Simplicity and conciseness of R,
with the blazing speed of Rust.
Chainable • Composable • Declarative • Semantic • Agent-Ready
pydplyr is a modern, expressive, and chainable data manipulation library built for humans and machines. Think:
- 🧠 Like dplyr in R
- ⚡️ With Python's ecosystem
- 🚀 On a path to a Rust-powered backend
- 🤖 Agent-ready for autonomous systems
Whether you're a data scientist, a developer building intelligent systems, or a machine 🤖 looking for clarity—this library speaks your language.
- ✅ Composable Verbs – clear, expressive syntax for fast prototyping & serious work
- 🧬 Chainable API – minimal boilerplate, max readability
- 🎨 Grammar of Graphics – familiar design for layered visualizations (Coming Soon!)
- 🔍 Simplified Regex – harness RegEx without the brain melt
- 🔗 Agent-First Thinking – semantically rich, logic-oriented operations
- 🦀 Rust Ambition – future versions will compile to Rust for blazing performance
Each verb is purpose-built and plays well with others. Start small, scale infinitely.
| Verb | Description |
|---|---|
arrange() |
Sort your data |
select() |
Pick columns |
filter() |
Subset rows |
mutate() |
Create or modify columns |
summarize() |
Aggregate and reduce |
group_by() |
Enable grouped operations |
distinct() |
Drop duplicate rows |
from pydplyr import *
result = (
Panel(df)
.arrange(desc("score"))
.filter("score > 80")
.mutate(score_plus_age="score + age")
.select("name", "score_plus_age")
.collect()
)Intuitive, readable, chainable. One thought per line.
Just like in ggplot2, our graphics philosophy follows this layered system:
- Data – the DataFrame
- Aesthetics – x/y mappings, color, shape
- Geoms – bar, point, line, etc.
- Stats – transformations like count or smooth
- Facets – split plots by category
- Coords – coordinate systems (polar, cartesian, etc.)
- Themes – polish for publication or dashboard
📊 Visuals should tell, not yell.
We believe RegEx shouldn't be a dark art.
Panel(df).filter_col("email", pattern=".*@example.com")You get the full power of re, simplified into expressive helpers for real-world usage.
The future of data is semantic, composable, and intelligent.
pydplyr is being designed with agentic AI frameworks in mind — where the code can be read and written by both humans and agents.
Whether it’s embedded in LLM-based agents or running as the logic core of autonomous data pipelines, pydplyr is made to be interpretable, traceable, and chainable.
- Core verbs (arrange, select, filter, mutate, summarize, distinct)
- Grouped operations
- Grammar of Graphics module
- Rust backend (via
pyo3orpolars) - Natural-language Regex builder
- LLM prompt-to-code interface
- Optional
asyncAPI for distributed computing - Plugin system for custom verbs and visual geoms
pip install pydplyrWe welcome contributors who care about:
- elegant APIs 🧼
- expressive code 💬
- performance 🔥
- semantic richness 🌐
- and dreaming big 💡
To get started, clone the repo and check the CONTRIBUTING.md guidelines.
Got ideas? Found bugs? Want to build the future of data science?
- Open an issue
- Start a discussion
- Or just drop by with a star ⭐️
"The art of data science is not in the numbers — it’s in the story they tell, and the tools that let them speak."
—