Semantic Web Interest Group IRC Chat Logs for 2004-12-19

00:12:07 <ows> Hi sh1mmer

00:59:06 <CaptSolo> sh1mmer: apparently no questions in the set theory :)

00:59:54 <sh1mmer> heh

01:00:10 <sh1mmer> talking to someone from #ai actually :)

02:28:13 <GreyLensman> sbp, u still working on the latest n3.n3 stuff? test suite et al.

02:31:48 <GreyLensman> @prefix x:y <http://www.bravias.org/test#> . # appears to be legal N3, but you can't ever actually use the prefix in a statement

02:46:22 <sbp> hi there GreyLensman. I am, though I don't have any control over n3.n3 so there's not much I can do to correct it unless I file a bug and correct it locally in the test suite

02:47:07 <sbp> we discussed the QName problem a little bit in the CWM/N3 meeting that was held here last Monday, but didn't get as far as splitting up prefix and localname

02:47:25 <sbp> which, it seems, is going to be necessary

02:47:29 <GreyLensman> yes

02:47:54 <GreyLensman> where is minutes of last meeting?

02:48:58 <sbp> from http://ilrt.org/discovery/chatlogs/swig/2004-12-13.html#T18-31-53 onwards

02:49:06 <GreyLensman> tanks

02:49:14 <sbp> I mention you at 18:39:44

02:49:43 <sbp> main discussion of the issue seems to be at http://ilrt.org/discovery/chatlogs/swig/2004-12-13.html#T19-14-23

02:50:19 <sbp> it takes a while for it to sink in (the others seem to think I mean that the localname can be zero length, which of course is fine)

03:52:28 <GreyLensman> interesting N3 meeting notes, lively discussion, lots of questions, few answers :(

04:05:10 <sbp> there was another one held a month beforehand, which was a bit better I think

13:56:52 <phurl> phurl is now known as md-afk

15:05:03 <balbinus_> balbinus_ is now known as balbinus

17:11:18 <sh1m> sh1m is now known as sh1mmer

18:06:06 <md-afk> md-afk is now known as PhUrl

18:19:17 <djav> hmm, any dublin core guru out there?

18:26:49 <PhUrl> PhUrl is now known as md-afk

18:32:38 <djav> Anyone know what is the best suited property to make a reference to the creator of a document, like a pointer to his FOAF file.

18:33:15 <Arnia> foaf:maker

18:33:29 <Arnia> Oh, to the file?

18:34:12 <djav> yes, at first I thought to use the dc:creator = "http://...."

18:34:29 <djav> but it seems dc:creator is about Literals

18:38:14 <d2m> djav: you could use dcterms:Creator

18:38:59 <kota> there's no dcterms:creator but dc:creator

18:39:44 <kota> and dc properties do not specify their range, so you can use it for resources

18:40:30 <djav> ok, nice. Thanks!

18:46:59 <djav> well, then if I have a FOAF file it's better for the rss:item in my rss feed to have a dc:creator="http://www.xxx.com/foaf.rdf" instead of dc:creator="Johann", isn't it?

18:47:55 <Arnia> dc:creator="http://www.xxx.com/foaf.rdf#me"?

18:48:14 <djav> yes

18:48:19 <Arnia> Point to the foaf:Person in the file I think...

18:48:27 <djav> sure

18:49:12 <kota> hmm, maybe you should <dc:creator><foaf:Person><foaf:name>blah</foaf:name><rdfs:seeAlso rdf:resource="uri"/></foaf:Person></dc:creator>

18:50:26 <djav> oh yes so I can combine best of both world

18:51:06 <Arnia> Although, I'd definitely include some IFP in that description too :)

18:51:26 <djav> IFP?

18:51:57 <Arnia> Inverse functional property... a property who's value uniquely identifies its subject

18:52:13 <Arnia> Like foaf:mbox, foaf:mbox_sha1sum, foaf:homepage etc

18:52:48 <djav> ok

18:55:50 <Arnia> How do I submit a paper to ESWC2005? The online submissions site doesn't appear to exist

19:46:35 <crschmidt> wow, building glibc with debug symbols is really helpful when doing backtraces...

