> Some problems are complex and require complex solutions.
And yet some very complex solutions were solved with very simple languages, before Java was conceived.
And yet some very complex solutions were solved with very simple languages, before Java was conceived.
So we're stuck with these complicated compile-time rules for defining everything in terms of a procedural language. We come up with complicated templates and generics for getting one set of methods to work with two types of data. Objects are needlessly serialized and deserialized to pass between systems, and programmers are needlessly spending time writing such methods.
I'm surprised that scala's paradigm of having actors work on data hasn't caught on more. When you separate them from the start, new modules are way easier to right, and integration time drops to near zero.
I do like the way jfb suggests using the concept of 'affordances' with languages/environments, to analyze what sorts of code they encourage. But to really do that, we'd have to actually identify these specific affordances, which i'm not sure how to either.
(And as far as your particular line of argument -- does anything in a processor know what an 'actor' is either? If not, I believe you that you find 'actor' a better abstraction than 'class', but it must not have anything to with either one being something the lower-level computer architectural abstractions 'know about')
There's a saying that one can write COBOL in any language, but nobody is born itching to POST INCREMENT C BY ONE. A given technology has a spirit; that spirit guides the usage of the technology; and people who use that technology, and share a sufficiency of that spirit, will find a welcoming home. That spirit is intrinsic to the technology; changing Java to not rely on Simula style classes for code namespacing would destroy Java -- it'd be something very different, and the usage of this new language would mean entirely new styles of conceptual organization.
I'm not taking a stance in this post on the aesthetics of the various ways to organize programs; but I do think that the forensic exercise of determining the -- I dunno, zeitgeist? -- of a technology is an interesting, rich one.
Or more accurately, the problem is that Java was a new language shoehorned into being "like" a language that was never meant to hold the concepts that Java demands. In some cases this manifests itself in more syntactical ways; C's pragmatic but unsophisticated approach typing works well for C, but starts to become a real burden when you start adding more sophisticated concepts onto the language. (Were generics really a concept that were ever appropriate for something like C? Generics deserve a more considered treatment; a language with structure and syntax designed with them in mind.) In other cases the impact is more clear but inexplicable; "In Java Everything is an Object"(tm)... except for the things that are not.
The end result is a language that is ugly and clumsy to work with without some pretty advanced tooling, all for reasons that really have no business being real problems. Shameful (primarily cosmetic) flaws forced upon Java by the business decision: "make it reminiscent of C".
I have the same complaint about C++. Had Java and C++ not burdened themselves with the design requirement of looking like C, they would be much better languages. Languages that stray further from the mold, like Scala or Go (I am thinking particularly of their abandonment of C inspired defines, in favor of a more Pascal approach), are more pleasant to work with for it. Perhaps this damages adoption.. but that is another matter I think.
The same could be said about Ruby and it's relationship to Perl and Python. It has cultural roots in those languages, but is often regarded to be better rather than worse. (As a die-hard Perl developer I disagree, but I don't hold a grudge.) All three are also considered to be C-like languages too, and they don't suffer the problems Java has.
One could argue that Java's problem was that it was designed for programming VCRs, DVD players, and Cable-TV boxes, and similar equipment, but the got stretched and pulled into environments it just wasn't ready for. That, plus Sun's hyper-aggressive marketing machine to turn Java into the language-for-everything, created the beast that it has become.
My personal view is that Sun turned Java into a programming language for charge-by-the-hour consultants. The culture of over-engineering exists because it takes a lot more time to design software that way, more time and larger teams to implement software that way, and specialized hard-to-transfer knowledge to maintain that kind of software. This all benefits the consultants and hurts the companies paying for the software to be built, which only makes sense if Sun was intentionally pushing Java that way.
I wouldn't describe C# as a language that I like, but I think it too has benifited from breaking the mold of C more than Java did (though I attribute most of it's superiority to its designers simply making better decisions in several places). Named and optional arguments stand out in my mind as ways C# did something that had no precedence in C, while Java still has not. (There is that double squiggly bracket trick, but you have to admit that is pretty shoddy).
For better or worse, Java has much simpler semantics.
PS: THE Java® SE 7 Edition of The Java Language Specification describes all the features that have been added to the Java programming language in Java SE 7. It also integrates changes made to the Java programming language under maintenance since the Third Edition in 2005. http://docs.oracle.com/javase/specs/jls/se7/jls7.pdf. It's 670 pages long.
Unfortunately, among the major industrial programming languages, that is far from the longest specification out there. Moreover, some of the major industrial programming languages don't even have a standard reference specification. Notwithstanding its length, the Java spec is actually fairly efficient and well-organised in comparison.
Unfortunately, Java is also not a very powerful language. Many of those OO gymnastics are only relevant because the language isn't expressive enough to achieve the same results in a simpler, more direct way. Java is the language where design patterns in the Gang of Four sense first became popular, and it's often been suggested that design patterns indicate missing language features.
Exactly right. In previous decades, had they been using modern terminology, calling conventions would no doubt be covered by plenty of "design patterns". We don't speak of calling conventions that way now because languages provide that for us.
And if we actually used those languages, we would probably be seeing much lower bug rates in production software already and we could start looking into what else could be improved next. Unfortunately, most of the popular industrial languages don't even get that far. :-(
Paradoxically, this assignment has given me a great deal of respect for the idiots who conceived of the damned thing in the first place.