Does CEDET work?

Does CEDET work?

[Jordi Gutiérrez Hermoso]

There's a running joke in #emacs that goes like this:

<JordiGH> ,cedet
<fsbot> From memory, cedet is [0] Collection of Emacs Development
Environment Tools
<fsbot> [1] at http://cedet.sourceforge.net/,
<fsbot> [2] http://xtalk.msk.su/~ott/en/writings/emacs-devenv/EmacsCedet.html,
..[Type ,more]
<JordiGH> ,more
<fsbot> [3] EDE, Semantic, SRecode, Cogre, Speedbar, and EIEIO,
<fsbot> [4] I totally know somebody who knows somebody who got cedet
working. They said it was awesome.,
<fsbot> [5] merged into Emacs in trunk, ..[Type ,more]
<JordiGH> ,more
<fsbot> [6] theoretical package nobody has ever managed to use for a
practical goal

Indeed, I keep hearing of the person in [4]. Does CEDET work? I've
tried it a few times, didn't get anything workable within a few
minutes, stopped trying. If CEDET doesn't work, why it it still in
trunk? If it does work, why can't apparently nobody make it work?

- Jordi G. H.

Re: Does CEDET work?

[joakim]

Jordi Gutiérrez Hermoso <jordigh@octave.org> writes:

> There's a running joke in #emacs that goes like this:
>
> <JordiGH> ,cedet
> <fsbot> From memory, cedet is [0] Collection of Emacs Development
> Environment Tools
> <fsbot> [1] at http://cedet.sourceforge.net/,<fsbot> [2] http://xtalk.msk.su/~ott/en/writings/emacs-devenv/EmacsCedet.html,..[Type ,more]
> <JordiGH> ,more
> <fsbot> [3] EDE, Semantic, SRecode, Cogre, Speedbar, and EIEIO,
> <fsbot> [4] I totally know somebody who knows somebody who got cedet
> working. They said it was awesome.,
> <fsbot> [5] merged into Emacs in trunk, ..[Type ,more]
> <JordiGH> ,more
> <fsbot> [6] theoretical package nobody has ever managed to use for a
> practical goal
>
> Indeed, I keep hearing of the person in [4]. Does CEDET work? I've
> tried it a few times, didn't get anything workable within a few
> minutes, stopped trying. If CEDET doesn't work, why it it still in
> trunk? If it does work, why can't apparently nobody make it work?
>
> - Jordi G. H.

This type of humour is sometimes funny in #emacs, but does not translate
at all well to emacs-develop where people are trying to be productive.


-- 
Joakim Verona

Re: Does CEDET work?

[Leo]

On 2012-04-19 06:05 +0800, joakim@verona.se wrote:
> This type of humour is sometimes funny in #emacs, but does not translate
> at all well to emacs-develop where people are trying to be productive.

Indeed. But the question is genuine. I have been comparing two python
editing environments: PyCharm (commercial) and Emacs. And the former got
decent parser for everything remotely related to python: js, html, xml,
css, sass, sql, coffeescript etc. etc not to mention python specific
stuff. I can appreciate such an intelligent environment and the enormous
productivity it enables.

Leo

Re: Does CEDET work?

[Eric M. Ludlam]

On 04/18/2012 05:39 PM, Leo wrote:
> On 2012-04-19 06:05 +0800, joakim@verona.se wrote:
>> This type of humour is sometimes funny in #emacs, but does not translate
>> at all well to emacs-develop where people are trying to be productive.
>
> Indeed. But the question is genuine. I have been comparing two python
> editing environments: PyCharm (commercial) and Emacs. And the former got
> decent parser for everything remotely related to python: js, html, xml,
> css, sass, sql, coffeescript etc. etc not to mention python specific
> stuff. I can appreciate such an intelligent environment and the enormous
> productivity it enables.

CEDET suffers in that it is pretty good in a fairly narrow band, ie - 
the band that I've worked in, and a few areas other people have helped 
with.  As I wrote it, I wanted to create a framework to make sure new 
support for these tasks could be added for other languages/projects. 
Sort of the way you always use GUD to write a debugger, you can use 
CEDET to do project management, parsing, or code gen.

