3 ms·
I think the reason this happens is that parser generators are probably never (ignoring AGI etc) going to produce code as fast and debugger-friendly as a human b
by scythe 8y ago
I think the reason this happens is that parser generators are probably never (ignoring AGI etc) going to produce code as fast and debugger-friendly as a human being. A really good parser generator might be a good way to "mock up" the design of a language, but for anything that gets used in production, the benefits of hand-rolling your parser will probably always outweigh the costs by a large margin: it's crucial code that affects thousands of projects, so it really should be perfect.
- cwzwarich 8y agoI don't think the performance part of your statement is true. There exist techniques for generating extremely efficient LR automata (e.g. https://link.springer.com/article/10.1007/BF00267043 https://link.springer.com/article/10.1007/BF00267043) where the equivalent handwritten code would be a goto-ridden mess.