-----Original message----- > From:Joe Zhang <[email protected]> > Sent: Fri 02-Nov-2012 17:55 > To: [email protected] > Subject: Re: URL filtering: crawling time vs. indexing time > > I'm not sure I get it. Again, my problem is a very generic one: > I think i get it.
> - The patterns in regex-urlfitler.txt, howevery exotic they are, they > control ***which URLs to visit***. They also control which URL's NOT to visit. If you have a regex prepended with a + it will pass but prepend it with a - and it will be filtered out. You have to filter out URL's you don't want and whatever remains you can let through by another regex. > - Generally speaking, the set of ULRs to be indexed into solr is only a > ***subset*** of the above. > > We need a way to specify crawling filter (which is regex-urlfitler.txt) vs. > indexing filter, I think. > > On Fri, Nov 2, 2012 at 9:29 AM, Rémy Amouroux <[email protected]> wrote: > > > You have still several possibilities here : > > 1) find a way to seed the crawl with the URLs containing the links to the > > leaf pages (sometimes it is possible with a simple loop) > > 2) create regex for each step of the scenario going to the leaf page, in > > order to limit the crawl to necessary pages only. Use the $ sign at the end > > of your regexp to limit the match of regexp like http://([a-z0-9]*\.)* > > mysite.com. > > > > > > Le 2 nov. 2012 à 17:22, Joe Zhang <[email protected]> a écrit : > > > > > The problem is that, > > > > > > - if you write regex such as: +^http://([a-z0-9]*\.)*mysite.com, you'll > > end > > > up indexing all the pages on the way, not just the leaf pages. > > > - if you write specific regex for > > > http://www.mysite.com/level1pattern/level2pattern/pagepattern.html, and > > you > > > start crawling at mysite.com, you'll get zero results, as there is no > > match. > > > > > > On Fri, Nov 2, 2012 at 6:21 AM, Markus Jelsma < > > [email protected]>wrote: > > > > > >> -----Original message----- > > >>> From:Joe Zhang <[email protected]> > > >>> Sent: Fri 02-Nov-2012 10:04 > > >>> To: [email protected] > > >>> Subject: URL filtering: crawling time vs. indexing time > > >>> > > >>> I feel like this is a trivial question, but I just can't get my ahead > > >>> around it. > > >>> > > >>> I'm using nutch 1.5.1 and solr 3.6.1 together. Things work fine at the > > >>> rudimentary level. > > >>> > > >>> If my understanding is correct, the regex-es in > > >>> nutch/conf/regex-urlfilter.txt control the crawling behavior, ie., > > which > > >>> URLs to visit or not in the crawling process. > > >> > > >> Yes. > > >> > > >>> > > >>> On the other hand, it doesn't seem artificial for us to only want > > certain > > >>> pages to be indexed. I was hoping to write some regular expressions as > > >> well > > >>> in some config file, but I just can't find the right place. My hunch > > >> tells > > >>> me that such things should not require into-the-box coding. Can anybody > > >>> help? > > >> > > >> What exactly do you want? Add your custom regular expressions? The > > >> regex-urlfilter.txt is the place to write them to. > > >> > > >>> > > >>> Again, the scenario is really rather generic. Let's say we want to > > crawl > > >>> http://www.mysite.com. We can use the regex-urlfilter.txt to skip > > loops > > >> and > > >>> unncessary file types etc., but only expect to index pages with URLs > > >> like: > > >>> http://www.mysite.com/level1pattern/level2pattern/pagepattern.html. > > >> > > >> To do this you must simply make sure your regular expressions can do > > this. > > >> > > >>> > > >>> Am I too naive to expect zero Java coding in this case? > > >> > > >> No, you can achieve almost all kinds of exotic filtering with just the > > URL > > >> filters and the regular expressions. > > >> > > >> Cheers > > >>> > > >> > > > > >