Anyway, I think folks pick up CEDET, see a reference to their favorite 
language, and the most complex feature (smart completion) doesn't work 
as they expected because they are outside the well supported band, and 
configuring it is too much of a challenge because they have never seen 
it work in areas where it does work well to even know what the reward is.

Python is a fine example.  There's a pretty good parser, but if it can't 
find system libraries, and if it doesn't understand a typical python 
project structure, then the smart completion is kind of lacking. 
Hooking in all the external tool support in really needs someone who 
loves the language to stick their nose in to complete the feature.

I recently picked up Android programming as a hobby to see what I could 
do, and immediately found the lack of direct support for Android 
projects and java system libraries meant very few features worked.  It 
took a little while to build up the support, teaching CEDET where 
directories are, and getting system libraries hooked so I can extract 
symbols from them, and tying it all into smart completion.  Now it will 
complete almost anything in Java with android libraries.  Nifty, but 
also took a couple days since I'm not an expert in either Android or 
Java.  Arduino support went a lot quicker since the language was C++, 
but their command line Make system was a bit baffling to me for a while.

Anyway, there is the beginnings of support for many languages, but a 
combination of developing project support and hooking in external tools 
to derive system symbol libraries is needed to really make the system 
work well.  As has been said many times, configuring CEDET can be a real 
challenge.  That's why tools that do work well often are very 
persnickety about how you lay out your project.  It is rare for a tool 
to try and "just work" in some arbitrary code blob the way CEDET does.

Eric

Re: Does CEDET work?

[Óscar Fuentes]

Leo <sdl.web@gmail.com> writes:

> Indeed. But the question is genuine. I have been comparing two python
> editing environments: PyCharm (commercial) and Emacs. And the former got
> decent parser for everything remotely related to python: js, html, xml,
> css, sass, sql, coffeescript etc. etc not to mention python specific
> stuff. I can appreciate such an intelligent environment and the enormous
> productivity it enables.

Mind your words. Some time ago I posted similar concerns here and
received so much flak that I wouldn't dare to repeat the experience.

Re: Does CEDET work?

[Andreas Röhler]

Am 19.04.2012 04:03, schrieb Óscar Fuentes:
> Leo<sdl.web@gmail.com>  writes:
>
>> Indeed. But the question is genuine. I have been comparing two python
>> editing environments: PyCharm (commercial) and Emacs. And the former got
>> decent parser for everything remotely related to python: js, html, xml,
>> css, sass, sql, coffeescript etc. etc not to mention python specific
>> stuff. I can appreciate such an intelligent environment and the enormous
>> productivity it enables.
>
> Mind your words. Some time ago I posted similar concerns here and
> received so much flak that I wouldn't dare to repeat the experience.
>
>
>

https://blueprints.launchpad.net/python-mode/+spec/productivity

Re: Does CEDET work?

[Tom Tromey]

Jordi> Indeed, I keep hearing of the person in [4]. Does CEDET work?

Yeah, it does.

Simplest way to try it, no configuring required:

1. Open a random .c file
2. M-x semantic-mode
3. 'C-c , j' then type the name of a symbol

The way this is better than imenu is that if you have opened multiple .c
files from the name directory, you can jump between tags in any of them.

It has other features too.

Ultimately for me it had some issue that caused me to turn it off.
I think it caused unpredictable pauses while I was editing.

Still, it is cool and I am looking forward to when the rough edges are
polished off.

Tom

Re: Does CEDET work?

[Les Harris]

Jordi Gutiérrez Hermoso <jordigh@octave.org> writes:

> Indeed, I keep hearing of the person in [4]. Does CEDET work?

I hope your question doesn't get buried by the tongue-in-cheek trappings
of your post.  CEDET does have a reputation for frustrating complexity
at times (although I do not believe it is as deserved as many would have it)

In my experience the answer is an emphatic yes, CEDET works.  Not only
that it works better now than it ever has before. 

Here is some 'proof': http://i.imgur.com/OGrq6.png

For the screenshot I opened up a random Emacs source file, went to a
random function, and tried to complete a structure.  In this case it was
glyph_matrix, and we can see that CEDET has provided a list of all the
members of the glyph_matrix struct.

There is a bit prettification going on there, I use AutoCompleteMode and
its AC-Semantic backend to do the overlay-based completion window but
the heavy lifting is all CEDET.

