Hi! On Fri, Dec 28, 2007 at 06:36:42PM +0100, Marc Espie wrote: >On Fri, Dec 28, 2007 at 07:37:54AM -0600, Marco Peereboom wrote: >> Wow one comprehensive list of the suckage that is C++. Thanks Miod now >> I dont need to type examples anymore.
>I will probably regret this, but all of these issues are known, and only >show that C++ is not a simple language. >Real C++ programmers know how to work around these. C++ is all about >giving you choices. >As a list: >- No compile time encapsulation: deliberate language choice. You want to >provide clean C++ interfaces ? you use the `compiler firewall/pimpl' idiom. >As a good example, look at the KDE libraries. It's all about NOT having to >pay the extra price if you want speed to the detriment of portability. >- complicated grammar: red herring. This does NOT account for any amount of >slowness of C++ compilers. It *does* account for somewhat bad error messages. >It is complicated to build a yacc parser for C++, and thus to add meaningful >error messages. Every slow C++ compiler out there is slow for other reasons. >Mostly, that you need efficient data structures to take care of namespace >lookup (C doesn't have that issue: it doesn't have any support for namespaces). >And also, that people abuse inline functions. As far as different compilers >interpreting rules differently, C++ was not standardized until 1998... >Did you try any new-fangled C99 constructs recently ? Like, variable >definitions work everywhere on every C compiler shipped with OpenBSD ? Yeah, >right... >- No easy way to locate definition: mix of stuff inherited from C (where you >can do anything with include files, and thus you're relying on people having >bad habits) and OO practice, where it is indeed often difficult to locate >definitions... That's why smalltalk has browsers... >- No run-time encapsulation. Yeah right... as if there are no smart pointers >out there. Not worse than C. >- No binary implementation rules... partly false. There are ELF ABIs for C++, >fortunately, and Unix implementations follow them these days. That's a bit >like looking at C from 20 years ago. Between COFF, a.out, and other shit, >you did not have a binary implementation either... >Rules for types and operators are fairly complex, granted, but there is >some actual logic. This is well documented in a few books. The main issue >with C++ there is that it it a big language. You want to understand those >rules ? You have to read Scott Meyers. >Exceptions can be managed. Again you have to read Herb Sutter. IIRC in Effective C++ there's good stuff about being safe with exceptions too. >And yeah, the standard library is somewhat small. But there are useful >libraries out there developed on top of C++, and they work (qt is the >one nice example, as far as I'm concerned). qt is GPL without L in front, which makes it difficult to use from a licensing perspective in quite some settings. >All of this shows the one thing we already know, which is that C++ is >a terribly complicated language that is hard to master. When it's used >properly, by people who know how to program in it, it can be really good >and really fast. *nods* It's difficult, indeed, but one *can* write stuff, for many application areas. BTDT. (Using OpenBSD as development platform, btw.) >I'm still happily using quite a few apps written in C++, >like most of kde, and I don't see them getting rewritten in anything else >yet. >You want OO ? you've got a choice of: >- C++, which *works*. >- java, which is really slow, and not portable at all (and written partly >in C++). Not slow in my experience. Used to be, perhaps. Not any more. On OpenBSD, btw, using JDK 1.5 or 1.6, and Eclipse as development environment. Good experience because of *fast* turnaround times and a somewhat leaner language. Yes, I had the same prejudices against Java for a long time. And there will be areas where I still will prefer other languages like C or C++, or sometimes scripting languages or functional languages over Java, when/because they're a better fit. >- smalltalk, which is beautiful, very nice, and that no-one uses because it >does not talk to the rest of the world, still locked in its ivory tower. >- eiffel or ada... yeah right, everyone uses them. >- python or perl, which work just fine... perl has about the same set of >defects as C++ (except for speed and reflection). python is probably about >the same, I don't use it enough to comment. When I tried, python felt slower than perl, even. >In the end, I still see C++ as a valid choice for *seasoned* programmers. >It is very, very hard to use, and thus I would never recommend it to a >beginner. Agree. Kind regards, Hannah.
