CEDET

On 06/08/2012 03:22 AM, Fredrik Malmros wrote:
> Hi,
>
> There seems to be an issue with visibility when completing members with
> default visibility i.e. members or classes without access modifier in Java.
> When calling semantic-analyze-possible-completion with point at ^ in the
> class below, I get:
> void baz(), but I expect to get both void baz() and void bar() since bar
> is visible in Foo (and all other classes in the same package as Foo).
>
> public class Foo {
>       void bar () {}
>
>       private void baz () {
> b
>            ^
>      }
> }

Hi,

Sorry for the late reply.

The java utilities return the protection of 'bar' as being 'nil', and 
'baz' as being 'private'.  I think all that is needed is for java to 
override semantic-tag-protection to return private if nothing was 
specified.  (If private is the right thing.  I'm not that familiar with 
java.

This patch should solve the problem if 'private is the right answer. 
Please let me know if unspecified protection means private.

Thanks
Eric

------------

=== modified file 'lisp/cedet/semantic/java.el'
*** lisp/cedet/semantic/java.el	2011-11-07 20:57:12 +0000
--- lisp/cedet/semantic/java.el	2012-06-16 02:14:07 +0000
***************
*** 173,178 ****
--- 173,187 ----
             (semantic-find-tags-by-class
              'type (semantic-find-tag-by-overlay point))))

+ ;; Tag Protection
+ ;;
+ (define-mode-local-override semantic-tag-protection
+   java-mode (tag &optional parent)
+   "Return the protection of TAG in PARENT.
+ Override function for `semantic-tag-protection'."
+   (let ((prot (semantic-tag-protection-default tag parent)))
+     (or prot 'private)))
+
   ;; Prototype handler
   ;;
   (defun semantic-java-prototype-function (tag &optional parent color)

Actually, no qualifier means that it has package visibility.  So it should
be visible to any class in the same package.  I'm not sure if semantic
knows the package for a java class, but it should not be difficult to
determine it.

Troy

On Fri, Jun 15, 2012 at 10:17 PM, Eric M. Ludlam <ericludlam@...:

> On 06/08/2012 03:22 AM, Fredrik Malmros wrote:
> > Hi,
> >
> > There seems to be an issue with visibility when completing members with
> > default visibility i.e. members or classes without access modifier in
> Java.
> > When calling semantic-analyze-possible-completion with point at ^ in the
> > class below, I get:
> > void baz(), but I expect to get both void baz() and void bar() since bar
> > is visible in Foo (and all other classes in the same package as Foo).
> >
> > public class Foo {
> >       void bar () {}
> >
> >       private void baz () {
> > b
> >            ^
> >      }
> > }
>
> Hi,
>
> Sorry for the late reply.
>
> The java utilities return the protection of 'bar' as being 'nil', and
> 'baz' as being 'private'.  I think all that is needed is for java to
> override semantic-tag-protection to return private if nothing was
> specified.  (If private is the right thing.  I'm not that familiar with
> java.
>
> This patch should solve the problem if 'private is the right answer.
> Please let me know if unspecified protection means private.
>
> Thanks
> Eric
>
> ------------
>
> === modified file 'lisp/cedet/semantic/java.el'
> *** lisp/cedet/semantic/java.el 2011-11-07 20:57:12 +0000
> --- lisp/cedet/semantic/java.el 2012-06-16 02:14:07 +0000
> ***************
> *** 173,178 ****
> --- 173,187 ----
>             (semantic-find-tags-by-class
>              'type (semantic-find-tag-by-overlay point))))
>
> + ;; Tag Protection
> + ;;
> + (define-mode-local-override semantic-tag-protection
> +   java-mode (tag &optional parent)
> +   "Return the protection of TAG in PARENT.
> + Override function for `semantic-tag-protection'."
> +   (let ((prot (semantic-tag-protection-default tag parent)))
> +     (or prot 'private)))
> +
>   ;; Prototype handler
>   ;;
>   (defun semantic-java-prototype-function (tag &optional parent color)
>
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> cedet-semantic mailing list
> cedet-semantic@...
> https://lists.sourceforge.net/lists/listinfo/cedet-semantic
>

