Pytra

is Pytra's input language — Pytra transpiles that code into multiple target languages.

· Supported Backends ·

Features

🐍 Python → native code in each target language

• 🌐 Transpiles to C++ / Rust / Go / Java / TS and many more
• 🧩 Preserves the original code structure almost entirely
• ⚡ Write in Python, generate high-performance code
• ✨ Simple Python subset input
• 🛠 Works with existing tools like VS Code out of the box
• 🔧 The transpiler itself is written in Python — easy to extend
• 🔁 Self-hosting capable — can transpile itself

Performance Comparison

Execution time of sample code written in Python versus execution time of the same code after transpilation. (Unit: seconds.) The Python column is the original code; PyPy is included for reference.

No.	Description
06	Julia set parameter sweep (GIF)	9.627	0.507	0.546	0.407	0.329	0.626
16	Glass sculpture chaos rotation (GIF)	6.847	0.606	0.277	0.246	1.220	0.650

Full data for all languages and all samples → Sample page

Sample code : 06_julia_parameter_sweep.py Full source: sample/py/06_julia_parameter_sweep.py Transpiled code (per language) C++ | Rust | C# | JS | TS | Dart | Go | Java | Swift | Kotlin | Ruby | Lua | Scala3 | PHP | Julia

Sample code : 16_glass_sculpture_chaos.py Full source: sample/py/16_glass_sculpture_chaos.py Transpiled code (per language) C++ | Rust | C# | JS | TS | Dart | Go | Java | Swift | Kotlin | Ruby | Lua | Scala3 | PHP | Julia

Python vs C++ vs Rust vs Pytra

Legend: ✅ = Good / 🔶 = Partial / limited / ❌ = Not supported / difficult

Aspect
Syntax	✅ Simple	❌ Complex	🔶 Ownership/ lifetimes	✅ Same as Python
Type safety	❌ Dynamic	✅ Static	✅ Static	✅ Static (Python-style annotations)
Execution speed	❌ Slow	✅ Fast	✅ Fast	✅ Fast (depends on target)
Memory management	✅ GC (easy but heavy)	❌ Manual/ shared_ptr	🔶 Ownership (safe but hard)	✅ RC-based automatic
Integer types	🔶 Arbitrary precision only	✅ int8–64	✅ i8–i64	✅ int8–64
float	🔶 64-bit only	✅ 32/64-bit	✅ f32/f64	✅ 32/64-bit
Build	✅ Not needed	❌ CMake etc.	🔶 cargo	✅ ./pytra --build --run
Multi-language output	❌	❌	❌	✅ 18 languages
Optimization	❌ Limited	✅ Rich	✅ Rich	✅ Leverages target
Distribution	🔶 Requires runtime	✅ Binary	✅ Binary	✅ Language-native
Single inheritance	✅	✅	❌ traits only	✅
Multiple inheritance	✅	🔶 Complex	❌	❌
Mix-in	✅	🔶 CRTP etc.	❌	✅
Trait/ Interface	🔶 Protocol	🔶 virtual base	✅ Native	✅ @trait
Exception handling	✅	✅	❌ Result/panic	✅ All languages
Templates/ Generics	❌	🔶 Cryptic errors	✅	✅ @template
Selfhost	❌	❌	❌	✅

Read the Docs

Getting started	Tutorial	チュートリアル
Guide	Guides	ガイド
Specification	Spec index	仕様書
Progress	Project Progress	プロジェクト進捗

Changelog

2026-04-10 — P0-ZIG-CREXC-S4 complete. Zig / Rust exception / try / with handling fully shared via CommonRenderer hooks. Zig toolchain_ dependency eliminated (all languages now at 0). Nim / Go / Lua new fixture parity complete. Lua copy elision done.

2026-04-09 — Zig / Rust handler binding / panic / block expression helpers consolidated in CommonRenderer.

2026-04-08 — All languages lint clear (697→0). 18 languages at 10/10 PASS. C# / Go / Nim parity restored. Nim emitter string-split workaround eliminated. P0-ZIG-CREXC S1-S3 started.

2026-04-07 — Lint down to 149 / 14 languages at 10/10 PASS (697→149). PyFile abolished, IOBase hierarchy in built_in/io.py. Emitter guide §12.7. Cross-language PyFile coupling removal.

2026-04-06 — With statement via __enter__/__exit__ protocol (CommonRenderer try/finally + hoist). 2 with fixtures added. TS/JS shim cleanup complete. Dart emitter guide compliance. JVM major progress. .east* removed from git.

2026-04-05 — containers.py mut[T] annotations for meta.mutates_receiver. C++ method name hardcode removed. mapping.json FQCN key unification. Toolchain rename complete.

Full changelog

License

Apache License 2.0