And lest you think that setup required extensive, painstaking
configuration I present to you the sum total of my CEDET specific
configuration:

;; Semantic
(global-semantic-idle-completions-mode t)
(global-semantic-decoration-mode t)
(global-semantic-highlight-func-mode t)
(global-semantic-show-unmatched-syntax-mode t)

;; CC-mode
(add-hook 'c-mode-hook '(lambda ()
        (setq ac-sources (append '(ac-source-semantic) ac-sources))
        (local-set-key (kbd "RET") 'newline-and-indent)
        (linum-mode t)
        (semantic-mode t)))

You should check it out yourself if you haven't in a while.

-- 
Do they only stand
By ignorance, is that their happy state,
The proof of their obedience and their faith?

Re: Does CEDET work?

[Jambunathan K]

Les Harris <lharris@gnome.org> writes:

> Jordi Gutiérrez Hermoso <jordigh@octave.org> writes:
>
>> Indeed, I keep hearing of the person in [4]. Does CEDET work?
>
> I hope your question doesn't get buried by the tongue-in-cheek trappings
> of your post.  CEDET does have a reputation for frustrating complexity
> at times (although I do not believe it is as deserved as many would have it)
>
> In my experience the answer is an emphatic yes, CEDET works.  Not only
> that it works better now than it ever has before.
>
> Here is some 'proof': http://i.imgur.com/OGrq6.png
>
> For the screenshot I opened up a random Emacs source file, went to a
> random function, and tried to complete a structure.  In this case it was
> glyph_matrix, and we can see that CEDET has provided a list of all the
> members of the glyph_matrix struct.
>
> There is a bit prettification going on there, I use AutoCompleteMode and
> its AC-Semantic backend to do the overlay-based completion window but
> the heavy lifting is all CEDET.

Could you please share your autocomplete configuration as well?  It
could serve as a good starting point for people like me interested in
exploring CEDET (and whatever else that goes well with it).

> And lest you think that setup required extensive, painstaking
> configuration I present to you the sum total of my CEDET specific
> configuration:
>
> ;; Semantic
> (global-semantic-idle-completions-mode t)
> (global-semantic-decoration-mode t)
> (global-semantic-highlight-func-mode t)
> (global-semantic-show-unmatched-syntax-mode t)
>
> ;; CC-mode
> (add-hook 'c-mode-hook '(lambda ()
>         (setq ac-sources (append '(ac-source-semantic) ac-sources))
>         (local-set-key (kbd "RET") 'newline-and-indent)
>         (linum-mode t)
>         (semantic-mode t)))
>
> You should check it out yourself if you haven't in a while.

--

Re: Does CEDET work?

[Les Harris]

Jambunathan K <kjambunathan@gmail.com> writes:

> Could you please share your autocomplete configuration as well?  It
> could serve as a good starting point for people like me interested in
> exploring CEDET (and whatever else that goes well with it).

Certainly, it's not much.

;; Autocomplete
(require 'auto-complete-config)
(add-to-list 'ac-dictionary-directories (expand-file-name
             "~/.emacs.d/elpa/auto-complete-1.4.20110207/dict"))
(setq ac-comphist-file (expand-file-name
             "~/.emacs.d/ac-comphist.dat"))
(ac-config-default)

That's my Autocomplete configuration.  It's available via ELPA or
EmacsWiki.

The AC-Semantic backend is handed in my c-mode-hook which I posted
previously.

Previous versions of CEDET (and Autocomplete mode) required much more
manual configuration. It all works more or less 'out of the box' now.

-- 
Do they only stand
By ignorance, is that their happy state,
The proof of their obedience and their faith?

Re: Does CEDET work?

