|Experiences using Holub's Compiler Design in C ? email@example.com (1990-11-29)|
|Re: Experiences using Holub's Compiler Design in C ? jmd@dlogics.COM (1990-12-03)|
|Re: Experiences using Holub's Compiler Design in C ? jmd@dlogics.COM (1990-12-04)|
|From:||jmd@dlogics.COM (Jens M. Dill)|
|Summary:||some impressions of the book|
|Organization:||Datalogics Inc., Chicago|
|Date:||3 Dec 90 13:53:58 GMT|
In article <9011291759.AA06388@neuron.tamu.edu>,
firstname.lastname@example.org (Tim McGuire) writes:
> Has anyone out there used Allen Holub's Compiler Design in C for
> a compiler course?
I haven't used Holub's book for teaching, but I have used it as a reference
for compiler development and I have taught compiler courses from Aho et al.
Here are my gut reactions to Holub's book.
1. It is a very readable, practical, down-to-earth view of the subject.
Students are going to find it easier to use than any of its competitors,
especially if the course focuses on practical aspects as opposed to
2. It shows signs of being hastily assembled. Minor errata are frequent,
and you need to be on your toes. Read everything first, then try it
out, and make sure you are sure that it works. Most of it does, but
you don't want to be caugth when it doesn't. By all means, contact
Gene Holub and make sure you have the latest errata listing. He's
been doing a good job keeping up with it, as near as I can tell.
But don't throw out the book due to the errata.
3. The book contains one serious technical error. The definition of
ambiguity, as applied to context-free grammars is just plain wrong.
It is the work of about five minutes for a language theorist to
construct examples of unambiguous grammars that satisfy Holub's
definition of ambiguity and ambiguous grammars that don't. [Gene,
if you are listening, I'd meant to open a discussion with you on
the subject, but my new employer's outgoing E-mail connections are
a pain. -- JMD]. Again, this is not a reason to avoid the book.
Hope this helps. And good luck.
Return to the
Search the comp.compilers archives again.