19:49:35 <dajobe> hi crschmidt

19:49:39 <dajobe> can't dup your bug, not enough info

19:49:40 <crschmidt> dajobe: are you aware with any "mysql_free_result" bugs in rasqal .9.3 ?

19:49:48 <dajobe> just install the glibc-debug package (debian)

19:49:50 <crschmidt> dajobe: yeah, i know, it's really weird

19:50:00 <crschmidt> dajobe: that's what i did, got a useful backtrace now (rather than lots of ??)

19:50:04 <dajobe> rasqal doesn't do sql, it's redland

19:50:08 <crschmidt> hm

19:50:42 <crschmidt> backtrace leads back to my_no_flags_free, mysql_free_result, librdf_storage_mysql_find_statements_in_context_finished

19:52:31 <crschmidt> I'm not surprised you can't dupe, since it worked fine for me until just recently ;)

19:52:47 <dajobe> it sounds more like a mysql storage bug

19:53:46 <crschmidt> I'm just confused as to why it would be a problem, when it wasn't as short as a week ago; I haven't changed anything (other than recompiling glibc, now) that I can remember

19:54:03 <crschmidt> Would mail to the mailing list with a backtrace be okay, seeing if anyone has seen anything similar?

19:54:16 <dajobe> that would at least give a record of it

19:54:49 <dajobe> fwiw we're thinking of setting up a redland bug/issue tracker. Once we manage to pick one ;)

19:54:55 <crschmidt> heh

19:55:14 <crschmidt> I'm afraid of how many bugs i might open and close an hour later with "My machine is jsut horribly sucky, damnit"

19:55:37 <dajobe> but you have found bugs in the query lang stuff, which has been helpful

19:57:07 <dajobe> I wonder if Arnia is around. I compiled dbus-sharp the other day, didn't get any further though

19:58:30 * Arnia manifests

19:59:16 <Arnia> Sorry, was reading Slashdot comments on old articles about RDF :)

19:59:57 <dajobe> I re-found http://www.ilrt.bris.ac.uk/discovery/chatlogs/rdfig/2004-11-21.html#T15-33-12

20:00:21 <Arnia> Ah cool

20:00:52 <dajobe> I've not investigated that but

20:00:53 <dajobe> bit

20:01:04 <dajobe> however, in DAWG I've been writing an XML format for the result set

20:01:09 <dajobe> redland should emit that at some point

20:01:19 <dajobe> but the original NULL pointer errors shouldn't happen, clearly

20:01:49 <crschmidt> dajobe: is there an example of a C program for doing RDQL queries (something like a "hello world", so my test-case isn't in python)?

20:02:34 <dajobe> crschmidt: in the sources for rasqal - roqet.c and redland - rdfproc.c and at the end of librdf_query.c

20:02:45 <crschmidt> Thanks

20:07:29 <Arnia> dajobe: Cool. Thanks

20:16:11 * crschmidt sends email to dev-list, makes lunch

20:52:27 * mortenf replies

20:52:29 <phenny> mortenf: 14:41Z <crschmidt> ask mortenf if I'm supposed to have a file at http://crschmidt.net/blog/archives/author/crschmidt/foaf.rdf once I install his foaf plugin? Or if there's something more I need to do (same for skos)

20:53:13 <mortenf> yep and no, assuming default setup

20:59:15 <crschmidt> mortenf: i use a different permalink setup than most people; would that cause problems/

20:59:32 <mortenf> perhaps, but shouldn't

20:59:45 <crschmidt> how does it create the file?

20:59:54 <mortenf> it doesn't create a file

21:00:01 <mortenf> it's a uri

21:05:13 * crschmidt doesn't know how to get it to work, then :(

21:06:54 <dajobe> I can only find rasqal_free_triples_match at the line you mention in rasqal 0.9.2, that's 2 versions old

21:08:13 <dajobe> yes, you are still using redland 0.9.18

21:08:59 <crschmidt> I'm using rasqal 0.9.3, redland 0.9.18 (according to (redland|rasqal)-config --version)

21:09:09 <dajobe> the source lines say otherwise

21:09:58 <dajobe> one cause of failure would be compiling against one and linking/running against a different version

21:12:45 <crschmidt> mortenf: The latest CVS must use slightly different regex values; Changing RewriteRule ^archives/author/([^/]+)/?$ to RewriteRule ^archives/author/(.+)/?$ fixed it, thanks

21:12:55 <mortenf> ah, good

21:13:43 <mortenf> just for the record, it has only been tested with 1.2

21:13:52 <crschmidt> dajobe: Well, I haven't even bothered trying to compile/install/link/run anything new in the past several weeks, but I'll try to compile the latest release and install that to see where it gets me

21:21:26 * crschmidt types make and crosses his fingers.