[Jordi Gutiérrez Hermoso]

On 19 April 2012 01:22, Les Harris <lharris@gnome.org> wrote:
> Jambunathan K <kjambunathan@gmail.com> writes:
>
>> Could you please share your autocomplete configuration as well?  It
>> could serve as a good starting point for people like me interested in
>> exploring CEDET (and whatever else that goes well with it).
>
> Certainly, it's not much.
>
> ;; Autocomplete
> (require 'auto-complete-config)
> (add-to-list 'ac-dictionary-directories (expand-file-name
>             "~/.emacs.d/elpa/auto-complete-1.4.20110207/dict"))
> (setq ac-comphist-file (expand-file-name
>             "~/.emacs.d/ac-comphist.dat"))
> (ac-config-default)
>
> That's my Autocomplete configuration.  It's available via ELPA or
> EmacsWiki.
>
> The AC-Semantic backend is handed in my c-mode-hook which I posted
> previously.
>
> Previous versions of CEDET (and Autocomplete mode) required much more
> manual configuration. It all works more or less 'out of the box' now.

Thanks, added your thing to the wiki:

    http://www.emacswiki.org/emacs/CEDET_Quickstart

- Jordi G. H.

Re: Does CEDET work?

[Aneesh Kumar K.V]

Jordi Gutiérrez Hermoso <jordigh@octave.org> writes:

> On 19 April 2012 01:22, Les Harris <lharris@gnome.org> wrote:
>> Jambunathan K <kjambunathan@gmail.com> writes:
>>
>>> Could you please share your autocomplete configuration as well?  It
>>> could serve as a good starting point for people like me interested in
>>> exploring CEDET (and whatever else that goes well with it).
>>
>> Certainly, it's not much.
>>
>> ;; Autocomplete
>> (require 'auto-complete-config)
>> (add-to-list 'ac-dictionary-directories (expand-file-name
>>             "~/.emacs.d/elpa/auto-complete-1.4.20110207/dict"))
>> (setq ac-comphist-file (expand-file-name
>>             "~/.emacs.d/ac-comphist.dat"))
>> (ac-config-default)
>>
>> That's my Autocomplete configuration.  It's available via ELPA or
>> EmacsWiki.
>>
>> The AC-Semantic backend is handed in my c-mode-hook which I posted
>> previously.
>>
>> Previous versions of CEDET (and Autocomplete mode) required much more
>> manual configuration. It all works more or less 'out of the box' now.
>
> Thanks, added your thing to the wiki:
>
>     http://www.emacswiki.org/emacs/CEDET_Quickstart
>

Another nice option is to use 

  (semanticdb-enable-gnu-global-databases 'c-mode)


And use gnu global for searching for tags project wide, So you start
with generated gtag file and then use semantic to search for a
tag. Semantic will internally fallback to global to find the file and
then parse the file to find the tag.

-aneesh

Re: Does CEDET work?

[Nix]

On 19 Apr 2012, Les Harris said:
> And lest you think that setup required extensive, painstaking
> configuration I present to you the sum total of my CEDET specific
> configuration:
>
> ;; Semantic
> (global-semantic-idle-completions-mode t)
> (global-semantic-decoration-mode t)
> (global-semantic-highlight-func-mode t)
> (global-semantic-show-unmatched-syntax-mode t)

So I just tried this. Random headers (e.g. stdlib.h, but not stdio.h)
were promptly highlighted in the
semantic-decoration-on-unparsed-includes face; other headers that
required -I flags to be added to the include path (e.g. glib.h) got
highlighted in the semantic-decoration-on-unparsed-includes face.
The latter I could fix with semantic-customize-system-include-path,
but the former was harder to fix.

Digging about in the info files, I found this:

,----
| 2.2.4 Create System Databases
| -----------------------------
| 
| If your supported language stores the system libraries in readily
| available parsable source code, you can pre-generate database files for
| them once, which will be used over and over for tools such as
| summary-mode, or the analyzer.
| 
|  -- Command: semanticdb-create-ebrowse-database dir
|      Create an EBROWSE database for directory DIR.  The database file
|      is stored in ~/.semanticdb, or whichever directory is specified by
|      `semanticdb-default-system-save-directory'.
`----

Great! I thought. A shame that function doesn't exist, nor does any
plausible replacement for it. I have no idea when Semantic built its
cache for /usr/include, nor why stdio.h didn't get included, nor how to
force it to rebuild anything.


But even despite this, Semantic is wonderful, ever so much better than
it was years ago when I looked at it last. I'll definitely be using it
constantly in future.

One other mode not mentioned:

(global-semantic-mru-syntax-mode t)

lets you hit C-x B to go back to any previous function / tag that you've
edited, with the most-frequently-changed stuff at the top of the list.
*This* is a feature that I've never seen any other editor provide, and
is seriously useful.


