CEDET

'sgoericke' writes:
> being a cedet user since a few years i;ve a question about C++11: i have an
> enum with a lot (~3000) members. Because i moved to the new C++ "enum class"
> feature cedet seems to have a big problem with that. Parsing that enum takes a
> lot of time (about 50min) and with the result that no symbols are found.
>  
> I guess cedet/semantic doesn't support those constructs yet. However, instead
> of changing my source i would like to help that semantic can parse "enum class"
> correctly. However i need a hint where to start in the parser. Maybe somebody
> can tell me which file or function is responsible to parse enums in C++?

First you'd have to change the grammar to recognize the constructs you
want. The current enum parsing is this rule in typesimple:

  | ENUM opt-class opt-name enumparts
    (TYPE-TAG (car $3) $1 $4 nil
              :parent (car ,$2))

And the members of the enums are parsed in those two rules:

enumparts
  : semantic-list
    (EXPANDFULL $1 enumsubparts)
  ;

enumsubparts
  : symbol opt-assign
    (VARIABLE-TAG $1 "int" (car $2) :constant-flag t )
  | LBRACE
    ( nil )
  | RBRACE
    ( nil )
  | COMA
    ( nil )
  ;

It's hard to tell how hard it would be to add parsing for C++11 enum
classes. It doesn't look too bad, but with C++ you can never be sure.

-David

Ok,

i wouldn't say it was simple but got it working :-)

I wouldn't say it's bug-free and of course i don't know what 
side-effects this change produces. However it works (and also "normal" 
enums are still parsed properly).

I've attached a patch for "c.by". Also attached is an archive with 
simple test-source and a screenshot ;-)

Regards
Sven

On 10/29/2012 08:10 PM, David Engster wrote:
> 'sgoericke' writes:
>> being a cedet user since a few years i;ve a question about C++11: i have an
>> enum with a lot (~3000) members. Because i moved to the new C++ "enum class"
>> feature cedet seems to have a big problem with that. Parsing that enum takes a
>> lot of time (about 50min) and with the result that no symbols are found.
>>
>> I guess cedet/semantic doesn't support those constructs yet. However, instead
>> of changing my source i would like to help that semantic can parse "enum class"
>> correctly. However i need a hint where to start in the parser. Maybe somebody
>> can tell me which file or function is responsible to parse enums in C++?
>
> First you'd have to change the grammar to recognize the constructs you
> want. The current enum parsing is this rule in typesimple:
>
>    | ENUM opt-class opt-name enumparts
>      (TYPE-TAG (car $3) $1 $4 nil
>                :parent (car ,$2))
>
> And the members of the enums are parsed in those two rules:
>
> enumparts
>    : semantic-list
>      (EXPANDFULL $1 enumsubparts)
>    ;
>
> enumsubparts
>    : symbol opt-assign
>      (VARIABLE-TAG $1 "int" (car $2) :constant-flag t )
>    | LBRACE
>      ( nil )
>    | RBRACE
>      ( nil )
>    | COMA
>      ( nil )
>    ;
>
> It's hard to tell how hard it would be to add parsing for C++11 enum
> classes. It doesn't look too bad, but with C++ you can never be sure.
>
> -David

Aneesh Kumar K. V. writes:
> I guess this is due to partial merge of senator. We don't have
> senator-completion-menu-point-as-event

All of the Senator completion functions were never merged to Emacs,
hence the bug is that `semantic-ia-complete-symbol-menu' was
accidentally merged by me. I will remove that again from Emacs. The new
completion framework is the semantic/complete package, with
semantic-complete-analyze-inline(-idle) as the main frontend functions.

If you find those mechanisms to be lacking, an alternative is to use the
external auto-complete or company package (the latter is available
through ELPA).

-David

David Engster <deng@...> writes:

> Aneesh Kumar K. V. writes:
>> I guess this is due to partial merge of senator. We don't have
>> senator-completion-menu-point-as-event
>
> All of the Senator completion functions were never merged to Emacs,
> hence the bug is that `semantic-ia-complete-symbol-menu' was
> accidentally merged by me. I will remove that again from Emacs. The new
> completion framework is the semantic/complete package, with
> semantic-complete-analyze-inline(-idle) as the main frontend functions.
>
> If you find those mechanisms to be lacking, an alternative is to use the
> external auto-complete or company package (the latter is available
> through ELPA).

