56:05.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2025-10-04T08:48:18.000Z (9 months ago)
• Last Synced: 2025-10-25T11:51:12.709Z (8 months ago)
• Topics: benchmark, data-table, pandas, polars, rolling-windows
• Language: R
• Homepage:
• Size: 20.5 KB
• Stars: 0
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          

# mini rolling statistics benchmark

Compares python pandas, R data.table and R polars by:

- input size: `1e6, 1e7, 1e8`

- rolling window size: `1e2, 1e3, 1e4`

- rolling functions: `mean` and `median`

- batching: `single` computation and `quadruple` (2 columns x 2 windows) computation

For comprehensive benchmark of rolling functions see [db-benchmark](https://github.com/duckdblabs/db-benchmark/pull/9) project.

## running

```sh

python pandas2.py

Rscript data.table.R

Rscript polars.R

```

## results

As of 2025-09-16 using

```

pandas 2.2.3

data.table@1.17.99

polars 1.3.1

```

![mean single](https://github.com/user-attachments/assets/666aea13-0947-4753-a1ab-214b8ec20f6e)

![mean quadruple](https://github.com/user-attachments/assets/f5f29a64-a58a-413d-bb12-1a32269b1ea7)

![median single](https://github.com/user-attachments/assets/9d94f0b7-b4c0-44da-9c34-54bc2a8fb784)

![median quadruple](https://github.com/user-attachments/assets/e328703b-fca9-45a8-b885-a8efb4bcd505)

On Fedora 42 and AMD Ryzen 7 7735HS CPU.

## license

MIT