#musicbrainz-devel

one important reason to not using pg_dump is that you need to export and update the database at the same time

otherwise you can't handle replication packets properly

the acoustid format for full export is almost identical to MB

but the replication format is different

Leo_Verto: git diff $(initial commit)

* Michael Wiencek: Make sure Entity::PartialDate sorts same years the same

* Michael Wiencek: Events in a series should be ordered by date, then name (MBS-7557, MBS-7987)

* Michael Wiencek: Release groups in a series should be ordered by first release date, then name (MBS-7557)

* nikki: Fall back to more languages before giving up and picking randomly

* Michael Wiencek: Do sorting from MBS-7557 on the client too

Michael Wiencek: Remove test Wikipedia link type from orderable_link_type

Yippee, build fixed!

Michael Wiencek: Correctly validate date ranges between fuzzy dates

Michael Wiencek: Use Date::Calc to simplify is_date_range_valid logic

lol

maybe not then I guess XD

hm, well seems unrelated to that, since it fails on hobbes if I revert that too. but need to sleep now so I guess we'll find out tomorrow

Freso, okay, if you have two branches with a common ancestor and no rebasing or merging, you should be able to use git diff otherbranch...

Leo_Verto: Not two branches. One branch.

Hm

Leo_Verto: One branch, no branching off. I want to compare the repository (or a single file in it) from the initial commit to the current situation (or any given point in between).

git rev-list --max-parents=0 HEAD gets the first commit

so git diff $(git rev-list --max-parents=0 HEAD) HEAD should work

I can also do some `git log` magic.

git log apparently shouldn't be used for scripting

This is not for a script, or a long-winded solution like that wouldn't be a problem.

mhm, "git diff $(git rev-list --max-parents=0 HEAD) HEAD" seems to work though

http://forums.musicbrainz.org/viewtopic.php?pid... in case someone can help

reosarevok: I received your pm. can you please save those for issues that need to be private?

reosarevok: please phrase your concerns in a way so that you can freely state them in public.

Ok, so, we have a "Good data being deleted" email that needs BDFL-looking-into

So you might want to take a look at it when you have some time :)

ok, thanks, will do. I'm catching up on email now.

if anyone's curious what maps like https://ddll.inf.tu-dresden.de/web/Wikidata/Map... would look like for mb, here's one: https://dl.dropboxusercontent.com/u/13195166/pl... :)

https://github.com/mineo/ipython-notebooks/blob... is the notebook I used for making it

Hello, guys I installed my own MB slave server with the github install.md instructions. All is working fine (including replication changes). However, I have some questions. 1. How can I disable the memory, time and data statistics when opening through the browser? 2. I'm trying to use the mbcontrol.sh startup script from this guide: http://www.tranquilbase.org/how-to-setup-your-o....

The script is not working for me, only when I use "exec plackup -Ilib -r" instead of "exec plackup -Ilib -s FCGI -E deployment --listen $SOCKET --nproc $NPROC --pid fcgi.pid --keep-stderr=1". What do I have to change to make the last part work?

hi, i'm having some trouble figuring out the id spaces and data schemas... i'm trying to go from searching for a recording to getting information about a work

but in the recording object, there are lots of ids, and none of them seem to resolve when i use them as work ids

rknLA: hey.

works are linked to recordings via what we call Advanced Relationships.

they are not traditional SQL table relationships.

ruaok: hm... ok.. so, is my approach even correct then?

basically, i have a track name and artist name, and want to look up information about the track

sorry, but i'm still super confused by these docs.. there's a lot there that looks vaguely relevant, but i don't understand how it all ties together

using the advanced search you can construct a search with artist and track name to find a recording.

you can send the same query to the web service as well.

so, use the web search to hone your query first.

ah, hm

are you looking for a recording or a work?

a work would be something like "beethoven's 5th".

a recording would be "Beethoven's 5th conducted by Karajan, movement I."

i'm looking for a work

but the docs weren't really clear on how to search for works, and the python library had a way to search for recordings, so i figured there must be some way to go from recording to work

this is how your search for work:

alastairp: does the python bindings support searching works?

there's a `get_work_by_id`, but no search

huh

if not, that's a bug

it also looks like not every recording has a work associated with it?

and a pretty big one at that

rknLA: correct

oh, there is a `search_works` method

alastairp is the person to help you with the python bindings.

however, perhaps you are after the browse methods

alastairp: looks like it's not documented at https://python-musicbrainzngs.readthedocs.org/e...

which let you say "give me items of one type based on relationships to another type"

it's there if i run `dir(musicbrainzngs)`, but i didn't think to do that

https://github.com/alastair/python-musicbrainzn... thanks :)

so, at a high level, my goal is to find the songwriter of a recording.. is there any kind of convention for which recordings have works associated with them?

I'll regenerate the docs

:) glad to help!

ruaok: heh. that last email was... interesting :p

the only convention is "if it exists and someone has added it"

the important part being this "if someone has added it"

word

reosarevok: yep. lots of interesting stuff coming. :)

(I think what other people call Advanced Relationships is https://en.wikipedia.org/wiki/Entity%E2%80%93at...)

ok, I need to run off now. back online later to try and tame my inbox some more.

completely off topic: once upon a time I read that reddit's entire database schema was EAV.

rknLA: https://gist.github.com/alastair/b55768a5c97503...

alastairp: thanks for helping rknLA

I think this is what you're trying to do?

ttfn!

however, you need to start somewhere I guess?

I don't think there's a way of saying "give me all works which have a lyricist or composer set"

alastairp: yeah, basically

right now i'm looking at that like 'search works for: "we will rock you" artist:(queen)' and hoping to get the work directly?

if you know the song name, work search is good

yeah, i know the song name of the recording

oh, perfect

whether or not that's also the name of the original is a different question, but yeah

ah

well, you could always do a recording search, and then add the include=["work-rels"] parameter to get information about works

for example: https://musicbrainz.org/search?query=%22one+of+...

yeah, i think recording search is the ticket

also, it looks like the search method does its damndest to return you _something_ even when there's no real strong/exact match?

also, search doesn't support `includes`, right? so there are at least 3 web calls per attempt here?

correct, search doesn't support includes

you could check for a score of 100 in the results

ahhh score, perfect

you'll probably find that many of the recordings in the search results will have the same work

but I don't think you can find that out until you query each of them

gotcha.

another question, about `id`:

what is the id in a `track-list` item? it didn't look like there was a concept of `track` in mb

yes, tracks are not really exposed in the website much

a track is "a recording in this specific release"

because recordings can appear on many releases

gotcha

hmm... how is `ext:score` calculated? i just got a score of 100 when the artist matched correctly, but the track that i searched for was not in the system

it's returned by lucene. I don't know how it's computed

note that you can use the advanced search syntax and tell it to match only on the title field

yeah, i saw that with the http requests... is that supported by the python lib?

ah, strict=True

great :)

mb.search_recordings("title:we will rock you", strict=True)

for example

gotcha