21:23:19 <crschmidt> I'd also like ot say that after manually going through the build process for a program to get it to work for OS X, I recognize entirely too much of what's going on in the ./configure, make output

21:23:39 <mortenf> sorry to hear that

21:23:54 <crschmidt> heh

21:24:18 <kasei> i did the same thing with redland on osx

21:24:39 <kasei> ...too much time digging around the configure/make process

21:25:34 <dajobe> the ibook I have access to has returned from apple for the 4th fix, so I can try osx problems again

21:26:03 <crschmidt> kasei: yeah, i did it for a game I was working on (SDLRoads)

21:26:12 <kasei> the problem was using the binary installer of mysql. fink solved everything.

21:26:45 <crschmidt> dajobe: okay, got the latest release installed, but my rasqal_free_triples_match line number still doesn't match yours (it says rasqal_engine.c:460)

21:27:31 <crschmidt> (rasqal|redland)-config say 0.94, 0.9.19 respectively

21:27:49 <dajobe> I quoted rasqal 0.9.2 which matched your code

21:27:56 <dajobe> 460 is correct for rasqal 0.9.4

21:28:23 <kasei> oh, dajobe: i haven't tested thouroghly, but I've been running into a problem with the perl bindings of redland ...

21:28:38 <kasei> ... that suggest that RDF::Redland::Node->literal_value is returning a utf8 encoded string that isn't tagged as utf8

21:29:14 <crschmidt> dajobe: Okay. Should I send the new backtrace to the list, then? It looks pretty much the same other than line numbers

21:29:18 <dajobe> how would perl tag that kasei ?

21:29:24 <dajobe> crschmidt: yes please

21:29:42 <kasei> i don't know how the swig interface deals with this, but a perl scalar (SV) structure has a utf8 flag that needs to be set.

21:30:07 <kasei> the data i get back is properly encoded, but isn't marked as utf8, so i get weird encoding problems when dealing with non-ascii

21:30:36 <dajobe> I guess this is hitting a swig deficiency, it's the same for python and C#

21:31:19 <dajobe> redland always uses utf8 (except in a few corner cases) and all char* in/out are utf8

21:31:26 <kasei> a workaround that forces the utf8 flag to be set is to use $literal = Encode::decode('utf8', $literal).

21:31:28 <dajobe> but swig thinks they are unknown encoding

21:32:09 <dajobe> "At this time, SWIG does not provide any special support for Unicode or wide-character strings "

21:32:22 <kasei> hrm

21:34:03 <dajobe> . o O (if only there was a #swig IRC channel to ask this on)

21:34:12 <kasei> haha

21:34:41 <crschmidt> heh

21:37:25 <crschmidt> mortenf: I did the sos->results=0; thing, bt is sent as a reply

21:37:37 <mortenf> k

21:37:51 <crschmidt> oh, the glibc error changed too; *** glibc detected *** double free or corruption (out): 0x402b0a3c ***

21:40:42 <dajobe> valgrind's the thing you need here. It's an x86 app only.

21:41:06 <crschmidt> ooh, i found an environment variable that seems helpful

21:41:26 <dajobe> the osx memory debugging does seem helpful, but in my experience, has never been so.

21:41:29 <crschmidt> (Turns the glibc error into just a segmentation fault, with no funky phtread_kills)

21:41:55 <dajobe> that might help if you can get gdb on it; or run inside gdb

21:42:08 <crschmidt> Yep

21:42:37 <crschmidt> looks like the funkyness is coming from mallopt, mallopt, calloc, from librdf_storage_mysql_find_statements_with_options ?

21:42:48 * crschmidt is definitely not a good gdb user.

