programming language design

Bilingual

Design and Research
3 months ago
- The author suggests answering 'What do you do?' with 'I'm designing a new dialect of Lisp' to avoid the question.
- Design focuses on what is good for the user, while research focuses on what is new, though they converge at the top.
- Good design requires understanding user needs, not just wants, similar to a doctor treating underlying causes.
- Designers should include themselves in the intended user group to avoid condescension and produce better results.
- Programming languages are for humans, and like arts, they must account for human cognitive limitations.
- Interactive development (e.g., a REPL) influences language features like mandatory type declarations.
- Prototyping and iterative refinement (Worse is Better) are crucial in both software and arts for effective design.
- Morale is key; designers should remain engaged and enjoy what they create, avoiding boredom.
- Good design typically requires a single decision-maker (dictator) rather than collaboration, as it must be coherent.
Being Popular
3 months ago
- Programming language popularity is driven by expert hackers, not by merits alone.
- Brevity is crucial for attracting hackers; verbose syntax hinders adoption.
- Hackability (allowing programmers to do what they want) is more important than brevity.
- Throwaway programs often evolve into major systems, so languages must be good for quick coding.
- Powerful, well-designed libraries are more valuable than core language features.
- Common Lisp's unpopularity stems from poor libraries and lack of a system to hack, not its syntax.
- Languages need time, a free implementation, a good book, and a target system to gain traction.
Two Case Studies of NaN
3 months ago
- Python's list equality optimization compares elements by identity first, which assumes an object always compares equal to itself, an assumption violated by NaN.
- In Lua's numerical for-loops, NaN leads to unintuitive behavior because the implementation's comparison logic breaks when checking limits and steps with NaN, which is likely an oversight.
- NaN's unique properties often cause programming language design choices or optimizations to fail, as seen in Python, Lua, and previously documented cases like Go map keys.

programming language design

Design and Research

Being Popular

Two Case Studies of NaN