Nano-parsing: A Data-parallel Architecture for Perverse Parsing Environments
summary
Proposes a data‑parallel parsing architecture that runs many tiny nano passes over the whole input, using APL-style tree wrangling and parent vector / depth vectors to incrementally refine a global parse tree. Demonstrations include a working APL parser (handles dialects and legacy codebases with millions of lines), a YAML parser, and a Metamath parser. Produces precise inline error underlines, supports stepwise debugging, and is amenable to GPU backends; runtime cost is multiple cheap passes rather than deep recursion.
AI-written summary · about
chapters (11)
- 0:00Nano-parsing architecture
- 1:05Traditional parsing challenges
- 3:35Nano-parsing approach
- 6:50Incremental parse tree refinement
- 9:23Nano-parsing in practice
- 13:36Nano-parsing benefits
- 16:25Developer advantages
- 18:34Scalability and platform independence
- 19:55Future research
- 22:04Q&A
- 24:11When to use Nano-parsing
related talks
40mType-safe Nanopasses: How to write a safe and modern compiler front-end
34mEfficient Nanopass Compilers using Cats and Matryoshka
41mWriting a Nanopass Compiler
58mParsing in C# from first principles
14mFast Incremental PEG Parsing
29mMy Little Parser: Tooling is Magic
29mNanopass Attribute Grammars
42mFrom Linear to Incremental