21:44:13 <crschmidt> and I'm on x86; all my hardcore hacking takes place here. The mac distracts me with all the pretty colors.

21:44:20 <dajobe> I think you've still got linking against different libraries.

21:44:36 <dajobe> because the rtm value turns from a value to 0x0 in calls

21:44:50 <dajobe> #13 0x4082a2e6 in rasqal_free_triples_match (rtm=0x80b7ab8)

21:44:50 <dajobe> at rasqal_engine.c:460

21:44:52 <dajobe> calls

21:44:55 <dajobe> #12 0x403528db in rasqal_redland_finish_triples_match (rtm=0x0,

21:44:56 <dajobe> user_data=0x80b7e30) at rdf_query_rasqal.c:407

21:45:06 <dajobe> the code is just one thing calling another

21:45:37 <dajobe> well then, run valgrind

21:45:57 <crschmidt> Installing it now

21:46:05 <dajobe> valgrind --tool=memcheck --leak-check=yes --num-callers=12 ..binary and args..

21:46:17 <dajobe> 12 is a random number, to give a bit of context - default is 4

21:46:46 <dajobe> if it's via python you'll get hundreds of errors in python, you'll have to ignore those

21:51:42 <crschmidt> Okay, I see lots of errors, (mostly "Invalid read of size 4") , and then at the end i see something that looks like a traceback with calloc, librdf_new_storage_from_factory, librdf_new_storage, _rwap_librdf_new_storage

21:52:07 <crschmidt> labeled: ==6304== 24 bytes in 1 blocks are definitely lost in loss record 8 of 35

21:52:27 <dajobe> if any of the invalid read lines have a librdf_ or rasqal call in them, can you email them?

21:52:45 <dajobe> we're not looking for lost memory here, but invalid reads - reading free()d memory

21:52:50 <crschmidt> okay

21:53:12 <dajobe> or just email the lot to me personally and I'll chop it down for redland-dev

21:58:48 <crschmidt> dajobe: how can I make sure I'm linking against the right libraries? Would `make uninstall` be enough?

21:59:06 <dajobe> I'd hope so

22:02:17 <crschmidt> Okay then, doing make uninstall, make clean, then another make; hopefully that'll link against the right libs

22:03:05 <dajobe> your last backtrace looked like the right libs

22:03:19 <dajobe> but only the valgrind output will tell you where the two frees were

22:05:09 <dajobe> i'll look at that tomorrow. too late for me today

22:05:15 <crschmidt> thanks

22:05:37 <crschmidt> do i need to install a new version of the bindings for 0.9.19?

22:05:54 <dajobe> it would be a good idea, otherwise I don't know what you have

22:06:00 <crschmidt> okay.

22:07:31 <mortenf> crschmidt, is this also with the zeroed instruction added?

22:10:46 <dajobe> looks to me that librdf_storage_mysql_find_statements_in_context_finished can be called multiple times

22:10:54 <dajobe> in rdf_storage_mysql.c

22:11:05 <mortenf> yeah

22:11:16 <dajobe> it's a destructor

22:11:27 <dajobe> so only safe to call there, or in the constructor if it partially fails

22:11:27 <mortenf> well, with the suggested fix (zeroing the result pointer)

22:12:02 <mortenf> hmm, not sure i understand what you mean now

22:12:40 <dajobe> it shouldn't be called inside your code

22:12:55 <dajobe> "return 1" will signify to rdf_stream_next that the stream is ended

22:13:06 <dajobe> and later on, it'll destroy the stream, calling librdf_storage_mysql_find_statements_in_context_finished

22:13:16 <dajobe> it guarantees to call it once only

22:13:37 <mortenf> hmm, ok, but i thought "better safe than sorry", and tried to make it recallable

22:14:15 <dajobe> you could try to do that but you can't have anywhere to store that state, since you free() every alloced resource :)

22:14:38 <dajobe> after LIBRDF_FREE(librdf_storage_mysql_sos_context, sos);

22:14:39 <mortenf> good point

22:14:45 <dajobe> there is no memory to store that it's been freed already

22:15:46 <mortenf> ok, problem located

22:16:41 <mortenf> however, the calls in librdf_storage_mysql_find_statements_with_options must stay, as they are executed before the stream is created

22:16:51 <dajobe> that's correct

22:16:55 <crschmidt> mortenf: everything since my email reply to you has been with the sos->results=0; thing in there

22:16:57 <dajobe> as long as return NULL follows

22:17:01 <mortenf> yep

22:17:52 <mortenf> i'll mail the list, do you want a patch or rather fix it yourself?

