#metabrainz

yes makes sense. 1.4 deprecates stuff which will be removed in 2.0

So I guess we should try and get these tests working, and then look at the warnings?

Well, maybe we should try to run the indexer on all entities first, to see if that works...

i'll try to fix the tests. it looks like a real issue to me. in the meantime you can test other entities?

> Are many of your listens still without mbids? :o

jasondk: None of my listens have MBIDs unfortunately. That's true locally and on test.lb. I mainly use Spotify and local music players posting to LB, I guess neither send the MBIDs…

reosarevok: Ok, how to setting links as ended?

What I would suggest is to wait until listens have MBIDs for all users before deploying the CB review feature. I know it's not ideal, but currently for a user like me the feature just looks broken and that can give off the wrong impression.

lucifer: ok, I'll do that after lunch

thanks!

MrClon: it used to be super complicated, but we have new code in beta for that now :) So if you go to beta.musicbrainz.org you can set links as ended by clicking the pencil button on the same place you'd add them / remove them

reosarevok: So i should POST to /ws/js/edit/create something like `{"edits":[{"id":3301206,"linkTypeID":75,"ended":false,"edit_type":91}],"makeVotable":false,"editNote":"some note"}`? `id` is relationship ID? How to get that ID?

reosarevok: hi, looks like you're busy, but can I add one more thing to your list and ask you to give a quick look over https://musicbrainz.wordpress.com/wp-admin/post... when you have time?

yyoung[m]: yes, unless relationships to such links are modified

MrClon: from the database, I don't know if there's another way. As I said, this is not meant to be a normal way of doing edits :)

alastairp: sure, in a bit :)

thanks!

MrClon: also, entering edits automatically should be done through a bot account: https://musicbrainz.org/doc/Bots

i know

If someone change broken link to correct, it will be new relationship with new reationship ID? Or it just update one row in urls table?

It depends: Small changes (such as http <-> https) can be made directly to the URL entity (just update one row in url table). Larger changes are expected to replace both the URL and the relationship with new ones.

But a relationship id cannot be reused to target another URL entity.

alastairp: "The fact that we can identify the same recording as being similar even with the MusicBrainz ID is different is interesting"

without?

"The data files generated by Annoy are stand-alone" this is the first mention of Annoy in the post. Either it's useful (maybe it can use more info) or it's not (and can be dropped) or it's a useful extra detail for people who know about that stuff (and then maybe it could be "The data files (generated by Annoy) are stand-alone")

"If you think that you might be interested in this, let us know" did you stop but want to write more, or is that the end? (since there's no punctuation, I'm not sure)

If it's the end, maybe "let us know!" ? :)

The general post seems good (and cool!)

reosarevok: "even when the MBID is different", will fix

yeah, that last paragraph was from the first draft, and I guess it survived. I'll add in a bit about annoy or make the last para more generic, and make a better ending

in fact, I think there was a ! at the end

thank you

reosarevok: can you please push your changes. i think i have a potential fix.

lucifer: sure, one sec :)

alastairp: i forgot my jenkins password 😓, do you know how i can reset it?

lucifer: done

👍 thanks!

Can you push to my branch?

Or should I just push this on top of the existing PR?

existing PR sounds good.

Ok

Wtf, I must be doing something stupid

It tells me everything up to date after a push but https://github.com/metabrainz/sir/pull/111/commits still doesn't have my commits

Oh, nevermind

Name clash

Now there

lucifer: all yours

👍

lucifer: please review https://github.com/metabrainz/musicbrainz-docke...

yvanzo: about that, i faced the issue again today. adding `/code/venv-musicbrainz-docker/lib/python2.7/site-packages` to PYTHONPATH fixed it.

i checked the default sys.path and it also has both /usr/local/lib/python2.7 and /usr/local/lib/python2.7/site-packages.

but i am not sure if this is indeed the right way so will look further into it and let you know.

Ok, thanks

"I could not find docs on this but doing a `vars` dump"

Sigh

:)

I wish stuff got documented properly, especially for huge libraries...

woo. tests pass now.

let's run it for all entities now.

I can try that on my sample data first

If that works I guess maybe we should try with non-sample?

But that'll take forever and I don't have it set up

let's do it on sample. i don't have the full dump either.

Ok, I'll run it

Yes, sample data would be good enough.

once it works on sample, we can either run on bono for a full run or sir_test.

I'm currently playing with sir-test.

ah ok

lucifer: uh-oh

yup i ran and saw

Oh, ok

a mess it is

load_only won't take the name, its wants the column object. undefer takes the name

Hmmm. But load_only seemed to happily work with the same defer did

And https://docs.sqlalchemy.org/en/14/orm/loading_c... suggests that defer and undefer take the same

yes. acc to my understanding we have column names. defer and undefer take those.

But load_only seemed to be taking those too, when I ran just instrument

right, it works for instrument which is surprising. the docs do not show load_only working with column names. even the args are named attrs and for defer and undefer its column key.

heh load_only has been available since 0.9.0!

Hmm

I tried to run work only, got

So, a second, different error ?

yeah looks like it

How the hell do all tests pass, in any case. Are the tests not actually testing indexing at all? :D

the wonders of mocks :)

I do feel, indeed, mocked

o.O

I wonder if it's somehow connected to PoolWorker-7 2021-08-31 12:49:03,663 DEBUG: Loading only [] on <class 'sir.schema.modelext.CustomLinkAttribute'>

(to it being passed nothing)

Nope

i am inclined to get rid of all this entirely. i see that the default for relationship loading is lazy but cannot find the default for columns but it looks like undeferred by default.

Get rid of all this as in, skip eager loading completely?

yes. for now. until we get this working.

Does it work if we remove the load_only call? Let's test

Looks like it's working

Slowly, but, sent 200 documents a few times

i am curently running for url. ok so far.

Work, here

Seems crazy slow, but of course, it might have been crazy slow either way

Since I'm certainly below speca

Specs too

Heh. It crashed my Chrome, even

Too little memory I guess

Still working tho

lucifer: did URL finish for you?

nope.

how many urls are there?

approximately in the sample dump

A lot, I guess :D

I can check, one sec