Re: Emacs learning curve

[Eric M. Ludlam]

On 07/13/2010 10:49 PM, Stephen J. Turnbull wrote:
> David Engster writes:
>
>   >  As you know, a language mode consists of different things. For instance
>   >  font locking, indentation, accessing some kind of documentation, or
>   >  communication with an interpreter. CEDET doesn't really immediately have
>   >  much to do with those, although font locking could probably profit from
>   >  a semantic parser and semantic tags can also contain documentation;
>   >  indentation is entirely another beast, though.
>   >
>   >  I think all that was said was that CEDET can provide stuff like semantic
>   >  parsing, project support and template generation under a general
>   >  framework, so that other tools (like ECB) can use that information, and
>   >  therefore it might be possible to unify those areas of a language mode.
>
> Maybe that's all that was intended to be said, but the requirement was
> for a unifying framework *now*, and it was claimed that CEDET fits the
> bill.  According to the description above, it does not.  Instead, it
> adds yet another set of semantics for overlaying on a text buffer.  It
> does not unify the existing sets (syntax highlighting, grinding, MMM
> for mixed-language files, maybe others like project membership), and
> it hasn't even been demonstrated that it *can*.

I feel the need to both agree and disagree.

I agree that CEDET is not a unifying force for *all* things a language 
mode may wish to do.

I had responded to a couple different assertions.  One was something 
David Kastrup wrote:

| The question is why the respective facilities are not part of the
| generic Emacs language support framework.  Support for every language
| has a completely disjoint set of features, keybindings, highlighting,
| and so on of wildly differing quality, design and usability.

My response was meant that CEDET is not a C/C++ only feature, but an 
infrastructure under which a mode can support code completion and the 
like.  CEDET doesn't tackle keybindings or font lock since Emacs already 
does that.  However, a suite of minor modes based on the infrastructure 
of CEDET can provide features (like code completion) which a mode author 
then does not need to implement themselves if a grammar is supplied. 
Thus, advanced major modes that try to solve some complex problems like 
code completion, compilation, or code generation could have a unified 
set of keybindings and behaviors for those features based on those minor 
modes.

Hopefully that is more clear.

A second thing I responded to was something Tom wrote:

| Since we don't have a killer feature which would attract new
| users like perfect code assist (context aware completion, instant
| display of documentation of elements, live indication of syntax
| errors, etc.) out of the box with near-zero configuration, we

to which this current thread branch originates.   The CEDET/Semantic 
tools can do these things such as code assist, display documentation 
(though you have to ask for it) and even live indication of some syntax 
errors.  (A feature that could use some help in CEDET.)

In theory, things like font-lock, or MMM mode features could work with 
the parsers CEDET/Semantic.  I've participated in discussions on these 
topics to try and solve them, but they are not solved, so I agree there 
is no quick fix there.

Eric