Re: struggling with running GIFT...
Henning Müller <[email protected]> Tue, 29 Jul 2008 17:10:42 +0200
| Newsgroups | gmane.comp.gnu.gift.general |
|---|---|
| Organization | HUG |
| Message-ID | <[email protected]> |
Hi, indeed, the collection ID only needs to be changed in the config file. There should be no other dependencies. Cheers, Henning Niles Johnson wrote: > Hi Neville, > > I haven't re-checked all the attachments you sent, but as far as I > remember, the output has included complaints about the collection ID > number each time . . . > > In your 'gift-config.mrml', note the line beginning with: > > <collection collection-id="c-38-27-1-27-6-108-0-208-0" > collection-name="imgs".. > > I have a collection with collection-name="Full-Image-Collection" and > collection-id="c-01"; in my 'gift-config.mrml', and the algorithms > reference this collection ID. This is probably why, in the output you > sent, you see the error about collection c-01 not being described in > your config file. > > So I'd suggest you update your 'gift-config.mrml' in two ways: > > First, Henning explained to me that the "Classical IDF" algorithm does > not perform as well as the "Separate Normalisation" algorithm, and his > 'gift-config.mrml' does not include either the "PERL link" or the > "Classical IDF" algorithms. Following his recommendation, I deleted > both of these from my 'gift-config.mrml' and am having better > performance from the GIFT--I'd encourage you to do the same :) > > To do this, just delete everything between <algorithm ... > and > </algorithm>. Be warned that XML doesn't parse nested comments like you > might expect, so even though I commented out the "PERL link" algorithm > by adding '<!--' at the beginning, and '-->' at the end, this won't work > for the "Classical IDF" algorithm, since there are comments within the > algorithm block. > > > Second, check the line beginning with: > > <algorithm algorithm-id="adefault" algorithm-type="adefault" > algorithm-name="Separate Normalisation" collection-id= > > to be sure that the collection-id matches the one listed in your > collection reference. The GIFT generates a collection ID number every > time it indexes a new collection, but as far as I can tell, you don't > have to use the number it generates. > > I can't remember if you need to update the collection ID in any of the > other GIFT files, but maybe you can just try it with these changes and > see how it works? Hopefully the new output can help us identify any > other problems . . . > > best, > Niles > > p.s. I think it's close to working for you, so hang in there! If you > still have trouble, you might also check to see that your GIFT client is > referencing the correct collection ID's, while we're at it. > > > > > On Sun, 27 Jul 2008, [email protected] wrote: > >> Hello all, >> >> Thanks for all the help. >> >> So i downloaded a moonsock client, compared these output files as much >> as i >> could (really gets monotonous after awhile... :) ) and i found some major >> differences the gift server output. So this is the output right after i >> start the server and load the opening page of the moonsock client. [ >> my_queryout is my output and your_queryout is your's..., i have only >> pasted >> where i thought the difference in output starts ]. >> >> The opening page doesnt load for awhile and after i click on stop it >> appears >> SOMETIMES with the CORRECT collection names and algorithms. However, >> at this >> point if i click on get random images nothing happens... (i know, its >> quite >> wierd)... >> >> This seems to be an error after which my output and your output seems >> to be >> going two different ways so i suspect there;s something wrong going on >> here... >> *** >> "CAccessorAdminCollection.pre-cc:249:throwing AnVENotFound occured: I was >> looking for a proxy with the ID: mIDToProxy _c-01_ . This probably means >> that the collection with the ID c-01 is not described in your config >> file." >> *** >> >> I overwrote your gift-config.mrml, just added one of my collections >> there. I >> am also attaching it in this email. >> >> Any idea whatsoever? >> >> Thanks >> Neville >> > > > _______________________________________________ > help-GIFT mailing list > [email protected] > http://lists.gnu.org/mailman/listinfo/help-gift > -- --------------------------------------- Henning Müller (PhD) University & Hospitals of Geneva Service of Medical Informatics 24, rue Micheli-du-Crest CH-1211 Geneva 14, Switzerland Tel +41 22 372-6175 Fax +41 22 372-8680 email [email protected]