GOB
GObject Builder (GOB) is a simple preprocessor for easily creating GObjects (glib objects). It reads a class description file with inline C code and produces .c and .h object files for you. It allows you to easily use most of the features of the GObject system and avoids typos by reducing the amount of code needed. In general, the amount of code is about the same as for OO languages such as Java or C++, but you still have to only use C and you get all the power of the GObject system.
| Tags | Software Development |
|---|---|
| Licenses | GPL |
Recent releases
- 20 Nov 2007 23:53
No changes have been submitted for this release.
- 05 Jan 2006 10:36
Changes: A major segfault has been fixed.
- 16 Dec 2005 14:04
Changes: This release fixes libglade support, adds abstract class support, has better boxed support, and has better spurious rebuild prevention.
- 22 Jul 2005 23:46
Changes: libglade support, properties updates, and other minor changes.
- 22 Sep 2004 21:21
Changes: This release added support for overriding properties and changing of file separators. Private structures now play nice with glib. C++ compilation fixes and M4 fixes were made, and string.h is now always included.
- All comments
Recent commentsRe: C vs. GTK object system
>
> It's not too different. That's good.
> But it gives you less code, clearer
> code, and compile-time type-safety.
> That's good too.
>
GOB gives you quite a bit of compile time type
safety as well. Though with a focus that this
will all work from C code (that is not all your
code has to be in GOB, and in fact shouldn't).
Though compile time typesafety is many times
overrated. I'm more of a fan of runtime
typesafety, since it will catch all type errors,
though with the caveat that the code must run for
you to find errors. That said a certain amount of
static typesafety always helps.
>
> I don't think it's an advantage that
> GTK+ in C makes subclassing difficult.
> Gtk-- in no way forces you to subclass,
> or emphasises that technique. That's
> just something that's available to you,
> whereas the difficulty of doing it in C
> means that people don't choose that
> design technique when they should. And
> code becomes less clear.
> I guess GOB is meant to fix this. It's
> just that C++ seems like a nicer way to
> do that.
>
I didn't mean that this is because GTK+ makes it
hard to subclass. Just that the focus is
different. Yes everything that GTK+ object system
does is possible in C++, and the other way around
as well (though if you wish to do some things in C
you get into some deep voodoo land).
I'm still way more profficent in C, and don't
enjoy C++ very much. You can usually stare at a
piece of code in C and figure out what it does
exactly. With C++ there could be many things
going on behind the scenes that you would have to
look up in the class definition. Example, in C no
code is run until you explicitly call a function.
In C++ things like constructors, destructors and
various overloaded functions get run even when the
syntax doesn't say something would run.
And this is where the original C++ (or cfront)
differ substantially from GOB. all GOB does is
write the boilerplate for you. It does not
attempt to parse or change C code. It does not
understand C almost at all (all it can do is count
parenthesis and braces).
I have made an object system that is statically
typesafe and easy to derive purely using the C
preprocessor. However GTK+ doesn't do this. I'm
very pragmatic on this point. I want to use C. I
like GTK+. Thus GOB. GOB doesn't have the things
that I don't like about C++, and it has an easy
syntax for making objects, while allowing me to
still just use C. I'd be happier if GTK+ object
model didn't require so much boilerplate and was
more c preprocessor friendly, but it isn't. GOB
isn't written for the purpose of writing GOB, it
is something to scratch an itch.
Re: C vs. GTK object system
> You can get
> things like
> signals for C++ with addon libraries,
> and you can
> get a type system for C++ with addon
> libraries.
> You can get named arguments as an
> addon too.
Yes, you can get them all from the same 'add-on library'. That's a language binding.
> But
> then it's not much different then
> coding with
> GTK+.
It's not too different. That's good. But it gives you less code, clearer code, and compile-time type-safety. That's good too.
> If a project for example uses
> GOB, it
> only uses GOB for those few classes it
> has. It
> does not emphasize subclassing for
> things it's not
> needed for.
I don't think it's an advantage that GTK+ in C makes subclassing difficult. Gtk-- in no way forces you to subclass, or emphasises that technique. That's just something that's available to you, whereas the difficulty of doing it in C means that people don't choose that design technique when they should. And code becomes less clear.
I guess GOB is meant to fix this. It's just that C++ seems like a nicer way to do that.
Re: C vs. GTK object system
>
> I'd be interested to know how you
> think that the C GTK object system is
> more flexible rather than just more
> long-winded.
>
>
It depends on your priorities for coding. C++ is
quite a different beast. You can get things like
signals for C++ with addon libraries, and you can
get a type system for C++ with addon libraries.
You can get named arguments as an addon too. But
then it's not much different then coding with
GTK+. The GTK+ (and GObject nowdays) comes with
all these and the unerlying language is much
simpler to understand. With GTK+ objects the
emphasis isn't given on subclassing and making
everything as an object. You still do mostly C
coding. If a project for example uses GOB, it
only uses GOB for those few classes it has. It
does not emphasize subclassing for things it's not
needed for. As owen once put it, it is more a
component system then an object system
I wonder why everybody mentions C++ as "the OO
language". If I was to mention a language that
GTK+/GOB would be closer to, it would be java or
Objective C. At some point in the future I plan
to use Java, when that gets more mature. Java is
very nice as a language, but I'm not too crazy
about it's libraries and compiler support.
Re: Back to the past
I meant 'saying that moc requires only C plus plus', but the plus sign doesn't show up.
Re: Back to the past
Quite right. I don't see how this can be said to require only C. That 's like QT saying that moc requires only C . If you want a better C, then it's already been done.