22:18:28 <dajobe> I've made an untested patch just by editing things

22:18:35 <mortenf> i wonder why i haven't run into this before

22:18:36 <mortenf> great

22:18:36 <dajobe> but it's for 0.9.18 as released, not current sources

22:18:44 <mortenf> ah

22:19:08 <mortenf> i doubt i'm up to date, btw...

22:19:23 <dajobe> (0.9.19 as released)

22:20:24 <crschmidt> Random curiosity: Are there any other RD Parsers that redland can use (other than raptor)?

22:22:14 * Arnia tries to figure out how to use Raptor independent of the rest of Redland in C#

22:22:23 <dajobe> crschmidt: no

22:22:46 <mortenf> dajobe, could you fix it in cvs as well, i'm behind on versions here?

22:23:01 <dajobe> yeah, I'll just email this out first to see if it *is* a fix

22:23:11 <mortenf> heh, ok

22:24:24 <mortenf> heh, just got your mail as i sent mine...

22:25:09 <dajobe> and you got the channel right...

22:25:13 <mortenf> :)

22:27:05 <dajobe> hmm, Bristol is colder than Helsinki today

22:27:41 <dajobe> got a DAWG meeting in Jan in HEL

22:28:01 <mortenf> stopping over in cph? there's a beer or two waiting for you...

22:28:18 <dajobe> heh, no - via AMS

22:28:38 <mortenf> ah, the dope thing ;)

22:28:50 <Arnia> Wycombe is pretty cold today too :(

22:28:58 <dajobe> no, the fly from BRS thing

22:29:14 * dajobe hates all london airports

22:29:32 <Arnia> Heh... I live 7 miles from Heathrow. I hear you on hating them :)

22:29:43 <Arnia> Mind you, I hate London too :)

22:30:11 <crschmidt> yay!

22:30:12 <crschmidt> it works!

22:30:23 <dajobe> great!

22:30:36 <mortenf> great, and sorry for the wasted time!

22:30:44 <crschmidt> pfft, it's sunday :)

22:30:47 <crschmidt> I'm glad I could help!

22:30:54 <crschmidt> I have no idea why I didn't run into the problem before...

22:30:56 <dajobe> hardly to blame, it's the undocumented redland internals

22:30:58 <dajobe> yeah

22:31:36 <mortenf> well, it's only a problem in case of an error, so perhaps your store is wasted somehow?

22:32:07 <crschmidt> hm

22:33:26 <mortenf> most likely cause: a node that doesn't exist, but is referenced from the statements table

22:34:05 * dajobe wanders off to watch a stupid film

22:34:16 <crschmidt> mortenf: hm, you have a script to clean those up, right?

22:34:36 <mortenf> well, not clean up that way, only the other way around

22:34:41 <mortenf> (unused nodes)

22:34:43 <crschmidt> ah

22:35:01 * mortenf checks som sql for locating missing nodes

22:37:03 <dajobe> C:fixed

22:37:03 <dc_swig> Added comment C5.

22:37:10 <mortenf> hmm, could be a context thing, you started with then skipped them right?

22:38:00 <mortenf> what do you get from "select count(distinct Context) from Statements;"?

22:38:48 <crschmidt> 244

22:39:06 <crschmidt> I've got some nodes with contexts; mostly from spidering, where i assign contexts to only hit URIs once.

22:39:18 <mortenf> what about "select min(Context) from Statements;"?

22:39:29 <crschmidt> 0

22:39:31 <mortenf> ah

22:39:42 <crschmidt> bad?

22:39:46 <mortenf> yeah...

22:40:01 <mortenf> well...

