vertti / daffy

Daffy — Validate pandas & Polars DataFrames with Python Decorators **Validate your pandas and Polars DataFrames at runtime with simple Python decorators.** Daffy catches missing columns, wrong data types, and invalid values before they cause downstream errors in your data pipeline. Also supports Modin and PyArrow DataFrames. • ✅ **Column & dtype validation** — lightweight, minimal overhead • ✅ **Value constraints** — nullability, uniqueness, range checks • ✅ **Row validation with Pydantic** — when you need deeper checks • ✅ **Works with pandas, Polars, Modin, PyArrow** — no lock-in --- Installation or with conda: Works with whatever DataFrame library you already have installed. Python 3.10–3.14. --- Quickstart If a column is missing, has wrong dtype, or violates a constraint — **Daffy fails fast** with a clear error message at the function boundary. --- Why Daffy? Most DataFrame validation tools are schema-first (define schemas separately) or pipeline-wide (run suites over datasets). **Daffy is decorator-first:** validate inputs and outputs where transformations happen. | | | | ------------------------ | -------------------------------------------------------------------------------- | | **Non-intrusive** | Just add decorators — no refactoring, no custom DataFrame types, no schema files | | **Easy to adopt** | Add in 30 seconds, remove just as fast if needed | | **In-process** | No external stores, orchestrators, or infrastructure | | **Pay for what you use** | Column validation is essentially free; opt into row validation when needed | --- Examples Column validation Regex column matching Match dynamic column names with regex patterns: Value constraints Vectorized checks with zero row iteration overhead: Available checks: , , , , , , , , , Also supported: , , , , Nullability and uniqueness Row validation with Pydantic For complex, cross-field validation: --- Daffy vs Alternatives | Use Case | Daffy | Pandera | Great Expectations | | ---------------------------- | :-----------------: | :----------------: | :-----------------: | | Function boundary guardrails | ✅ Primary focus | ⚠️ Possible | ❌ Not designed for | | Quick column/type checks | ✅ Lightweight | ⚠️ Requires schemas | ⚠️ Requires setup | | Complex statistical checks | ⚠️ Limited | ✅ Extensive | ✅ Extensive | | Pipeline/warehouse QA | ❌ Not designed for | ⚠️ Some support | ✅ Primary focus | | Multi-backend support | ✅ | ⚠️ Varies | ✅ | --- Configuration Configure Daffy project-wide via : --- Documentation Full documentation available at **daffy.readthedocs.io** • Getting Started — quick introduction • Usage Guide — comprehensive reference • API Reference — decorator signatures • Changelog — version history --- Contributing Issues and pull requests welcome on GitHub. License MIT