Google Summer of Code 2025 Final Report

Student: Swaminath Shiju
Project: Expanding OOP capabilities of LPython for porting common CPython modules
Organisation: Python Software Foundation
Mentor: Ondřej Čertík
Weekly Reports: Link to weekly reports

Abstract

The original goal was to enhance LPython’s object-oriented programming (OOP) capabilities by implementing dunder methods and supporting non-dataclass imports. However, due to significant updates in LPython and its shared backend with LFortran, the project scope was revised during the bonding period in consultation with my mentor.

The revised objective focused on implementing Python-style container types (lists, sets, dictionaries, and tuples) in LFortran’s frontend and runtime. This effort improved test coverage for Python-specific backend components, ensured smoother synchronization with LPython’s libasr backend, and supported a major synchronization cycle.

The project spanned frontend syntax, backend code generation, and data structure intrinsics.

Final Deliverables
Timeline Summary
List of Contributions (PRs)
Future Work
Acknowledgements

Final Deliverables

Implemented and stabilized Python-like container types (list, set, dict, tuple) in LFortran’s frontend and runtime.
Refactored backend to reduce reliance on deprecated mechanisms and prepared for string/struct changes.
Introduced and iterated on Union types: syntax, access, and ASR refactor.
Improved C backend behavior (string arguments, dict copy) and enabled robust unsigned integer support.
Unblocked and synchronized LPython with latest libasr changes; restored integration tests.

Timeline Summary

Week	Key Achievements
Week 1 [25/5 - 1/6]	Initial list/set syntax implementation First tests in frontend & runtime
Week 2 [1/6 - 8/6]	Improved syntax for lists and sets Added for tests to update backend
Week 3 [8/6 - 15/6]	Fully ported several LPython List tests Proper syntax for dicts and tuples implemented
Week 4 [15/6 - 22/6]	Completely ported all tuple tests & several list tests Simplified syntax for LP types Backend refactoring
Week 5 [22/6 - 29/6]	Fixed string-type keys in dict type Exposed Union type in LF frontend Optimized list.pop Refactored backed for less reliance on create_gep, typecode Successfully merged a month long sync PR for LPython
Week 6 [29/6 - 6/7]	Implemented Union type access Union ASR type refactor
Week 7 [6/7 - 13/7]	Fixed Unsigned integer properly in libasr Made progress towards having C-introp working in LP
Week 8 [13/7 - 20/7]	Began work on fixing more complicated bugs in post sync LP
Week 9 [20/7 - 27/7]	Uint in LF frontend Test coverage for Uint
Week 10 [27/7 - 3/8]	Uncommented several Integration tests in LP Union refactor Fixed list regression after string refactor
Week 11 [3/8 - 10/8]	Began work on large scale LP sync after struct and string refactors in LF backend
Week 12 [10/8 - 17/8]	Finishied sync to a large extent Fixed Dict regression after string refactor

List of Contributions (PRs)

Current State

LPython’s continuous integration (CI) supports approximately 200 tests with the latest libasr backend.
Ported all tuple tests and nearly all list, set, and dictionary tests to LFortran. Most tests affected by the backend string refactoring broke. I fixed list tests and all but one dictionary test fixed.
Adapted the LPython frontend to align with recent backend refactoring efforts.

Future Work

Complete LPython synchronization to achieve approximately 300 passing tests, integrating LFortran-specific features as tests to prevent backend changes from breaking LPython.
Implement dunder methods and non-dataclass imports in LPython, aligning with the original project goals.
Begin porting additional CPython standard library modules to LPython.
Benchmark performance using Advent of Code 2025 challenges.

Learnings

Before I started committing to the project I had little to no experience with the LLVM toolchain - but along the course of the project I have picked up a good amount of LLVM fundamentals to begin exploring on my own.
I've learnt how to use tools like lldb to help debug large complex projects
I've also learnt how to effectively communicate with other open-source contributors in the community.
With my college semester starting in the final 3–4 weeks of the program, I developed effective time management skills to balance academic responsibilities and project commitments.
Git version control, bison, flex, conda, mambda etc

Acknowledgements

Thanks to my mentor Ondřej Čertík, the LFortran/LPython community, and PSF GSoC coordinators for accepting me into this program and providing guidance and reviews throughout the project.

GSoC 2025 – Final Work Report