22:40:21 <mortenf> no, that should be ok

22:40:21 <crschmidt> I apparently have 788980 with Context 0

22:44:46 <mortenf> try "select count(*) from Statements left join Bnodes as SuB on Subject=SuB.ID left join Resources as SuR on Subject=SuR.ID left join Resources as PrR on Predicate=PrR.ID left join Bnodes as ObB on Object=ObB.ID left join Resources as ObR on Object=ObR.ID left join Literals as ObL on Object=ObL.ID where (SuB.Name is

22:44:46 <mortenf> null and SuR.URI is null) or PrR.URI is null or (ObB.Name is null and ObR.URI is null and ObL.Value is null);"

22:45:09 <mortenf> if you get results from that (might take a while), something's missing

22:45:57 <crschmidt> Running

22:46:31 <mortenf> here too, about 90+% cpu...

22:46:48 <mortenf> done: 0

22:47:06 <crschmidt> As usual, I'm disk bound

22:47:13 <mortenf> (from 1.2M triples)

22:47:19 <mortenf> :/

22:47:32 <crschmidt> 70% CPU, 528K/s from the disk

22:48:03 <mortenf> what do you use for disk measuring?

22:48:31 <crschmidt> gkrellm

22:48:36 <crschmidt> returned 25148.

22:48:39 <crschmidt> :/

22:48:47 <mortenf> oops

22:49:07 <mortenf> and that query doesn't include missing contexts, so yep, that's the origin of the error

22:49:48 <mortenf> of course, if you were using contexts, you'd know which resources to re-GET ;)

22:49:54 <crschmidt> So what does that mean? Select all Statements where Subject, Predicate, or Object is null?

22:50:14 <mortenf> yeah

22:50:35 <crschmidt> So, none of those will ever return correctly, right?

22:50:50 <mortenf> erh?

22:51:02 <mortenf> "those"?

22:51:09 <crschmidt> Those 25148 statements

22:51:11 <mortenf> right

22:51:21 <crschmidt> So there's no reason I shouldn't just delete them, right?

22:51:42 <crschmidt> s/select count(*)/DELETE/ ?

22:52:13 <mortenf> hmm, if you can, not sure if it works that way

22:53:15 <crschmidt> hm, apparently not :)

22:53:41 <crschmidt> i suppose that makes sense

22:54:07 <mortenf> well, i think you can in the recent versions

22:54:55 <crschmidt> i'm using 4.0.21 or something like that; maybe only in 4.1

22:55:35 * crschmidt is just happy to have julie working again

22:56:38 <Arnia> Aww. That's cute :)

22:57:12 * crschmidt goes to figure out how to add sparql support, now that he's using the latest code (hooray!)

22:57:55 <Arnia> crschmidt: Will you be using the new interpolation stuff from py2.4?

22:59:09 <crschmidt> [3 crschmidt@peanut ~]$ python -V

22:59:09 <crschmidt> Python 2.3.4

22:59:13 <crschmidt> ;)

23:00:34 <crschmidt> hm, does sparql not define anything like a USING after the where clause?

23:01:16 <Arnia> PREFIX?

23:01:24 <Arnia> bparsia: Hey

23:03:28 <crschmidt> Arnia: so far as I can tell, PREFIX has to be before the SELECT

23:03:50 <Arnia> ok

23:04:07 <crschmidt> [1] Query ::= PrefixDecl* ReportFormat PrefixDecl* FromClause? WhereClause?

23:09:05 <md-afk> md-afk is now known as PhUrl

23:13:41 <bparsia> Morning

23:13:51 <Talliesin> Evening.

23:14:14 <crschmidt> Howdy!

23:32:57 <KjetilK> crschmidt: julie is missed on irc.perl.org again... :-)

23:35:41 <crschmidt> KjetilK: yeah, and oftc, too

23:36:13 <KjetilK> any chance of getting her back?

23:37:02 <crschmidt> yep,

23:37:24 <crschmidt> just copying namespace definitions and command storage between the networks

23:38:02 <KjetilK> oh, she has some new code that requires that?

23:38:40 <crschmidt> nah, i just never sync the command lists between the networks like i should

23:39:12 <KjetilK> ok

23:39:16 <KjetilK> thanks a lot!

23:39:44 <crschmidt> np

23:48:42 <crschmidt> yay, fixed my perl bindings query problem too!

23:48:52 <crschmidt> I think today can be considered a good day.