Removed emacs devel from the cc,

So I tried to use ac-source-semantic. Is there anything specific I need
to do to get this to work ? I just have this in my c-mode-common-hook
  (setq ac-sources '(ac-source-semantic))

Any other changes needed to get it working ?

-aneesh

Aneesh Kumar K. V. writes:
> So I tried to use ac-source-semantic. Is there anything specific I need
> to do to get this to work ? I just have this in my c-mode-common-hook
>   (setq ac-sources '(ac-source-semantic))
>
> Any other changes needed to get it working ?


Please read the manual.

http://cx4a.org/software/auto-complete/manual.html#Installation

-David

David Engster writes:

> Aneesh Kumar K. V. writes:
>> So I tried to use ac-source-semantic. Is there anything specific I need
>> to do to get this to work ? I just have this in my c-mode-common-hook
>> (setq ac-sources '(ac-source-semantic))
>> 
>> Any other changes needed to get it working ?


> Please read the manual.

> http://cx4a.org/software/auto-complete/manual.html#Installation

You might also be interested in this fork (awaiting merge into upstream):

    https://github.com/llvilanova/auto-complete

Relevant code is in the "semantic-completion-placeholders" branch.

If you have "yas/minor-mode" enabled in your buffer, completing a function will
insert placeholders for the arguments.


Lluis

-- 
 "And it's much the same thing with knowledge, for whenever you learn
 something new, the whole world becomes that much richer."
 -- The Princess of Pure Reason, as told by Norton Juster in The Phantom
 Tollbooth

Lluís <xscript@...> writes:

> David Engster writes:
>
>> Aneesh Kumar K. V. writes:
>>> So I tried to use ac-source-semantic. Is there anything specific I need
>>> to do to get this to work ? I just have this in my c-mode-common-hook
>>> (setq ac-sources '(ac-source-semantic))
>>> 
>>> Any other changes needed to get it working ?
>
>
>> Please read the manual.
>
>> http://cx4a.org/software/auto-complete/manual.html#Installation
>
> You might also be interested in this fork (awaiting merge into upstream):
>
>     https://github.com/llvilanova/auto-complete
>
> Relevant code is in the "semantic-completion-placeholders" branch.
>
> If you have "yas/minor-mode" enabled in your buffer, completing a function will
> insert placeholders for the arguments.
>

That made it really nice. I had to use a larger ac-delay value,
otherwise filling the args value and giving \t tried to auto complete
the args in a wrong way. Hopefully this will be part of 1.4 ? I also
have to use ac-semantic-raw in ac-sources

-aneesh

Aneesh Kumar K V writes:

> Lluís <xscript@...> writes:
[...]
>> You might also be interested in this fork (awaiting merge into upstream):
>> 
>> https://github.com/llvilanova/auto-complete
>> 
>> Relevant code is in the "semantic-completion-placeholders" branch.
>> 
>> If you have "yas/minor-mode" enabled in your buffer, completing a function will
>> insert placeholders for the arguments.
>> 

> That made it really nice. I had to use a larger ac-delay value,
> otherwise filling the args value and giving \t tried to auto complete
> the args in a wrong way. Hopefully this will be part of 1.4 ?

I'm not sure. All I know is the auto-complete people are planning to split the
whole project into multiple per source sub-projects, and that's when they're
planning to merge the placeholders code.


> I also have to use ac-semantic-raw in ac-sources

I didn't play much with the implementation of candidate selection, except for
making the symbol documentation available.

AFAIR, ac-semantic-raw will provide you symbols for the cases where you're not
completing "members" (see `ac-prefix-cc-member').


Lluis

-- 
 "And it's much the same thing with knowledge, for whenever you learn
 something new, the whole world becomes that much richer."
 -- The Princess of Pure Reason, as told by Norton Juster in The Phantom
 Tollbooth

Hello Oleg

What is a language that you're using? Do it have it's own build tool?