One other feature not mentioned: C-x , G gives you a *very* nice menu
for the current tag, far nicer than that in the Emacs package provided
by e.g. GNU global (which C-x , G can interact with, and which is itself
better by miles than the standard tagging tools in my opinion).

-- 
NULL && (void)

Re: Does CEDET work?

[Philipp Haselwarter]

This is what most of the folks that see me using Emacs for C code (which
is arguably the language best supported by CEDET) want to have for
/their/ setup. Apart from a functioning CEDET, auto-complete is *very*
appealing visually and a for many (novice-) users a more attractive way
to be presented with context-completion than a completion-buffer
(actually the interface is provided by popup.el). It'd be nice to have
this in stock Emacs.

As Eric has pointed out himself, the entry barrier is quite high though
and support for other languages often requires difficult customizations
(if possible at all).

A better integration by default for some of the most-used languages
would increase Emacs's usefulness tremendously for the majority of
people writing code, which in my understanding is the main purpose of
Emacs. The framework is there, now it needs some love and attention from
the core people. A stronger endorsement might also lead a to better
integration with external tools that now write their own glue-code to
tie into Emacs (eg. rope, to stick with Eric's Python example). The same
is true for language modes.
Providing a unified entry point is important to make the features these
tools provide accessible.

Without raising the question what Emacs wants to be or can be, I think
this is an area that could use some more attention and dev-time.

-- 
Philipp Haselwarter

Re: Does CEDET work?

[Lennart Borgman]

[-- Attachment #1: Type: text/plain, Size: 1093 bytes --]

On Thu, Apr 26, 2012 at 20:14, Philipp Haselwarter
<philipp@haselwarter.org>wrote:

> This is what most of the folks that see me using Emacs for C code (which
> is arguably the language best supported by CEDET) want to have for
> /their/ setup. Apart from a functioning CEDET, auto-complete is *very*
> appealing visually and a for many (novice-) users a more attractive way
> to be presented with context-completion than a completion-buffer
> (actually the interface is provided by popup.el). It'd be nice to have
> this in stock Emacs.
>
> As Eric has pointed out himself, the entry barrier is quite high though
> and support for other languages often requires difficult customizations
> (if possible at all).
>
> A better integration by default for some of the most-used languages
> would increase Emacs's usefulness tremendously for the majority of
> people writing code, which in my understanding is the main purpose of
> Emacs.
>
> The CEDET framework is there, but the support for a language could be on
many levels. Is there somewhere a table over the support for different
languages?

[-- Attachment #2: Type: text/html, Size: 1454 bytes --]

Re: Does CEDET work?

[Philipp Haselwarter]

On Thu, Apr 26 2012 20:26 (@1335464766), Lennart Borgman wrote:

> On Thu, Apr 26, 2012 at 20:14, Philipp Haselwarter <philipp@haselwarter.org> wrote:
>
>     This is what most of the folks that see me using Emacs for C code (which
>     is arguably the language best supported by CEDET) want to have for
>     /their/ setup. Apart from a functioning CEDET, auto-complete is *very*
>     appealing visually and a for many (novice-) users a more attractive way
>     to be presented with context-completion than a completion-buffer
>     (actually the interface is provided by popup.el). It'd be nice to have
>     this in stock Emacs.
>    
>     As Eric has pointed out himself, the entry barrier is quite high though
>     and support for other languages often requires difficult customizations
>     (if possible at all).
>    
>     A better integration by default for some of the most-used languages
>     would increase Emacs's usefulness tremendously for the majority of
>     people writing code, which in my understanding is the main purpose of
>     Emacs.
>
> The CEDET framework is there, but the support for a language could be
> on many levels. Is there somewhere a table over the support for
> different languages?
>  


There is, at http://cedet.sourceforge.net/languagesupport.shtml

I just caught up on the cedet-ml and saw that the newtrunk branch got
merged (yay!).  This could be a Big Thing if it means that integration
and accessibility become an Emacs issue (as opposed to a
not-an-emacs-but-someone-other-package's-issue).


-- 
Philipp Haselwarter

Re: Does CEDET work?

