#musicbrainz-devel

pom pom

nikki: ping

hi

hi :)

nikki: did you make any changes to toc.txt?

not neccesarily today.

(and someday I will learn how to spell that word).

That, or get an IRC client with spellchecking and English dictionary. :p

i don't like spellcheckers

I always feel I should be able to know the languages I use, and not need computers to help me.

(except my native language, I don't know that very well at all, which is a somewhat interesting situation)

you can always ignore the angry red line though

sure

and you don't even need to use the thing that offers suggestions...

luks, cycles I just ran through the releasegroup example http://wiki.musicbrainz.org/Next_Generation_Sch... expecting the release group to link to two different artists, (portishead and massive attack) via the artist credit table but it just links to one artist called 'Portishead & Massive Attack' , I don't get what artist credit is providing

where it links to one artist?

http://musicbrainz.org/artist/8644d878-2cb4-42e... <- no relationships

oh, you mean in the current (automatically converted) DB?

yep

are you saying the ngs db isn't quite right yet

as nikki said, there are no ARs so the script doesn't know it should convert it

you can't expect the DB to be automatically "fixed" :)

so should eventually it work as I imaginn

you can use e.g. http://test.musicbrainz.org/release/34e0d88d-94... for testing

or some other collaboration

fixing the DB will require a huge amount of manual work

the script does only the totally obvious fixes

thx, got it

luks: the psql text search seems broke :(

http://127.0.0.1:3000/search?type=label&que... go through the pages, and watch the "total hits" count

i have no idea why it does that though

I think it's because the hard search limit varies as you change the offset

yes, it's intentional

I did this to keep it fast enough even for queries with too many results

Hrm, then we're going to need something different for the lookup search then I think

Also, it makes the pagination completely useless, even on the main search results

I disagree that it's completely useless

ok

not completely

what you need most of the time is on the first few pages

and as you look through them, new search results are appearing

if you get 10 pages, you will almost never go straight to the 7th (for example)

right

because you can't possibly guess that your results is there

it would be better if it said "found over x results" then...

what about the lookup control? As the limit is about 10 results per page the chance of needing to go through pages is more likely

Could we use a minimum hard limit of 100?

I don't really see the problem, honestly

I guess the only time I hit a problem was when searching for something like "Smith" and by the time you get to page 10, clicking next page doesn't seem to do anything at all

but for smaller searches it doesn't matter, so I could leave it

it seems to behave fine for "Warner" as a label search too, which is probably as extreme as it would get

I mean, what is the result that you want to get?

if the number of results is going to change, can we get rid of the ">>" link and the "n pages" bit?

$hard_search_limit += $limit * 3;

this code makes sure that there are always at least three page visible

do you want to increase this?

nikki: I think it's unclear yet how will we use the psql search yet

ok

but yes, that link and "n pages" doesn't make sense there

it was just easier to re-use the existing system

I was just thrown when I searched for "records" in the label search and the pagination started to break when I got to "L"

I don't see a problem with the behaviour though, google works fine for me without saying exactly how many pages of results there are

it's not the count

hmm

by the time I got to "L" i couldn't seem to actually get past L

ok

but for small searches it's fine

I just imagined someone complaining about just this when we go for testing, but if it's expected behaviour, then we can leave it as is

is it the same on test? 'cause it works for me there o_O

oh, wait

it's not expected if the UI shows the number as the total count of matching result

but it should not do that

it doesn't matter much for artists or labels

I got to R and it started adding in results starting with B to M

yea, that sounds similar to what I got

but if you search for recording containing 'the', it helps a lot

(don't try it with the current direct search on the main server :))

heh

hmm... it was only that page that I saw though, the rest were fine, but once I got to Z, it started from A again

I'm not sure I understand it, where are you seeing it?

if you are seeing inconsistency in the results, it's because of $fuzzy_search_limit option

I thought 10000 is high enough for any search

oh, duuuh... they're in the middle of R because people stupidly used "Records, Whatever" as the sortname for "Whatever Records"

and it started repeating itself or something after something silly like 150 pages

but of course, nobody would seriously be checking 150 pages...

ok, we'll leave that

other question -- any one know how to test if a string is written using the latin alphabet?

in perl?

JavaScript

where do you need it?

yuck

People mentioned only showing the sort name if the first name is not latin

brian had some horrible regex for it

yea

i don't wanna bring that library in

see, in perl you'd just need something involving \p{Latin} :/

heh

you can make the server to return you the info

good point!

nikki: so how'd I do it in Perl :)

I'm not sure if it makes sense to decide whether or not to show the sort name based on this

at the moment I just always show the sort name

I would probably use /^[\p{Latin}\p{ASCII}\p{Common}\p{Inherited}]+$/ which would match anything which only contains ascii, latin or symbols

hmm... I don't think \p{ASCII} is needed, all the useful characters would be covered by the other ones...

Cool, works a treat

brian: ping?

hm, that didn't really need a question mark.

luks: do you have a second?

luks: Is the name field in artist_credit_name ever different to the name in artist that the artist_credit row links to

or is it just there for performance reasons

if I understood you correctly...

nikki:any idea when ?

when an artist is credited differently to the main name they use

Your'e right, http://pastebin.com/d1c8eb0fb

:)

so its a bit like an alias but specific to a release(group)

warp: brian emailed me yesterday saying he'll be afk for 24 hours ish

Cycles`: yesterday is 24 hours ago! ;)

nikki, ruaok: In the release group example at http://wiki.musicbrainz.org/Next_Generation_Sch... why does the first name-credit have a name element , but the second element tonly has one inside the artist element, is this right ?

no idea, I don't deal with the xml

that is not a good example, but technically correct.

hehe, its the only example I have

the name element there is for cases where the credit on the liner notes is different from the commonly recognized name.

in this case, imagine the credit having "Jethro Toe" and it should have "Jetho Tull".

this was one of the cases that we kept pounding on in the schema summit.

it will be used very infrequently

I see , so it shouldnt be shown if both name and artist are the same

correct

Ok, also the join phrase seems to be shown against the second term ( & Massive Attack), but in db its held against the first term (Portishead &) , be best if use same syntax (pre or post) in oth db and xml

actually we already discussed this and luks shot it down.

I haven't looked that the nuances, but luks says that the way it is now matches the schema.