#musicbrainz

hello hello HELLO!

hi

okay... I am curious to the benifit of having the albumjoin table. Couldn't we have albumid and squence fields put into the track table ?

the answer is probably something related to optimising stuff

hhmmm...

Knio-sleep is now known as Knio

Knio is now known as Knio-sleep

damn!

germany is winning over japan over release dates.

http://lmfao.org.uk/.files/ccodes.txt old ones, which were probably something like the 15th of december, http://lmfao.org.uk/.files/ccodes-new.txt ones as of the other day, http://lmfao.org.uk/.files/ccodes-difference.txt the difference between the two

cikkolata, if there's any way to search for these:

26 = France, Metropolitan

should all be changed to France

(i.e. FX -> FR)

StupidMop: you're right, at the moment the albumjoin table isn't really necessary because MB only uses 1:n relationships between album and track

the idea is to switch to a m:n relationship later where a track may be on multiple albums

hm, is it possible to query mb server for related artists?

cikkolata: http://www.audioscrobbler.com/group/Automoderat... care to chip in there at all with a good explanation of how MBIDs work?

Knio-sleep is now known as Knio-school

eek

DBD::Pg::st execute failed: ERROR: column "current_replication_sequence" is of type integer but expression is of type character varying

what did you do?

.

break it, obviously :P

during data import

that was only 10mins after the load started

but I did a RDF dump after that

on the redhat machine?

yup

and got 1.8Gb out

so you didn't import yesterday. hm.

let me have a look

well I think it was towards the end of the import

so maybe it doesn't matter

phone

dupuy: I had wondered whether that should be done (had contemplated a mailing list post), I'll fix them now.

back

towards the end of the import? were indexes already created?

going to run it again, but think so

phone. again. argh!

back again

changing line 173 in MBImport.pl might work

currently the DB driver seems to quote the $iReplicationSequence because it thinks it's a string

ok

I will wait for this import to finish and might try again in a sec

adding 0 makes it an integer. but if the variable contains something wrong we should at least get a warning

yup

wow, really does only take 10mins to import

now it worked?

nope, same failure

I will make your change now

also moving the postgres database to a different disk

shut it down first :-)

have done :)

strange. on my postgres 7.4.6 I can assign strings with a number in it to integer columns

I would have expected an error message

dupuy: I think all of the FX ones are now FR.

the user europop seems to have been responsible for 19 of them.

watch, and now you'll find that there were actually different versions released in parts of France and they were actually correct ^^;

heh

metropolitan france is covered by france though

so they're still right.

yes, but it would be amusing if there was say a paris release different from the rural France release

uncensored version for the big cities, censored for the small towns

I guess it would be amusing...

but I don't think that happened.

in any case, the FX country code has been retired (it's not longer used)

gah, university network lost internet access

wow this server is bloody fast

did a comeplete import and indexing in 6mins

what about the error?

still there :(

damn.

hmm, I have 26 FX->FR changes, so it seems like I got them all.

as I said, I can assing strings to integer columns in postgres 7.4.6

but it looks like the value is fine, most odd

yup

well, it is the last thing that the import script does

so you can at least use the DB without replication until we find out what's wrong

yeah

I am still in shock that it only took 6mins

I never heard of an import faster than 1 hour

took 4 hours on other server

well, it was all in the buffer cache after all

full import?

probably no disk io was required for reading

could be, yeah

cikkolata: almost. mbdump + derived

although it was unbziping it all again

hm. anyway, it's very fast

ah

damn.

I did mbdump on here in and half and a half (exactly)

errr

hour and a half

oh, I have been loading mbdump-moderation by accident

np, that's rather small

closed moderations are very large

yeah, not bothering with that for sure ;)

next is to see if JFS can cope with 387462 text files

what for??

going to create one file of RDF per artist and album

makes loading into triplestore a lot more efficient

but is going to bring your machine to its knees

if it was a sane machine, that is

and can also refer to an individual artist's RDF file

well we shall see !

bloke downstairs seemed to think it was a good idea !

would be tricky to keep triplestore up to date that way

we have the disc space... and it is his server... !

good luck :-)

argh, I really don't know what is wrong with this replication SQL problem

neither do I

7340 looks like an integer to me

what is it set to atm?