Functional Programming with Serialized Data

on YouTube ↗32m

" summary

Gibbon compiles a subset of Haskell/ML into code that operates directly on serialized bytes, claiming safe, pointer-free traversals by enforcing symbolic locations and an end witness relation so every written region proves out its serialized shape. The compiler runs traversal effect inference, inserts offsets or indirections when needed, and emits cursor-passing low-level code. Benchmarks show Gibbon beating alternatives on tree walks; extensions add parallelism, GC, layout optimization and a Haskell library.

AI-written summary · about

" chapters (8)

1. 0:00Functional programming intro
2. 3:28Motivation for serialized data
3. 12:01Gibbon compiler goal
4. 16:15Symbolic locations enforce serialization
5. 22:14Safe add-one in middle language
6. 25:28Handling incomplete traversals
7. 28:42High-level to cursor-passing code
8. 30:28Conclusion and future

" related talks

• 46mFunctional Programming with GoDylan Meeus · GopherCon 2020
• 36mConcrete Functional ProgrammingKamila Szewczyk · Lambda World 2024
• 01h 44mFunctional Programming: Technical Reasons to AdaptVenkat Subramaniam · JavaDay Kyiv 2015
• 55mModern Software Development with HaskellRúnar Bjarnason · Devoxx US 2017
• 41mEvolving Functional Programming LanguagesErik Osheim · Lambda World 2016
• 44mRethinking compilers with live codingTomas Petricek · Lambda World Seattle 2018
• 33mAnalyzing Functional ProgramsDavid Cleaver · Scale By the Bay 2017
• 42mFunctional Programming in 40 MinutesRuss Olsen · GOTO Berlin 2018