CEDET

Johann "Myrkraverk" Oskarsson <myrkraverk@...> writes:

> Among my other adventures with semantic, I'd be interested in using
> IBM DB2 as a backend for SemanticDB (or even SemanticDB itself?) --
> mostly to relieve my 32 meg PS2 (my #1 workstation) from cache
> handling and second to get some speed increase -- if possible.
>
> I'd just like some slight pointers in the right direction, would I be
> looking at semanticdb-file.el or something else perhaps?

If you look in the mailing list archives (either on the web, or via
news.gmane.org) you'll see a brief thread between me and Eric Ludlam,
where he describes a way to do it.  

You'll also see a more recent email from me where I describe something
that I've actually started, but which is done differently.

> I doubt I'll look into this very soon -- unless it's just a lot
> easier than I thought.

No, I doubt that it's easier than you thought.  For one thing, you'll
want some kind of Emacs interface to DB2 (I cheated and used
PostgreSQL, for which there is an interface, albeit a slow one).

[...]

Hi,

Sorta off-topic, but I don't seem to have gotten a copy of my post to
the list (I've gotten used to it in the past).  Is there some default
that has changed?  Can someone confirm that it's just a setting I need
to change?

And for the on-topic stuff:

Bruce Stephens writes:
 > Johann "Myrkraverk" Oskarsson <myrkraverk@...>
 > writes:
 > 
 > > Among my other adventures with semantic, I'd be interested in
 > > using IBM DB2 as a backend for SemanticDB (or even SemanticDB

[...]

 > If you look in the mailing list archives (either on the web, or via
 > news.gmane.org) you'll see a brief thread between me and Eric
 > Ludlam, where he describes a way to do it.
 > 
 > You'll also see a more recent email from me where I describe something
 > that I've actually started, but which is done differently.

K, I'll take a look

[...]

 > No, I doubt that it's easier than you thought.  For one thing,
 > you'll want some kind of Emacs interface to DB2 (I cheated and used
 > PostgreSQL, for which there is an interface, albeit a slow one).

Well, there is a sql-db2 interface, but I'm not sure I'd like to use
it -- I think it'll just slow me down (and I'd be using this to speed
things up).

Now since I've a habit of adding projects to my todo list faster than
I finish them off, I'm not sure *when* I'll start this but I'm
estimating somewhere between now and in 3 months; no hurry to respond
with detailed discussions -- and I'll be letting you know ;)

[...]

Johann

-- 
Emacs is not a text editor -- it's a way of life

>>> Bruce Stephens <bruce+cedet@...> seems to think that:
>Johann "Myrkraverk" Oskarsson <myrkraverk@...> writes:
>
>> Among my other adventures with semantic, I'd be interested in using
>> IBM DB2 as a backend for SemanticDB (or even SemanticDB itself?) --
>> mostly to relieve my 32 meg PS2 (my #1 workstation) from cache
>> handling and second to get some speed increase -- if possible.
>>
>> I'd just like some slight pointers in the right direction, would I be
>> looking at semanticdb-file.el or something else perhaps?
>
>If you look in the mailing list archives (either on the web, or via
>news.gmane.org) you'll see a brief thread between me and Eric Ludlam,
>where he describes a way to do it.

Here is a pointer:

https://sourceforge.net/mailarchive/forum.php?thread_id=3557240&forum_id=796

>You'll also see a more recent email from me where I describe something
>that I've actually started, but which is done differently.

http://sourceforge.net/mailarchive/forum.php?thread_id=3785516&forum_id=796

>> I doubt I'll look into this very soon -- unless it's just a lot
>> easier than I thought.
>
>No, I doubt that it's easier than you thought.  For one thing, you'll
>want some kind of Emacs interface to DB2 (I cheated and used
>PostgreSQL, for which there is an interface, albeit a slow one).
  [ ... ]

There are many ways it could be approached.  For example, a separate
application could accept STDIN as a series of commands and data to put
into the database, or do queries.  The more specific it is to
semantic the easier the lisp code could be.  Pushing the data across
this wire would not be very speedy.

Of course, writing such an application would prove challenging, as
well as the schema for the tables.  Communicating with Emacs may well
be the easy part.

To get reasonable speed and the advantage of off-machine storage of
tags, you would have to have speed Dependant actions store tags in
Emacs, and have time based events (based on usage) occasionally flush
tags out the to database.  You would need to balance your work flow
with when Emacs can spare the cycles.  Right now, databases are only
ever flushed when Emacs exits.  A more intermittent save might require
some additional framework in the core libraries.

Eric

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org

Johann "Myrkraverk" Oskarsson <myrkraverk@...> writes:

> Sorta off-topic, but I don't seem to have gotten a copy of my post
> to the list (I've gotten used to it in the past).  Is there some
> default that has changed?  Can someone confirm that it's just a
> setting I need to change?

I think it may be a sourceforge problem.  Sometimes I don't seem to
receive posts for a few days either.

[...]

> Well, there is a sql-db2 interface, but I'm not sure I'd like to use
> it -- I think it'll just slow me down (and I'd be using this to
> speed things up).

Right, sql-db2 doesn't look like quite the right sort of thing to
use---it looks like something you'd use as a human interface to DB2
rather than something that would allow an interface between Emacs-lisp
and DB2.  Whatever, it looks like this'll involve some work I haven't
(yet) had to do for PostgreSQL.

> Now since I've a habit of adding projects to my todo list faster
> than I finish them off, I'm not sure *when* I'll start this but I'm
> estimating somewhere between now and in 3 months; no hurry to
> respond with detailed discussions -- and I'll be letting you know ;)

[...]

Hi,

Thanks for the pointers.

Eric M. Ludlam writes:

[...]

 > There are many ways it could be approached.  For example, a
 > separate application could accept STDIN as a series of commands and
 > data to put into the database, or do queries.  The more specific it
 > is to semantic the easier the lisp code could be.  Pushing the data
 > across this wire would not be very speedy.

I had already thought of this aproach, but I'll look at the details
when I actually start.

 > Of course, writing such an application would prove challenging, as
 > well as the schema for the tables.  Communicating with Emacs may
 > well be the easy part.

Most likely ;)

 > To get reasonable speed and the advantage of off-machine storage of
 > tags, you would have to have speed Dependant actions store tags in
 > Emacs, and have time based events (based on usage) occasionally
 > flush tags out the to database.  You would need to balance your
 > work flow with when Emacs can spare the cycles.  Right now,
 > databases are only ever flushed when Emacs exits.  A more
 > intermittent save might require some additional framework in the
 > core libraries.

I'm itching to start this -- but until some of my current projects are
getting close to stable -- I'm sitting firmly on my hands.

A dream would be to have this mixed with a vc backend and store tag
tables with versions (at least the most recent ones).  But that will
probably just contine to be dream ;P

Have fun!

Johann

-- 
Emacs is not a text editor -- it's a way of life