My idea is to create a base for java-based project that have some
common things, like project file in the root of project, classpath,
etc.
I plan to commit some Maven-related changes over weekend (although I'm
not sure that I'll have much time due surprisingly good weather ;-) or
maybe during next week. After Maven support I plan to add support for
Clojure's Leiningen, although we don't have parser for Clojure yet...

Simon Brown <lists@...>, c
On Thu, Oct 18, 2012 at 2:13 PM, Left Right <olegsivokon@...> wrote:
> Not sure if my comment belongs here, if not, I'll start a different thread.
>
> I'm right now reading on extending EDE and the basic concepts. I'm
> looking for a tool to provide project support for a language mode I'm
> trying to improve. I've tried first eproject from elpa, but I don't
> like it for many reasons not really important to mention here.
> I would be very interested to see an EDE Java project, because the
> language I'm working on has similar concepts in terms of how sources
> are arranged. Please, if it is possible to share it, Alex, it would be
> perfect! :)
>
> One particular thing that I'm not feeling content with EDE projects is
> the requirement of Project.ede file to reside in each subdirectory of
> the project tree. The typical for Java style of arrangement of the
> sources is to have deeply nested hierarchies of directories, many of
> them would appear empty or be created only because the codestyle guide
> says so. I.e. it is typical to have
> /src/java/com/enterprise/project/module/components/group/ui/impl/Button.java
> kinds of names :| Where most of this path would appear as not
> containing anything else but other directory trees.
> So, does it sound reasonable to only have one Project.ede file per
> actual project? Also, note that project sources aren't necessary under
> the same physical directory tree on the file system. In fact, it is
> common to  reference sources which don't share a common directory root
> under the same project.
> Perhaps, if I need to elaborate, the tools working with the sources in
> these kinds of projects assume this state of events to be typical and
> normally don't require sub-directories and files to be provided to
> them one by one, instead, they would expect to be given a reference to
> a directory tree, which contains the sources. Thus having Project.ede
> files in the "intermediate" directories would serve no purpose from
> the stand point of compilers / debuggers etc. Another impediment would
> be that other IDEs don't save project info in this way and sharing
> source code with people not using Emacs would make it problematical to
> maintain a consistent project structure.
>
> Or maybe have I goten a wrong understanding of it?
>
> Best.
>
> Oleg
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_sfd2d_oct
> _______________________________________________
> Cedet-devel mailing list
> Cedet-devel@...
> https://lists.sourceforge.net/lists/listinfo/cedet-devel



-- 
With best wishes,                    Alex Ott
http://alexott.net/
Twitter: alexott_en (English), alexott (Russian)
Skype: alex.ott

Hello Eric

I thought about caching to avoid re-parsing of Jars on each
invocation, although maybe it would work fast enough (although I see
differences in performance on mac & linux)

On Fri, Oct 19, 2012 at 4:58 AM, Eric M. Ludlam <ericludlam@...> wrote:
> If I remember, db-javap will extract data from your .jar file on a per class
> basis once, and remember it during the current session.  If you restart
> Emacs, it will have to reload everything again.  It was speedy, so I didn't
> setup any other kind of persistence.
>
> If you need to reset things, just set semanticdb-javap-directory-tracker to
> nil.
>
> Eric
>
>
> On 10/15/2012 05:32 AM, Alex Ott wrote:
>>
>> Hi again
>>
>> I wrote simple prototype on the top of proj-maven2, that fetches
>> classpath information from Maven, and use it. I was able to get
>> completions working for fairly complex Hadoop project. I think, that
>> I'll refactor common parts into separate class, and will inherit all
>> JVM-based projects from it.
>>
>> And I have following question - does semantic/db-javap.el will store
>> extracted information in cache or it will perform lookup in jar every
>> time I'll call completion?
>>
>> On Sun, Oct 14, 2012 at 1:20 PM, Alex Ott<alexott@...>  wrote:
>>>
>>> Hi all
>>>
>>> Last months I'm spending more time working on JVM-based projects
>>> (mostly Clojure, and Java). I see, that situation with Java in CEDET
>>> is improved, but it still lacks such things, as automatic classpath
>>> detection, etc.
>>> I'm thinking about restoring "obsolete" maven EDE project, but besides
>>> maven, I also use Leiningen for Clojure code, and I'm also thinking
>>> about support for other JVM-related build tools, like SBT for Scala,
>>> Graddle, etc. So, I want to ask developers on how it's better organize
>>> hierarchy for all these projects, so we'll avoid code duplication, and
>>> Java completion/SemanticDB would be able to obtain classpath
>>> information from any of these projects.
>>> I'm thinking about creating generic-java-project, that will have
>>> "classpath" slot inside, and it will be used by SemanticDB. And
>>> specific project types will fill this slot using project-specific
>>> tools, like 'lein classpath' for Leiningen, 'mvn
>>> dependency:build-classpath' for Maven, etc.
>>>
>>> What do you think?
>>>
>>> --
>>> With best wishes,                    Alex Ott
>>> http://alexott.net/
>>> Twitter: alexott_en (English), alexott (Russian)
>>> Skype: alex.ott
>>
>>
>>
>>
>



