At 05:44 PM 7/14/00 +0100, [EMAIL PROTECTED] wrote:
>Hi folks,
>
>I have a couple of questions and observations that I would be grateful for
>
>answers to/reflections on;
>
>1) I have a like like "private Dimension screenSize = null;". The
>"Dimension" part is of color A. I change "Dimension" to "dimension" and
>the color changes. This makes me a bit nervous, as it impiles that
>coloring is done based on coding standards. This is usually fine by me,
>but I can not help but think that things should be coloured after what
>they are (keyword/builtin/etc) and not by what they look like they are.
>Is this so? If so, I'd be interested in the explanation =)
>
Syntax coloring is done by font-lock, the standard Emacs syntax coloring
package. The coloring done by this package is based primarily on syntax.
However, Java, like other object-oriented languages, presents a problem for
syntax coloring: its type system is extensible. It is very easy to
determine whether any given symbol refers to a built-in type. You simply
check it against a list of built-in types. But how do you determine if a
symbol refers to a user-defined type? The font-lock package solves this
problem, I believe, by taking advantage of the common Java coding
convention that capitalized symbols refer to user-defined types. AFAIK, the
only alternatives would be to search the classpath for every symbol not
otherwise defined or constantly build and update a database of user-defined
types. I prefer the existing solution as I believe the others would be too
slow.
>2) The speedbar is great for navigating. However, as a Java programmer I
>would very much like it to be an "object browser". Browsing Java code by
>files somehow doesn't feel very elegant, especially as with
>inner/anonymous classes code files tend contain more code. I know of a few
>different browsers, but shouldn't something like that be a priority in a
>Java development environment?
>
This question comes up fairly often. The BeanShell, which comes with the
JDE, includes a browser. To use it, type M-x jde-browse-class with point on
any JDK class name (e.g., String). Unfortunately, the BeanShell class
browser has a limitation, it only knows about classes that exist in jar
files on the classpath. The author of the BeanShell is aware of this
limitation and promises to remove it in a future release. As far as the
class browser being a priority, sure, but the JDE basically has one
part-time developer to assign to the problem and that developer would
rather work on some other issues at the moment, like the debugger and the
completion facility
>3) jde-complete-at-point usually takes quite some time (I use a PIII 500).
>
>Will this change?
>
The completion facility is basically a partially functioning prototype as
I've noted more than once before on this list. It is a priority of mine to
complete its implementation, including addressing performance issues.
>4) jde-help-symbol takes you to the page for the specified class, like
>".../lang/IllegalArgumentException.html". I'd like to see the frames
>version instead. I imagine this could be done by creating a frameset as
>all the information on what pages to display is at hand. Has anyone
>thought of this? Anyone else interested?
>
I'm more interested in having it take you to a particular field or method
in the page for the class, which is another planned enhancement. I could
take a look into the issue of trying to open the javadoc frameset and
displaying the right page within the frameset.
>5) jde-help-symbol doesn't seem to work for static methods. Correct?
>
I don't know. I'll take a look.
>6) When I compile and use next-error, the point is inserted the the
>beginning of the line with an error. Is there a way to make it insert the
>point at the same line but where the problem actually is (according to the
>compile)?
>
The Emacs compile.el package handles error hyperlinking. Perhaps you could
investigate how compile works and either provide an enhancment to do this
or provide a recommendation for how to implement the required enhancement.
- Paul
