Sorry for double post... the web interface has lag. I thought my last one didn't get through, so I rewrote the message.

"John Reimer" <John_member@pathlink.com> wrote in message news:bs4fa9$lau$1@digitaldaemon.com...
> I assume you would still recommend this book, right?

I have read the earlier two editions of it, and they are excellent introductions to compiler writing. I've also taken a compiler course instructed by Profs. Ullman & Hennessey, and I learned a great deal from that course that went far beyond their book. I hope the new edition will have that extra info in it!

"John Reimer" <John_member@pathlink.com> wrote in message news:bs4o1f$12n8$1@digitaldaemon.com...
> In article <bs4kuc$u1f$1@digitaldaemon.com>, Charles says...
> >I actually heard the dragon book was abit outdated and disorganized.  I
more
> >of a hands-on person then theory, hopefully the new book will contain
alot
> >of examples.  Next year is along time away though :/
> I heard the same about the original and am hoping the new book will fix
these
> issues. I also prefer implementation examples to back up theory, so it
would
> beneficial if the new book provided a small language implementation.  I
know
> other books that do. At the same time I understand the need for a good theoretical base to comprehend complicated compiler design. The book I
have is
> the old dragon book (1979 -- believe it or not; hey, it was cheap!).  So
I'm
> kind of interested to see what the new book has to offer.  Lots of
examples in
> it would indeed be welcome!

Writing a good compiler means sweating the details and getting the basics right. Those basics are the same now as they have been for 40 years. DMC++ is the fastest C++ compiler out there, and its based on 1982 technology.

Object-orienting the guts of the compiler has made it better (see the DMD front end source for example), but that isn't really a new *compiler* technology.

>I have read the earlier two editions of it, and they are excellent introductions to compiler writing. I've also taken a compiler course instructed by Profs. Ullman & Hennessey, and I learned a great deal from that course that went far beyond their book. I hope the new edition will have that extra info in it!
>

That's encouraging because the basics are all I need at this point, so the old book will be a great starting point.  I will likely grab the new book at some point because intro describes "new" or "more indepth" chapters on: "syntax-directed translation, data-flow analysis, code optimization, just-in-time compiling, and garbage collection" among other things.  I didn't think those topics were too new though...but perhaps "new" to the book.  JIT and garbage collection were around pre-Smalltalk days from what I recall.  Perhaps the techniques taught are modernized.

It's quite an opportunity you had learning under those professors.  I imagine information covered in those classes might not readily make it into any textbook.  I'd like to be pleasantly surprised to the contrary, though :-).

Later,

John

>Writing a good compiler means sweating the details and getting the basics right.

You're telling me!  It's no easy topic from the stuff I've been reading so far! But I guess a hard study effort will pay off eventually.

>Those basics are the same now as they have been for 40 years. DMC++
>is the fastest C++ compiler out there, and its based on 1982 technology.
>

Not to blow your own horn or anything ;-D.  It is pretty darn fast, though.

>Object-orienting the guts of the compiler has made it better (see the DMD front end source for example), but that isn't really a new *compiler* technology.
>

Thanks for the suggestions!  I've glanced at the D front end source at one point and noticed that you used some OOP in that.  I'll definitely have a good reference for real-world compiler lexical analysis and parsing technology. Thanks for making this source open.  It's sure better than trying to study something like say... gcc ...lol.

Later,

John

"John Reimer" <John_member@pathlink.com> wrote in message news:bsiikr$m4f$1@digitaldaemon.com...
> >Object-orienting the guts of the compiler has made it better (see the DMD front end source for example), but that isn't really a new *compiler* technology.
> Thanks for the suggestions!  I've glanced at the D front end source at one
point
> and noticed that you used some OOP in that.  I'll definitely have a good reference for real-world compiler lexical analysis and parsing technology. Thanks for making this source open.  It's sure better than trying to study something like say... gcc ...lol.

You're welcome. The fastest way to learn how to write a compiler is to study an existing one.

"John Reimer" <John_member@pathlink.com> wrote in message news:bsig6j$i1u$1@digitaldaemon.com...
> >I have read the earlier two editions of it, and they are excellent introductions to compiler writing. I've also taken a compiler course instructed by Profs. Ullman & Hennessey, and I learned a great deal from that course that went far beyond their book. I hope the new edition will have that extra info in it!
> That's encouraging because the basics are all I need at this point, so the
old
> book will be a great starting point.  I will likely grab the new book at
some
> point because intro describes "new" or "more indepth" chapters on: "syntax-directed translation, data-flow analysis, code optimization, just-in-time compiling, and garbage collection" among other things.  I
didn't
> think those topics were too new though...but perhaps "new" to the book.
JIT and
> garbage collection were around pre-Smalltalk days from what I recall.
Perhaps
> the techniques taught are modernized.

None of those are new, except perhaps just in time compiling (a technology I believe was invented by Steve Russell of Symantec).

> It's quite an opportunity you had learning under those professors.

What they offered was their full year compiler course compressed into two weeks one summer. It was a lot of fun to attend.

> I imagine
> information covered in those classes might not readily make it into any
> textbook.  I'd like to be pleasantly surprised to the contrary, though
:-).

I think it could be, I don't know why they haven't put it in long ago.