-- 
With best wishes,                    Alex Ott
http://alexott.net/
Twitter: alexott_en (English), alexott (Russian)
Skype: alex.ott

Hello Alex,

I'm working on HaXe, here's the language homepage http://haxe.org/ if
you are interested.

The language's compiler can use a special kind of files to build
projects, these are basically a set of compiler options (+ compiler
can forward calls to shell / perform multiple compiles in succession).
For a large range of projects this is sufficient.
What is different from C/C++ compilers is that there is no need to
specify each file that has to be compiled and linked into the program.
The compiler will analyse the dependencies on its own and will find
(or fail to and error) the required dependent sources.
The primary use of the language is the web projects, and a typical
output is not a single binary, it is more of a what would you put into
a WAR file if you were to deploy a Java application into a Java web
server. I.e. a project would most commonly consist of a sources to
compile a web applet, HTML pages (perhaps generated), maybe some
JavaScript code, maybe resources such as images or sound clips etc.

What I was hoping to store persistently per project is a bunch of
customizations such as: compile-command, compiler's location, standard
library location, contributors library location etc (it is possible to
have multiple copies on your PC and it is a common practice to have
more then one). And a few variables defined by the user (such as
author, version) or formatting settings (such as number of spaces in a
tab, number of tabs to indent, maybe even color theme). From the
coding point of view, All I have to know is the project's root
directory, the list of root source directories and the name of the
file which is the program's entry point. All of these I could infer
from the build script, but this doesn't scale well, because,
obviously, I can't add stuff like source formatting in there.

What I did like in EDE is the interface :) I.e. I like very much the
customization buffer. It looks like a lot of work to me. I'm not sure
EDE is the right tool for what I want, I just liked how it works for
other things :)

Best.

Oleg

Hi

I think, that you can start by looking to ede-cpp-root-project and
other simple ede-project types. I'm using ede-cpp-root-project for
similar tasks (https://github.com/alexott/emacs-configs/blob/master/rc/emacs-rc-cedet.el#L186)
- I only specify project's root, compilation command & other options,
and retrieve this information when I need to invoke compiler, setup
some variables in hook, etc.

You can look to my config starting with this line:
https://github.com/alexott/emacs-configs/blob/master/rc/emacs-rc-cedet.el#L109
- it automatically fetches compilation command from EDE, and executes
it

On Fri, Oct 19, 2012 at 11:42 AM, Left Right <olegsivokon@...> wrote:
> Hello Alex,
>
> I'm working on HaXe, here's the language homepage http://haxe.org/ if
> you are interested.
>
> The language's compiler can use a special kind of files to build
> projects, these are basically a set of compiler options (+ compiler
> can forward calls to shell / perform multiple compiles in succession).
> For a large range of projects this is sufficient.
> What is different from C/C++ compilers is that there is no need to
> specify each file that has to be compiled and linked into the program.
> The compiler will analyse the dependencies on its own and will find
> (or fail to and error) the required dependent sources.
> The primary use of the language is the web projects, and a typical
> output is not a single binary, it is more of a what would you put into
> a WAR file if you were to deploy a Java application into a Java web
> server. I.e. a project would most commonly consist of a sources to
> compile a web applet, HTML pages (perhaps generated), maybe some
> JavaScript code, maybe resources such as images or sound clips etc.
>
> What I was hoping to store persistently per project is a bunch of
> customizations such as: compile-command, compiler's location, standard
> library location, contributors library location etc (it is possible to
> have multiple copies on your PC and it is a common practice to have
> more then one). And a few variables defined by the user (such as
> author, version) or formatting settings (such as number of spaces in a
> tab, number of tabs to indent, maybe even color theme). From the
> coding point of view, All I have to know is the project's root
> directory, the list of root source directories and the name of the
> file which is the program's entry point. All of these I could infer
> from the build script, but this doesn't scale well, because,
> obviously, I can't add stuff like source formatting in there.
>
> What I did like in EDE is the interface :) I.e. I like very much the
> customization buffer. It looks like a lot of work to me. I'm not sure
> EDE is the right tool for what I want, I just liked how it works for
> other things :)
>
> Best.
>
> Oleg



