Re: RFR cc-langs.el: change syntax for @ in Java

[Alan Mackenzie <acm@muc.de>]

Hello, Filipp.

On Thu, Apr 08, 2021 at 20:26:49 +0300, Filipp Gunbin wrote:
> Hi Alan,

> On 08/04/2021 10:15 +0000, Alan Mackenzie wrote:

> > I've just run the CC Mode test suite with this amendment, and it
> > produces the following successes on annotations-1.java:

> > annotations-1.java:4: Expected analysis ((inclass 1) (topmost-intro 21)), got ((inclass 21) (topmost-intro 21))
> > annotations-1.java:4: Expected indentation 0, got 32
> > annotations-1.java:5: Expected analysis ((class-close 1)), got ((class-close 21))

> > ..  So, there are things to sort out before we can apply the change.

> I think these breakages are not real breakages, but rather the resfile
> could be updated.

I've had a more thorough look at it.  In cc-langs.el, there are several
places where @ is treated as a symbol character, namely:

L375: c-make-mode-syntax-table (which you have proposed modifying)
L419: c-identifier-syntax-modifications
L636: c-symbol-start
L1969: c-paragraph-start (I think this one only applies in comments)
L2198: c-class-decl-kwds (The "@interface" keyword)

We need to consider whether to amend some or all of these places, so
that @ ceases to be a (starting) character of an identifier.

I'm also a bit worried that in

    @NonNull
    @TestClass
    @FooBar
    public class Annotations {

, the last line gets parsed as ((annotation-top-cont 21)), where it
really ought to be a topmost-intro.  But this problem is present without
your patch, anyway.

In summary, I think the motivation for the change (highlighting symbols
correctly) is valid, but the change is going to be more involved than
the patch you've supplied.

Do you want to look at all these things, or should I?


> This class:

> --8<---------------cut here---------------start------------->8---
> @NonNull
> @TestClass
> @FooBar public class Annotations {

> }
> --8<---------------cut here---------------end--------------->8---

> previosly "started" at the @NonNull line, but now "starts" at @FooBar
> line, which is absolutely correct.

Are you saying here that @FooBar is actually a part of the declaration
of class Annotations?  If so, why not @NonNull and @TestClass as well?

> So here's the updated patch (now against the cc-mode repo).
> Thanks.

Thank you!

> diff -r 10abd4587dad cc-langs.el
> --- a/cc-langs.el	Mon Mar 29 15:20:53 2021 +0000
> +++ b/cc-langs.el	Thu Apr 08 20:22:56 2021 +0300
> @@ -365,12 +365,14 @@
>         (let ((table (make-syntax-table)))
>  	 (c-populate-syntax-table table)
>  	 ;; Mode specific syntaxes.
> -	 ,(cond ((or (c-major-mode-is 'objc-mode) (c-major-mode-is 'java-mode))
> +	 ,(cond ((c-major-mode-is 'objc-mode)
>  		 ;; Let '@' be part of symbols in ObjC to cope with
>  		 ;; its compiler directives as single keyword tokens.
>  		 ;; This is then necessary since it's assumed that
>  		 ;; every keyword is a single symbol.
>  		 `(modify-syntax-entry ?@ "_" table))
> +		((c-major-mode-is 'java-mode)
> +		 `(modify-syntax-entry ?@ "'" table))
>  		((c-major-mode-is 'pike-mode)
>  		 `(modify-syntax-entry ?@ "." table)))
>  	 table)))
> diff -r 10abd4587dad tests/annotations-1.res
> --- a/tests/annotations-1.res	Mon Mar 29 15:20:53 2021 +0000
> +++ b/tests/annotations-1.res	Thu Apr 08 20:22:56 2021 +0300
> @@ -1,8 +1,8 @@
>  ((topmost-intro <0,0>))
>  ((annotation-top-cont <1,0>))
>  ((annotation-top-cont <1,0>))
> -((inclass <3,0>) (topmost-intro <1,0>))
> -((class-close <4,0>))
> +((inclass <1,0>) (topmost-intro <1,0>))
> +((class-close <2,0>))
>  ((topmost-intro <1,0>))
>  ((topmost-intro <2,0>))
>  ((inclass <1,0>) (topmost-intro <1,0>))

-- 
Alan Mackenzie (Nuremberg, Germany).