Re: [h-e-w] c++ editing , variable name completion

[Paul Kinnucan]

Richard Y. Kim writes:
 > I've been using tags based completion for many years.
 > 

My impression is that Emacs tags-based completion blindly provides any
strings in the tags file that match the symbol at point, regardless of
whether the strings correspond to any method or field defined by the
class of the object whose method or field is sought. If this is the
case, tags-based completion falls far short of what people who are
seeking an intelligent completion facility like those found in
commercial C++ and Java IDEs are looking for in Emacs. AFAIK, the only
Emacs package that provides an intelligent completion facility
comparable to those found in commercial IDEs is the Java Development
Environment for Emacs. It is possible that ebrowse provides a
comparable facility for C++ but it is impossible to tell from the
ebrowse documentation or from the sketchy responses to my query for
information from who have used ebrowse. My impression from the
doc is that ebrowse provides the same primitive level of completion
as tags-based completion.

Intelligent completion requires that the completion system parse the
expression at point--far from a trivial task--to determine the class
of the object whose field or method is to be completed and then look
up the class to determine whether any of its classes or fields match
the expression at point. It appears that ebrowse like tags-based
completion simply provide a string database for Emacs' rudimentary
string-matching completion algorithm. If this is the case, ebrowse
completion can be no better than tags-based completion and very far
from context-sensitive completion systems, like the JDEE's, that
provide only completions appropriate to the type of the object at
point.

Never having used ebrowse, I could be mistaken in this assessment.
I'd be glad to be corrected if it is incorrect.

- Paul