-- 
With best wishes,                    Alex Ott
http://alexott.net/
Twitter: alexott_en (English), alexott (Russian)
Skype: alex.ott

Thanks for replies,

I was at first, too, thinking I'd go with the example of the SCons
project, but then I realized I would need to save some info into the
project file and things like central repository (on one's PC) don't
quite cut it because, ideally, I'd like the project to be portable
between different computers, (portable literally, as in you can copy
the project file with the sources into another computer and given
there's EDE + haxe-mode installed in their Emacs on the other
computer, then it should contain all the necessary info to start
working without additional setup.)

I'm looking into Android project sources now. Thanks for the hint!

Best.

Oleg

Hi all

I just committed initial version of EDE support for Maven projects.
I took old version of proj-maven2, and extended it with automatic
classpath calculation using Maven itself. Now I'm able to open .java
file, and corresponding project will be loaded, providing name
completions using semantic/db-javap

There is a lot of todo in my list, but I hope, that even in this form,
it will help java devs.
The next thing that I want to do, is to refactor proj-maven2 to use
common base class for all JVM projects, plus to add ability to
re-calculate classpath if pom.xml was changed.

If you have any suggestion, please, write me directly, or (it's
better) to write here...

On Mon, Oct 15, 2012 at 6:19 PM, Alex Ott <alexott@...> wrote:
> Here is example of working name completion, where classpath
> information is obtained automatically from Maven:
> https://plus.google.com/photos/106643849267592516873/albums/5799553437092351793/5799553435408628978
>
> On Mon, Oct 15, 2012 at 11:32 AM, Alex Ott <alexott@...> wrote:
>> Hi again
>>
>> I wrote simple prototype on the top of proj-maven2, that fetches
>> classpath information from Maven, and use it. I was able to get
>> completions working for fairly complex Hadoop project. I think, that
>> I'll refactor common parts into separate class, and will inherit all
>> JVM-based projects from it.
>>
>> And I have following question - does semantic/db-javap.el will store
>> extracted information in cache or it will perform lookup in jar every
>> time I'll call completion?
>>
>> On Sun, Oct 14, 2012 at 1:20 PM, Alex Ott <alexott@...> wrote:
>>> Hi all
>>>
>>> Last months I'm spending more time working on JVM-based projects
>>> (mostly Clojure, and Java). I see, that situation with Java in CEDET
>>> is improved, but it still lacks such things, as automatic classpath
>>> detection, etc.
>>> I'm thinking about restoring "obsolete" maven EDE project, but besides
>>> maven, I also use Leiningen for Clojure code, and I'm also thinking
>>> about support for other JVM-related build tools, like SBT for Scala,
>>> Graddle, etc. So, I want to ask developers on how it's better organize
>>> hierarchy for all these projects, so we'll avoid code duplication, and
>>> Java completion/SemanticDB would be able to obtain classpath
>>> information from any of these projects.
>>> I'm thinking about creating generic-java-project, that will have
>>> "classpath" slot inside, and it will be used by SemanticDB. And
>>> specific project types will fill this slot using project-specific
>>> tools, like 'lein classpath' for Leiningen, 'mvn
>>> dependency:build-classpath' for Maven, etc.
>>>
>>> What do you think?
>>>
>>> --
>>> With best wishes,                    Alex Ott
>>> http://alexott.net/
>>> Twitter: alexott_en (English), alexott (Russian)
>>> Skype: alex.ott
>>
>>
>>
>> --
>> With best wishes,                    Alex Ott
>> http://alexott.net/
>> Twitter: alexott_en (English), alexott (Russian)
>> Skype: alex.ott
>
>
>
> --
> With best wishes,                    Alex Ott
> http://alexott.net/
> Twitter: alexott_en (English), alexott (Russian)
> Skype: alex.ott



