3 ms·

I still recommend spending the time to (re-)learn bottom-up parsing though, even if you never use it again. Specifically, LR and then GLR (no need to bother wit

I still recommend spending the time to (re-)learn bottom-up parsing though, even if you never use it again. Specifically, LR and then GLR (no need to bother with LALR/SLR/all those variants). Maybe even recursive ascent too, or recursive ascent-descent if you find those interesting. It's painful and confusing and takes a long time to pick them up, but it's quite gratifying once you do learn it. The algorithms are super clever and cool, and help you gain a lot of appreciation and intuition for what is/isn't possible to do (whether efficiently or inefficiently). Using recursive descent for everything is like using Python (or Go or pick your favorite easy language to learn) for everything. Yeah you can totally do it, and a heck of a lot of people do precisely because of how nice and convenient it is, but it's good to have a sense what else is out there and how you can think about a problem differently IMO, even if you rarely use other languages in practice.

recursivedoubts 5y ago
i agree, but for an undergraduate understanding of parsing and grammars, recursive descent is perfect i think the more complex parsing algorithms were good for producing papers, which explains their success in academia but relative lack of success in industry
samus 5y ago
Recursive parsing is a completely valid concept, but it's unnatural at first to hand-write parsers with it. I have tried it - even wrote a prototype DSL in Haskell (the CPS monad is very useful for that), but what I end up doing is using the DSL to write down the LR state machine, bottom up. This is eerily familiar to embedded and GUI programming where internal state matters and the program reacts to input. It might even be a better way to incorporate ad-hoc code that has side effects. However, most people think about programming languages top-down and use grammars to describe them.

recursivedoubts 5y ago

i agree, but for an undergraduate understanding of parsing and grammars, recursive descent is perfect i think the more complex parsing algorithms were good for producing papers, which explains their success in academia but relative lack of success in industry

samus 5y ago

Recursive parsing is a completely valid concept, but it's unnatural at first to hand-write parsers with it. I have tried it - even wrote a prototype DSL in Haskell (the CPS monad is very useful for that), but what I end up doing is using the DSL to write down the LR state machine, bottom up. This is eerily familiar to embedded and GUI programming where internal state matters and the program reacts to input. It might even be a better way to incorporate ad-hoc code that has side effects. However, most people think about programming languages top-down and use grammars to describe them.