🚀 go-pugleaf

In article <1992Sep30.101144.11382@dutrun2.tudelft.nl>, rcpshdb@dutrun2.tudelft.nl (Han de Bruijn) writes:
|> In article <1992Sep27.022750.8620@leland.Stanford.EDU> Laurence James Edwards:
|> > Wow! I had forgotten what a godly book Preparata&Shamos is.
|> 
|> In article <1992Sep28.083843.23616@dutrun2.tudelft.nl> Han de Bruijn:
|> > Bad taste.
|> 
|> In article <1992Sep28.231335.10973@leland.Stanford.EDU> Laurence James Edwards:
|> > While I wouldn't recommend it for recreational reading, Preparata&Shamos does
|> > present a comprehensive and coherent account of computational geometry,
|> > and I think it is the first book devoted to the subject.
|> 
|> Trivialiter, because Preparata&Shamos are the _inventors_ of "Computational
|> Geometry". Same story with "The Fractal Geometry of Nature", by Mandelbrot,
|> which is, equally trivialiter, the "first" book about fractals.

Well the term, Computational Geometry, was first applied to this
particular field of study in a 1975 paper by Shamos. The book is
copyright 1985, so it is not trivially evident that it would be the first
book on the subject.

|> Actually, the stuff covered by Preparata & Shamos IMNSHO barely justifies the
|> creation of a new, separate, area of investigation. A most important part of
|> it (Voronoi tesselation) is nothing new & belongs to classical mathematics.

I didn't realize they did big oh analyses of algorthms in classical
mathematics.

|> The book doesn't contain obvious generalizations to multiple dimensions
|> in the first place. Even worse. Solutions in 2-D are presented in such a
|> way that a generalization from that point to, say, three dimensions is
|> made impossible beforehand.

That simply is not true. For example, convex hull algorithms are generalized
to n-dimensions, and there are a number of sections on extending 2-D results
to higher dimensions.

|> One shouldn't use specific properties of the
|> plane (like orientation of angles) in a theory that could be generalized
|> to three and more dimensions.

I assume you mean the point in polytope problem here. The theory is
not about general algorithms that work in all dimensions but about worst
case analysis for a given problem. The authors point out that very
little is known about point location problems in E^3 and higher dimensions.

|> Overall, the attacks presented in the book
|> are quite unprofessional. Any experienced programmer could do better.

Ahem, "unprofessional"? First off, the book is *an introduction*. Second,
they do not claim to present professionally written code, the point
of the book is the theoretical worst case analysis of possible algorithms
answering a number of related geometric queries. Big oh worst case analysis
does not always lead to algorithms well suited for a particular application,
the authors specifically admit that.

Larry Edwards