-- 
With best wishes,                    Alex Ott
http://alexott.net/
Twitter: alexott_en (English), alexott (Russian)
Skype: alex.ott

Hi Eric,

>From reading more and a little bit into the code itself I realized
that projects have to do with Speedbar (I didn't know that before, but
it looks cool). Where can I read more about the [possible] interaction
with Speedbar / displaying outlines in its buffer? (Outlines =
information extracted from TAGS).

Thanks.

Oleg

On 10/20/2012 10:11 AM, Left Right wrote:
> Hi Eric,
>
>> From reading more and a little bit into the code itself I realized
> that projects have to do with Speedbar (I didn't know that before, but
> it looks cool). Where can I read more about the [possible] interaction
> with Speedbar / displaying outlines in its buffer? (Outlines =
> information extracted from TAGS).

Hi,

   There is a speedbar manual for just using it, but there are many 
random tools with their own support.  Usually I just use the context 
menu in speedbar, and you can see some of the supported display types, 
and then you can pick your favorite.

   If you mean how to add your own speedbar support, doing it from 
scratch is a bit of a pain (I made bad API design choices IMO) but if 
you use eieio, and inherit from eieio-speedbar, then the rest is pretty 
easy.

   Lastly, ECB is another nifty tool if you are looking for file/tag 
displays.  You can find that at ecb.sf.net.

Eric

Hi Eric.

That's probably something I don't know about eLisp yet, what I was
asking is: ede-project extends some Speedbar class and some methods
that specialize on it read slots which are declared in the Speedbar
class it extends (keybinding for example). I just wanted to find a way
to see what slots there are, and, ideally accessors / readers and
such.

Perhaps, in the future, I'd like to customize Speedbar to group
entities from tags hierarchically / display type information for
variables and methods, but it's less important at the moment.

I've been using ECB for a while, but as per what I understood, it's
not been updated for a long time / gradually gets out of sync with the
CEDET code base, or am I wrong?

Best.

Oleg

Hi Eric.

Sorry, may not be related to the thread title, perhaps I need to start
a new thread.
I was wondering, could you give me a hint on what is the meaning of
the :file slot in ede-project-autoload? Where is it trying to search
for it / what is it expecting to find in that file?

https://code.google.com/p/haxe-mode/source/browse/trunk/ede/haxe.el#71

I have tried to replicate Android project with minimal changes, and
this place is something I can't figure out.

Thanks.

Oleg

I see I hadn't updated the doc for the :file since I first wrote EDE. 
I'll updated that.

EDE uses the :file slot to unique identify where the project is located 
in the directory hierarchy (AndroidManifest.xml).  For some project 
types, it is also the file that stores the configuration file (Project.ede)

It could be a file that exists in each subdirectory (like a Makefile) or 
it can exist only at the toplevel of your project (like 
AndroidManifest.xml).  The code that deals with that difference is 
related to the project autoloader which I usually keep at as the first 
few functions in any ede project code.  (The stuff that goes into the 
ede-project-autoload definition.)

Usually the project autoloader :proj-file, and :file will match in some 
way, mostly for sanity purposes.

Good Luck
Eric

On 10/21/2012 10:46 AM, Left Right wrote:
> Hi Eric.
>
> Sorry, may not be related to the thread title, perhaps I need to start
> a new thread.
> I was wondering, could you give me a hint on what is the meaning of
> the :file slot in ede-project-autoload? Where is it trying to search
> for it / what is it expecting to find in that file?
>
> https://code.google.com/p/haxe-mode/source/browse/trunk/ede/haxe.el#71
>
> I have tried to replicate Android project with minimal changes, and
> this place is something I can't figure out.
>
> Thanks.
>
> Oleg
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_sfd2d_oct
> _______________________________________________
> Cedet-devel mailing list
> Cedet-devel@...
> https://lists.sourceforge.net/lists/listinfo/cedet-devel
>