Re: Minor bug in cc-menus.el: cc-imenu-java-generic-expression does not match all Java 1.5+ function definitions.

[David Kastrup <dak@gnu.org>]

Alan Mackenzie <acm@muc.de> writes:

> sorry about the delay replying - I'm the guy who does the CC Mode
> things, and sadly there's only one of me at the moment.
>
> On Sat, Jul 18, 2009 at 08:27:10AM -0700, Nathaniel Flath wrote:
>> I was using Imenu, and noticed that it currently doesn't recognize Java
>> functions with either:
>> Generics e.g : public List<String> foo() { ...
>> or annotated arguments, e.g : public void foo( @NonNull bar ) {
>
> CC Mode has got a bit behind in handline C++ templates and Java generics
> (which are essentially the same thing from Emacs's point of view).
> There's quite a bit of work to be done on updating Java Mode's support
> for recent enhancements to the language.
>
> <rant>
> Misusing "<" and ">" as template/generic delimiters was an egregious
> "design" decision by the C++ people,

It was actually a design decision by the Ada people.  Ada uses <...> for
constraints in lots of contexts, and generics parameters are
constraints.

It was abysmally stupid to let C++ templates look like Ada generics, in
particular since C++ has << and >> operators.  But then C++ syntax is a
collection of stupidities, parseable only by hand-written recursive
descent parsers and they still tend to get the corner cases wrong (while
human readers tend to get surprised by quite more cases).  And it is
already stupid to have a preprocessor with a different syntax, but that
particular stupidity is inherited.

C++ is an attempt to cram not only everything that can be somehow made
to fit into the functionality, but also into the syntax.  And that's
insane.

> There are pieces of C++ code which need a compiler (or some equivalent
> tool) to distinguish "less than" from "template start".  There might
> also be similar Java code, I'm not sure.  As a result, syntactic
> analysis of C++ (and maybe Java) code which contains "<" and ">" is
> either difficult or impossible in the general case.

And nowadays, >> for ending two nested template parameter sets is
supposed to be handled properly by the compiler.  Making the problem
still more complex.

In a well-designed language, recognizing syntactic elements does not
need much context.  Nobody wants to read computer languages in the style
of classic Greek where a single sentence spreads itself over more than
one page and happens not to actually have a verb predicate, so you go
back and forth until you finally manage to make sense of the nested
somewhat arbitrary constructs.

If I wanted to do that, I would learn German.  Oh, wait...

-- 
David Kastrup