2012/6/16 Eric M. Ludlam <ericludlam@...>

> On 06/08/2012 03:22 AM, Fredrik Malmros wrote:
>
>> Hi,
>>
>> There seems to be an issue with visibility when completing members with
>> default visibility i.e. members or classes without access modifier in
>> Java.
>> When calling semantic-analyze-possible-**completion with point at ^ in
>> the
>> class below, I get:
>> void baz(), but I expect to get both void baz() and void bar() since bar
>> is visible in Foo (and all other classes in the same package as Foo).
>>
>> public class Foo {
>>      void bar () {}
>>
>>      private void baz () {
>> b
>>           ^
>>     }
>> }
>>
>
> Hi,
>
> Sorry for the late reply.
>
> The java utilities return the protection of 'bar' as being 'nil', and
> 'baz' as being 'private'.  I think all that is needed is for java to
> override semantic-tag-protection to return private if nothing was
> specified.  (If private is the right thing.  I'm not that familiar with
> java.
>
> This patch should solve the problem if 'private is the right answer.
> Please let me know if unspecified protection means private.
>
> Thanks
> Eric
>
> ------------


> === modified file 'lisp/cedet/semantic/java.el'
> *** lisp/cedet/semantic/java.el 2011-11-07 20:57:12 +0000
> --- lisp/cedet/semantic/java.el 2012-06-16 02:14:07 +0000
> ***************
> *** 173,178 ****
> --- 173,187 ----
>            (semantic-find-tags-by-class
>             'type (semantic-find-tag-by-overlay point))))
>
> + ;; Tag Protection
> + ;;
> + (define-mode-local-override semantic-tag-protection
> +   java-mode (tag &optional parent)
> +   "Return the protection of TAG in PARENT.
> + Override function for `semantic-tag-protection'."
> +   (let ((prot (semantic-tag-protection-**default tag parent)))
> +     (or prot 'private)))
> +
>  ;; Prototype handler
>  ;;
>  (defun semantic-java-prototype-**function (tag &optional parent color)
>
> Hi,

no modifier i.e "package private" makes a field or method
visible everywhere in its own package.  So unspecified and private does not
mean the same thing; however, when completing members defined in the class
we are in, the the modifier doesn't really matter since everything is
visible in its own class.

I think the following sums the rules up:

public - accessible everywhere.
protected - accessible to the package classes and subclasses (that might be
defined in other packages).
no modifier "package private" - accessible to classes in the same package.
private - accessible within the class.


/Fredrik

On 06/18/2012 04:05 AM, Fredrik Malmros wrote:
> Hi,
>
> no modifier i.e "package private" makes a field or method
> visible everywhere in its own package.  So unspecified and private does
> not mean the same thing; however, when completing members defined in the
> class we are in, the the modifier doesn't really matter since everything
> is visible in its own class.
>
> I think the following sums the rules up:
>
> public - accessible everywhere.
> protected - accessible to the package classes and subclasses (that might
> be defined in other packages).
> no modifier "package private" - accessible to classes in the same package.
> private - accessible within the class.


Hmmm.   There is no 'package' concept in Semantic.  Is this a common 
concept for different languages, or unique to Java?

I fiddled with this some and created the attached patch.  Checking for 
package level protection is done by comparing the package of the 
discovered tag against that of the current buffer.  (ie - the buffer the 
completion is being done in.)   Is that accurate.  It is for your 
previous simple use case, but that is less clear if someone strung a 
bunch of things together.

It would be great if someone could check the correctness of this patch.

Thanks
Eric

I checked in the patch I had proposed below.

There should now be some rudimentary support for package protected slots 
in Java.

For others, the overload methods semantic-tag-full-name will now 
possibly start returning different values, as it accounts for package names.

Eric

On 06/21/2012 11:35 PM, Eric M. Ludlam wrote:
> On 06/18/2012 04:05 AM, Fredrik Malmros wrote:
>> Hi,
>>
>> no modifier i.e "package private" makes a field or method
>> visible everywhere in its own package. So unspecified and private does
>> not mean the same thing; however, when completing members defined in the
>> class we are in, the the modifier doesn't really matter since everything
>> is visible in its own class.
>>
>> I think the following sums the rules up:
>>
>> public - accessible everywhere.
>> protected - accessible to the package classes and subclasses (that might
>> be defined in other packages).
>> no modifier "package private" - accessible to classes in the same
>> package.
>> private - accessible within the class.
>
>
> Hmmm. There is no 'package' concept in Semantic. Is this a common
> concept for different languages, or unique to Java?
>
> I fiddled with this some and created the attached patch. Checking for
> package level protection is done by comparing the package of the
> discovered tag against that of the current buffer. (ie - the buffer the
> completion is being done in.) Is that accurate. It is for your previous
> simple use case, but that is less clear if someone strung a bunch of
> things together.
>
> It would be great if someone could check the correctness of this patch.
>
> Thanks
> Eric
>
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>
>
>
> _______________________________________________
> Cedet-devel mailing list
> Cedet-devel@...
> https://lists.sourceforge.net/lists/listinfo/cedet-devel