[CWB] info and documentation on the development of CWB's client-server interface

[Serge Heiden]

Hello Luca,

We have also developed a client-server architecture based on CQi and
CQPserver in Java, during a 3 years research project that is ending
this December (See http://textometrie.ens-lyon.fr/?lang=en)

We developed three different client side applications :
- Eclipse RCP based for Windows and Linux (Mac prototype only)
- Grails based for the web (See the current prototype at
http://txm.risc.cnrs.fr/txm : you need to access the 'Editions' link
and then the 'Outils' button to access the CWB based concordancer
form) [this application and prototype were developed by a Politecnico
di Torino student]
- GWT based for the web also, mimicking the RCP interface, but with
login based fine grained access control (prototype under beta testing
for the moment)
The server side combines CQPserver (for corpus query) and R (for
statistics model experiments)

Everything is licensed under GPL v3.0 and can be downloaded at
http://sourceforge.net/projects/textometrie
(see also the developers' web site at
http://sourceforge.net/apps/mediawiki/textometrie)

All the sources (Toolbox, RCP, Grails and GWT) are ready for coding in
the Eclipse IDE directly from SVN (See the developer's wiki for
instructions).

To have a preview of what final users can do with our applications
(based on CWB and R), you can read the 'Reference Manuals' here:
https://sourceforge.net/projects/textometrie/files/documentation

The Javadoc, and the Groovydoc for scripting the platform, is at
http://textometrie.sourceforge.net/javadoc/index.html
You will find the documentation of our Java implementation of the CQi
client side at
http://textometrie.sourceforge.net/javadoc/java/org/textometrie/searchengine/cqp/package-summary.html
The server side documentation you will find concerns the code we use
to control the CQPserver process from Java (start, stop).

It may be useful to you to know that we are not completely satisfied
by the present CQPserver implementation of the CQi API.
Most importantly:
- error diagnostics are not transmitted back by the protocol yet and
our clients can't report CQL syntax errors properly for example.
(search the CWB mailing list to find threads about this)
- Windows sockets transmission is TOO slow for the moment on some
architectures (we also had to develop a Windows version of CQPserver).
We planned to implement a flat Java CQi client/server API where the
CQPserver binary is dynamically linked to the client JVM and data is
exchanged through JNI to prevent communication lags. But nothing has
started on that for the moment.
We may also propose extensions to the CQi API in the near future.

Best,
Serge



Selon Luca Procopio le 30/11/2010 10:18:
> Hello,
> I'm a student of Turin working with prof. Evert,
> I would work on the development of  CWB's “client-server interface” 
> because I think that it's a very intresting job and I hope that the 
> final result will can be useful for whole CQP comunity.
>
> Prof. Evert wrote :
>
> *>Develop a good networking client-server interface, possibly based on 
> the *
> *>existing CQi specification and CQPserver implementation, which would 
> also *
> *>offer "virtual corpora" combined on the fly from multiple "physical" *
> *>subcorpora. While this incurs the overhead I have warned about 
> above, it *
> *>opens the possibility of distributing corpora over multiple servers 
> for *
> *>parallel processing and load balancing; for very complex queries 
> that are *
> *>not limited by disk access speed, it would also enable us to make 
> use of *
> *>multiple CPU cores on a single server.  Your tasks in this case 
> would be *
> *>to revise the CQi specs, rewrite CQPserver to optimise performance 
> (e.g. *
> *>making use of pipes or other communication tricks if client and server *
> *>run on the same machine), and then implement virtual corpora within *
> *>CQPserver.*
>
> There is someone else that already is working on the development of 
> the client - server interface?
> Does exist some documentation?
>
> Best regards
>
>
>
> _______________________________________________
> CWB mailing list
> CWB at sslmit.unibo.it
> http://devel.sslmit.unibo.it/mailman/listinfo/cwb
>    

-- 
Dr. Serge Heiden, slh at ens-lyon.fr, http://textometrie.ens-lsh.fr
ENS de Lyon/CNRS - ICAR UMR5191, Institut de Linguistique Française
15, parvis René Descartes 69342 Lyon BP7000 Cedex, tél. +33(0)622003883

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://liste.sslmit.unibo.it/pipermail/cwb/attachments/20101130/a008b0a6/attachment.htm