[Lennart Borgman]

On Thu, Apr 26, 2012 at 21:13, Philipp Haselwarter
<philipp@haselwarter.org> wrote:
> >
> > The CEDET framework is there, but the support for a language could be
> > on many levels. Is there somewhere a table over the support for
> > different languages?
> >
>
>
> There is, at http://cedet.sourceforge.net/languagesupport.shtml

Thanks Philipp. Then people can start filling the gaps and expanding the list.

However the list is a bit incomplete. I think it also should mention
how the support is implemented since there are many ways to implement
it.

Re: Does CEDET work?

[Eric M. Ludlam]

On 04/26/2012 03:17 PM, Lennart Borgman wrote:
> On Thu, Apr 26, 2012 at 21:13, Philipp Haselwarter
> <philipp@haselwarter.org>  wrote:
>>>
>>> The CEDET framework is there, but the support for a language could be
>>> on many levels. Is there somewhere a table over the support for
>>> different languages?
>>>
>>
>>
>> There is, at http://cedet.sourceforge.net/languagesupport.shtml
>
> Thanks Philipp. Then people can start filling the gaps and expanding the list.
>
> However the list is a bit incomplete. I think it also should mention
> how the support is implemented since there are many ways to implement
> it.

You can get a high level view of how to add support at this page:

http://cedet.sourceforge.net/addlang.shtml

Or were you asking about how support was implemented for the existing 
language support listed in the table?

Eric

Re: Does CEDET work?

[Lennart Borgman]

On Fri, Apr 27, 2012 at 01:36, Eric M. Ludlam <ericludlam@gmail.com> wrote:
> On 04/26/2012 03:17 PM, Lennart Borgman wrote:
>>
>> On Thu, Apr 26, 2012 at 21:13, Philipp Haselwarter
>> <philipp@haselwarter.org>  wrote:
>>>>
>>>>
>>>> The CEDET framework is there, but the support for a language could be
>>>> on many levels. Is there somewhere a table over the support for
>>>> different languages?
>>>>
>>>
>>>
>>> There is, at http://cedet.sourceforge.net/languagesupport.shtml
>>
>>
>> Thanks Philipp. Then people can start filling the gaps and expanding the
>> list.
>>
>> However the list is a bit incomplete. I think it also should mention
>> how the support is implemented since there are many ways to implement
>> it.
>
>
> You can get a high level view of how to add support at this page:
>
> http://cedet.sourceforge.net/addlang.shtml
>
> Or were you asking about how support was implemented for the existing
> language support listed in the table?

Thanks Eric, yes I was asking about how support was implemented now. I
think knowing that could be great when searching for examples. (Right
now and for a considerable length of time I do not have time to try to
implement anything.)

Re: Does CEDET work?

[David Engster]

Philipp Haselwarter writes:
> This is what most of the folks that see me using Emacs for C code (which
> is arguably the language best supported by CEDET) want to have for
> /their/ setup. Apart from a functioning CEDET, auto-complete is *very*
> appealing visually and a for many (novice-) users a more attractive way
> to be presented with context-completion than a completion-buffer
> (actually the interface is provided by popup.el). It'd be nice to have
> this in stock Emacs.

I agree. It would be great to have a completion system based on textual
overlays. At least, the also excellent company-mode is in ELPA, but it's
currently unusable on Emacs24 with an activated header-line (yes, I
reported it).

CEDET has its own completion system with various display styles, but
they're all not as good as company-mode or auto-complete.

-David

Re: Does CEDET work?

[Stefan Monnier]

> A better integration by default for some of the most-used languages
> would increase Emacs's usefulness tremendously for the majority of
> people writing code, which in my understanding is the main purpose of
> Emacs.

100% agreement: the integration of CEDET in Emacs was one step in
that direction.  The new completion-at-point-functions is also a step in
that direction.  Integration of company-mode in ELPA is also a step in
that direction.  Hopefully auto-complete will be added to ELPA as well.
And hopefully those completion mechanism will merge (i.e. auto-complete
and company-mode should rely on completion-at-point-functions to get
their completion data, which will need enhancing
completion-at-point-functions along the way).

> The framework is there, now it needs some love and attention from
> the core people.

I'd love to do that, but am swamped, so help would be most welcome.


        Stefan