Comment by d0m
10 years ago
I agree. I remember my compiler class where we went through the whole Dragon book and we had a lot of theoretical knowledge, but we were disappointed because we practically couldn't build a compiler. We went in too much detail too fast without having a good understanding and practical examples with compilers.
> the whole Dragon book
I learned compilers from Al Aho, and believe me, the class was no better in person. Mostly in-person waxing poetic about his time at Bell Labs...
Dinner with Al Aho was one of the highlights of when I visited Columbia a while ago. He's simply an awesome person. And he's done some fantastic work - Aho-Corasick is a great algorithm. Alas, the Dragon Book was one of the low points of my undergraduate education, though I had a great teacher who pulled out a terrifically fun class despite the book. (Thanks, +Wilson Hsieh.)
In fairness, though -- or more an admission of how wrong we can be about things like this -- I subscribed to the "parsing is boring and solved" belief until Bryan Ford's Packrat Parsing work made me realize that it had just gotten stuck at a traffic light for a few decades.
Veering a bit offtopic, but does anyone have any pointers to important recent work on parsing? There are alot of papers out there and I guess I don't know how to sift through them. I've heard the "parsing is solved" line before, but so much of my time is spent doing some type of parsing that even incremental improvements are extremely interesting.
12 replies →
I went through a undergrad compiler class with the Dragon book, not the whole thing though, just enough chapters to build a compiler. We did build a full fledged compiler for an OO language, with all the parts of a compiler all the way to code generating assembly. It's kind of fast pace with too many stuff crammed into a class. Some theoretical topics were kind of fuzzy at the time. Still it was one of the best classes I have ever attended. Later on when I re-read the book on my own pace, I thought the book was the best thing since slice bread. Got a much better understanding of the theories behind the compiler topics. A lot of the things I learned I can still apply to this day.
Interesting, I took compilers where the Dragon book was the standard text book, we implemented a PASCAL like language (which compiled for the DEC10 as a target) and it was one of the really transformative classes I had taken. It brought together algorithms and data structures into a curriculum where they joined up to convert text into executable code. Really one of those moments of sudden clarity for me.
In my compiler class we also used the Dragon book and we did write a compiler for a C like language... I don't think we covered the book in its entirety though but sounds like that was a better mix. I felt quite capable of writing simple compilers after I finished the course...
I loved the Dragon book, but I agree not a practical book. I probably loved it because at the time I'd already been experimenting with compiler writing for 5-6 years, and it gave me a theoretical grounding for a lot of stuff I'd been doing or doing subtly "wrong".
I think it - and similar books - should be banished from first compiler classes, and instead be used later. Much more practical books like Wirth's that actually walk you through writing a compiler would be much better for most beginners.