@alastairp ok . I have already looked the lists but i had some doubts. I will review the list once again and select a ticket to work on and ask my doubts.
alastairp
if you want to work on AB, tell me and I can pull up some simple tasks
for LB, perhaps ishaanshah or iliekcomputers or shivam-kapila might be able to recommend you something
abhinavohri
I have already opened a pr for CB-367.
Please tell some tasks for AB also.
alastairp
I saw the open CB tickets, thanks. I'll get to those soon
File "/Users/robert/metabrainz/troi-recommendation-playground/.test/lib/python3.7/site-packages/troi-0.0.1rc0-py3.7.egg/troi/utils.py", line 22, in discover_patches
so it's using the version in the venv
and I suspect it's trying to import troi.tools.area_lookup from the egg
when you import from a python shell it imports it from the source tree
ruaok
ahh, I see.
so add . to path?
alastairp
no, that's a recipe for disaster (and will probably fail when people pip install without having access to the rest of the source tree)
did you `pip install -e .` ?
ruaok
no.
python setup.py install
alastairp
ah, right
it's possible that setuptools didn't find troi.tools properly
if you eye over `python setup.py build` does it "look right"? e.g. does it seem like it's finding all of the packages
ruaok
oh, troi.tools actually doesn't seem to be installed.
alastairp
yeah, that's what I was getting at
ruaok
it has no __init__.py
ok, onward to the next problem. thanks for yer help.
alastairp
... that's weird that `import` in a python shell worked, then
ruaok
actually it does make sense -- recent pythons no longer always need __init__.py
how are you planning on running it during development? install as an editable pip, and then use the `troi` executable?
ruaok
yes
alastairp
I _think_ you should also be able to get away with `python -m troi.cli <x>` without using pip install -e or python setup.py something. Let me check that
you'll need to exclude it from find_packages in setup.py
and there shouldn't be an issue with overlapping patches with the same name, right? Because they'll have different import paths (troi.patches.x vs patches.x)
ruaok
not sure.
Mr_Monkey: is the export playlist to spotify features for BP still on your radar?
Yes, but probably after implementing a playlist/queue of sorts in BP.
ruaok
Mr_Monkey: 👍
kori has quit
kori joined the channel
kori has quit
kori joined the channel
yvanzo
:)
Mr_Monkey
ruaok: The PR i just opened above is to pass on any extra error information we get from the spotify player, following your report of 'internal server error' messages last Friday. Not sure I'll be able to do anything about the errors themselves (they're thrown by spotify's sdk, not by us), but at least we should have more info displayed in the error message where available.
ruaok
perfect, thanks!
alastairp
yvanzo: remind me the syntax to link PRs to jira?
ruaok
at least this will help us understand what is going on...
alastairp
does it have to be at the beginning of the title?
shivam-kapila
ruaok: Mr_Monkey: This also happen ed with me
But consple showed me that unable to refresh token
Console*
That returned 404 for spotify api
yvanzo
alastairp: Preferably yes, for example: LB-XXX: Trigger BrainzGit!
More complex titles are possible: LB-XXX, LB-YYY: Trigger BrainzGit again!
what's the best way to say that a PR is related to this ticket (and I want it to show up on the jira list), but I don't want it to close the issue when the PR is merged?
yvanzo
it's not currently implemented in our automation rules.
alastairp
right
yvanzo
It could be changed to still add a comment with a link (if no such link has been posted already) to issues mentioned in parenthesis. (?)
alastairp
ah, I like that idea
so it'll link if the issue is mentioned, but will only close it if it has the : ?
yvanzo
yup, I will check if that is doable, cause the scripting syntax is quite limited.
oh, I guess because when you run `troi`, the current directory isn't in sys.path :(
ruaok
python is not able to import patches.do_not_include_this, it can't find the patches dir.
hmmm.
alastairp
that's a bit more hairy. in this case I guess it _does_ make sense to add `.` to sys.path, but I think we would have to be really careful about doing it
ruaok
and undoing it asap.
alastairp
unless you can somehow import by absolute file path
I see that `importlib.import_module` has a 'package' parameter, not sure if you can specify '.' as that or somethign
BrainzGit
[troi-recommendation-playground] mayhem opened pull request #16 (main…add-local-patches-dir): Add support for loading patches from the patches directory in the current directory. https://github.com/metabrainz/troi-recommendati...
ruaok
nice and productive day. :)
thanks alastairp !
alastairp
yeah, you're right that it just required a few days of concentrated effort to pick up a bunch of loose ends
iliekcomputers they are still wrong for people who import from lastfm
I am writing a fix for that
iliekcomputers
Huh
Quick explanation on why?
ishaanshah
Theres a bug in incremental dump creation
You can check the chatlogs above
iliekcomputers
Created and listened at timestamps are supposed to be different
That's by design
pristine___
ruaok: so I was thinking about artist similarity. What would be the training set? [Artist 1 id, artist 2 id, count] i.e how many collaborations artist 1 and artist 2 have. So if artist 1 appeared in let's say 10 collaborations with artist 3 and artist 1 and artist 2 are similar then artist 2 and artist 3 will also be similar, something like that.
ruaok
I was planning a different tack....
given a user and all of their listens, we can make a histogram of artists based on their listens.
rather than a histogram based in their recordings
ishaanshah
iliekcomputers i know that but while creating incremental dumps we are filtering by listened at rather than created
Which wont work for lastfm imports
Cause that means only listens from the day when import was done would be included
Which is happening with my data, I checked in the dumps
pristine___
ruaok: do we want to solve *what are the artists similar to artist A* or *what are the artists user A might like*?
ruaok
oh. good point. my solution solves the wrong problem. thanks for catching that.
I mean that is useful too.
pristine___
Yeah, but one at a time :p
ruaok
in fact, it might work really well. hard to say.
pristine___
We can try both.
ruaok
I'm thinking that the version I described ought to be easy to implement...