this is far closer to what we are looking for. Using Ant is an interesting idea, although it probably won't help us for the UI tool. But we could try to layer things so we could use them for both -- we want to get some more sophisticated index management anyway. The option to create the index at one place and use it somewhere else would be great -- during testing and demoing we ran into the problem that we wanted to demo on a Windows box but using a Unix filesystem mounted via SMB/Samba. Symlinks are no fun in this case :-( To work around this we need to develop some notion of a base URL, then we could easily mount an index created on one machine on another -- even if the underlying OS changes. To go Enterprise we would still need some security concept, which we probably won't do before someone is willing to pay for it :-) It might be better to go intranet for that one anyway -- we should be able to take it all to the Web.
Two differences between the Ant project and what we do right now:
- the Ant project doesn't have a notion of an explicit file filter. I think this is important if you want to extend the filter options to more than just extensions and if you want some UI to manage the filter mappings. BTW: does anyone know of a Java implementation for file(1) magic?
- the code creates Documents as return values. The reason we went away from this is that we want to use the same document handler with different index options. One of the core issues here is storing the body or not. I don't think there is any true answer for this one, so it should be configurable somehow. The two options I see are either returning a data object and then turning that into a Document somewhere else or passing some configuration object around. Both are not really nice, the first one needs to create an additional object all the time, while the second one puts quite some burder on the implementer of the document handler. Ideas on that one would be extremely welcome.
Two ideas we will probably pick up from this are:
- use Ant for creating indexes if we go larger than personal document retrieval
- use JTidy for HTML parsing (we missed that one and used Swing instead, which is no good)
So thanks again, that was quite helpful.
Peter
Erik Hatcher wrote:
If you are after a pure file system indexing abstraction, check out the 'ant' project in the sandbox. It's got a DocumentHandler abstraction allowing it to be a bit pluggable. Its not perfect, but it has worked for me for quite some time quite sufficiently.
Erik
On Monday, June 30, 2003, at 08:26 PM, Peter Becker wrote:
Clemens Marschner wrote:
There's an experimental webcrawler in the lucene-sandbox area calledI've done that by now -- my first problem was to identify LARM as the relevant project, but then things were reasonably easy to find.
larm-webcrawler (see
http://jakarta.apache.org/lucene/docs/lucene-sandbox/larm/ overview.html),
and a project on Sourceforge (http://larm.sf.net) that tries to leverage
this on a higher level. I want to encourage you to go on that side and read
through the specs in sourceforge's CVS.
It concludes pretty much everything that Andy wrote in his proposal, andIn many ways the project aims too high for us. We are interested only in the file system part and our time is limited. My hope was that someone would say there would be a basic framework somewhere where we can put our code, but due to the time limitations we will rather do our own thing. But this is maybe not as bad as it sounds since (a) our original plan was very close to what you describe in certain parts of the system, (b) we have read your documentation and (c) our code will be BSD-licensed.
more. The project only contains conceptual documents at this time, but if
you're willing to contribute actively, that's very appreciated.
The main ideas we have are:
- map file types to document processors
- use the java.io.FileFilter interface as base for the mappings
- the document processors will probably have a two method interface:
DocumentSummary processDocument(URL);
String getDisplayName();
- the DocSummary class will model the common attributes like author, title, text, etc. with a Properties object to be extensible. It's main purpose is to separate indexing concerns like stored/unstored and tokenized/untokenized from the document processors
- the display name will be used in the UI to create lists of FileFilter->DocumentProcessor mappings
- there will be some crawler code for the file system, but of course that is a lot easier
Many of these things will not extend straightaway into the web context, but I think the main work we will do will be in implementing the different DocumentProcessors. That part should be reusable. The mapping idea should be reusable, although FileFilter would have to be replaced with something more abstract, at least a URLFilter. My experience with Java networking is not good enough to judge the complexity of that.
We expect to have the relevant parts of this done next week. Code will be on Sourceforge (http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/toscanaj/docco/), it might be at least useful as inspiration :-) We are also looking into alternatives for parsing PDF and other formats. We have a lot of problems with PDFBox at the moment, and there might be other candidates (http://www.cs.berkeley.edu/~phelps/Multivalent/). And we are looking into the option to use the UDK for indexing (http://udk.openoffice.org/), although that most likely will complicate deployment and increase program size quite a bit. One of the problems we have is that we have some interesting test cases for the parsing tools, but we can't give them away and don't have the time to debug ourself. We have a file which causes PDFBox to get stuck without any feedback and an XLS file which causes POI to loop with funny messages for a long time until we run out of memory (with -mx500m). But that is something we have to talk to the other projects about.
The point of this waffle is: if you think some of our ideas are not as good as they should be or there are things that might affect reuse, please shout now :-) We start coding this right now.
Unfortunately I have to stop my efforts regarding LARM. Long story short: MyFair enough. I guess as professional developer you can never be completely free from considering IP issues.
future employer says it's too close to their business. But in contrast to
other open source projects, there's already lots of ideas in that document
and lots of code in the old crawler. If you wish to contribute, it's now up
to you.
Gr�sse, Peter
Clemens
----- Original Message ----- From: "Andrew C. Oliver" <[EMAIL PROTECTED]>
To: "Peter Becker" <[EMAIL PROTECTED]>
Cc: "Lucene Developers List" <[EMAIL PROTECTED]>
Sent: Friday, June 27, 2003 2:53 AM
Subject: Re: Lucene crawler plan
On 6/26/03 8:33 PM, "Peter Becker" <[EMAIL PROTECTED]> wrote:others
Hi Andrew,Yes I wrote it.
are you the Andy signing this:
http://jakarta.apache.org/lucene/docs/luceneplan.html? If no -- do you
know who wrote the page and could you forward this email? Thanks. BTW:
your website link on http://jakarta.apache.org/lucene/docs/whoweare.html
is dead.
The question is: is there some code already? If yes: can we get it? CanI never wrote any code, but there is code in lucene-contrib which realized
we join the effort? If no: what are things we should consider doing to
increase our chances that you guys accept our code in the end? We are
not really interested in maintaining the crawler bits and pieces, our
interest is in the visualization. We are happy to get something going as
part of our little demonstrator, but then we'd give it to you and hope
someone picks up maintenance.
most of what is in this document. I was going to write code, but someone
beat me to the punch and I was like "wow I have things I can do that
won't do for me" and moved on :-)
I'm cc'ing lucene developers list. You'll find plenty of folks interested
in working with you on this.
-Andy
Is this all an option anyway? It is ok to say no ;-)
Regards, Peter
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
