Re: first-class support for csharp in cc-mode.el

[Alan Mackenzie <acm@muc.de>]

Hi, Dino,

On Wed, Dec 30, 2009 at 08:29:14AM -0800, dino chiesa wrote:

> Hi Alan, nice to meet you.
> Stefan, thanks for the input. 

> acm@muc.de wrote:
> > The current policy is not to add any more languages to the core of CC
> > Mode, since it's hassle enough supporting the seven that are already
> > there. There are certainly lots of users of C, C++, Java, several
> > users of Objective C and AWK, but I just don't know about Pike and
> > IDL. The problem is, when somebody implements support for a language
> > directly in CC Mode and they move on a few years later, support for
> > that langauge effectively vanishes. It was to fix this that Martin
> > Stjernholm (my predecessor) implemented the c-lang-const mechanism.

> Very clear. In that case I should be petitioning to get csharp-mode put
> into the emacs progmodes distrib separately, rather than petitioning to
> get csharp-mode folded into cc-mode.

I think so.  I doubt Stefan will take that much persuading.  ;-)

> > If the c-lang-const stuff is inadequate to support, say, C#, then I
> > would prefer to amend CC Mode so that C# can be supported. It looks
> > like the @"....." syntax will need such an amendment. This will
> > probably involve a large change, since the use of \ as an escape
> > character is firmly embedded in the very fabric of CC Mode. Is @ used
> > for anything else in C#? Is it forbidden to have whitespace between @
> > and "?

> Yes, according to the lang spec, the @ must immediately be followed by
> a double-quote. The symbol is not used for anything else in the
> language. I just looked and this is called a "verbatim string literal".
> http://www.go-mono.com/docs/index.aspx?link=ecmaspec%3a9.4.4.5

> > In fact, I've just thought of a way of fixing this: in an
> > after-change function, splat the syntax (using syntax-table
> > text-properties) of anything within @"....". I think there's already
> > a suitable hook in CC Mode for this.

I've had a look at this.  The two hooks you want are
c-get-state-before-change-function and c-before-font-lock-function, both
of which are c-lang-defconsts.  In the function you put in the first
hook, you should see whether or not you're about to disrupt the status of
one or more @"......\" constructs.  There'll be some cunning special
cases to check here.  The function on the second hook will actually add
the text-property to the pertinent \, or possibly remove it.

> Is this the approach used for handling of slashes within awk? (can be
> either a divide operator or a string delimiter). I tried wading through
> that code but didn't yet figure it out.

:-)  I wrote that code some while ago, and I'm not sure I could figure it
out any more.  It's a fairly involved finite state machine implemented by
regular expressions.

> In csharp, it is not only a trailing slash that is special - the
> verbatim string also allows verbatim newlines in the string. This is
> legal:

> var s = @"The quick brown fox
> jumps over the lazy dog";

> and results in a string with a \n between the 4th and 5th words.

> I don't know how that might affect the final design of a solution, or
> if it recommends or disrecommends Stefan's suggestion to worry only
> about the final slash. 

I couldn't say off-hand either.  I suspect it would just work.

> >> I have no idea if this is a friendly proposal or not, but it would
> >> make things nice for people who write C# and use emacs.

> > Thanks for the proposal! Constructive suggestions are always welcome at
> > CC Mode. By the way, CC Mode has its own mailing list at
> > bug-cc-mode@gnu.org and is hosted on sourceforge for historical reasons.


> Ah, very good. Is bug-cc-mode intended to serve discussion related only
> to bugs or potential bugs, or is it also good for random discussion or
> questions from cc-derived modes? I mean, if I were to ask specifically
> about the verbatim string literals, should I ask on emacs-devel or is
> it better on bug-cc-mode ?

Probably better on bug-cc-mode, but more people will see it on
emacs-devel.

> >> There are still a few issues with that C# mode - lambdas have no
> >> special support, and the new dynamic keyword in .NET 4.0 isn't there
> >> (simple to add), and there is the literal string thing I described
> >> in my other mail thread - but mostly it works well. Including this
> >> would provide a good baseline of support in emacs for C#.

> > Those are all issues for C# Mode's maintainer.


> Right, those are issues for me (and friends) to own. I wasn't really
> asking for *that*, although I may ask for specific advice on some of
> those in the future, after I take a run at them, I hope you don't mind.

> My real goal regarding "first class support" is to make csharp-mode
> official and standard, so that someone downloading emacs 24 or 23.x
> will just get it. I had originally thought that meant major surgery on
> cc-mode, but you're telling me it could be simpler than that.

It should be, and if it's not, I should be able to make it so.

-- 
Alan Mackenzie (Nuremberg, Germany).