#beets

Indeed. :/

AppVeyor is so slow compared to Travis

2 of the builds haven't even started yet for me

Yeah, it stays "queued" for a really long time.

good work so far on the windows tests sampsyo

We're getting close!

are you gonna be around later?

Maybe a little, but not too much later, alas—I've got some plans later on

so everythong from say somelibraryentry.path shouldbe a bytestring right?

Yes.

after py3 is merged, the first thing to do is reverse most of the work i'm doing right now.. with pathlib

Yes, I suppose so. Also a good chunk of work.

well the more i work with the codebase, the easier it'll be

True!

btw.. sqlite won't accept bytestrings

:(

err i mean the sqlite connect method

Oh! That's better.

(But still bad.)

but at least we'll mostly be ok with that.. compared to the library directory

so mediafile properties.. are we doing an unnecessary decode now on py2 (it's certainly not possible on py3)

py2 propertes can't be unicode can they?

Property names? No, they're bytes on python 2.

…but if they're being used somewhere as unicode, then yeah, they'll need decoding

if isinstance(descriptor, MediaField):

- yield property.decode('utf8')

+ yield property

it passes the tests without it on py2

yeah those aren't true properties.. it's actually dict keys specifically there

Yeah, but it produces a dict containing bytes instead of text on Python 2, which could cause problems. :/

oh, but they are used with getattr() later i think... so they are

i'm now up to like 18 commits on my py3 branch .. minus the one you just merged

That is, we'd need to check what the clients expect: if they want "native strings" exclusively (for property lookup), then it's fine.

so, i ran into a problem last night. where does the translation to bytes happen that mutagen expects when saving a property via mediafile?

like say the 'customtag' example

What does everyone think of just running python 2.7 and 3.3 tests?

I can't exactly imagine anything changing between 3.3, 3.4 and 3.5

jrobeson: There are lots of places where encoding happens, unfortunately. Does grepping for `.encode` in mediafile.py help?

i just know that in sanitize it doesn't do anything for bytes there

jackwilsdon: That seems fine for now—when we actually get Python 3 compatibility, it is nice to know we didn't run afoul of anything in the version changes.

and we really shouldn't be turning them into bytes until we send it to mutagen should we?

i guess that's another whole can of worms

but the way it is no.. it gets to try to save it.. and _none_value() doesn't send bytes

and none_value is what gets saved

mutagen expects bytes

nothing from here onwards https://github.com/beetbox/beets/blob/master/be... seems to send bytes if the value is none at least

and it doesn't look like the storage styles coerce to bytes either

It's not always the case that Mutagen wants bytes, though—some formats want Unicode.

ah

well it does when sending setting mb_trackid

which is where the first failure is

unless maybe it's the key and not the value that it's expecting to be bytes

i'll check that

ah, it's 26 commits..

Eugh I really don't like how weird Tox is

I can't believe there's no other virtualenv testing systems like it though

seems like the problem is more with the tox configuration parsing than tox itself tho at least?

It's the way they've designed the file I think

It's just odd

maybe instead write a script to build a tox config manually :)

hah

Also they seem to loosely match environment names which is silly

I can do something like tox -e py27-flake8-test-docs

And it will try and jam the commands end to end

And do some weird shit

it definitely doesn't fit how other tools work

Yeah

Ok so I seem to have got a really nice configuration now

Where chaining works

So I can do tox -e py27-test-flake8 and it will run nose and then flake8

I'm starting to understand tox

didn't we want flake8 before nose? or is that too much trouble?

You can just do tox -e py27-flake8-test :D

We're going to run them separately anyway for now

It's just something nice

does it fail now?

Yeah everything seems to be alright

just updating travis

And going to push it and try it

so what about the min-version issue that sampsyo mentioned?

I'm going to re-add that to the config

Our command line argument overrides it anyway

So we can keep it for support sake

Not entirely sure how we can abstract away the --show-skipped

Not sure what to call the env for it

why not SHOW_SKIPPED ?

For the test name?

AHA

If we export

NOSE_SHOW_SKIPPED

Nose already shows skipped

It's a "hidden" feature that's not documented of showskipped

So we can set that

well then :)

add a comment mentioning that it's undocumented way to show skipped tests

Yeah I will do

you might need to pass it down through the tox env variables

Hmm yeah IDK if tox forwards them

Good point

it doesn't

Eugh

but you can tell it to forawrd specific ones

it's a whitelist

and now i think you've exhausted everything i know about tox

Aha, passenv

Thanks :)

uggh.. i feel bad about this comment that's about to show up

not sure how i missed that one

What's up with it?

it was meant to be part of the previous commit,but since it doesn't break on py2 i iddn't notice it

Ah

at least i keep finding backportable stuff

i still have a few more backportable things like providing a __str__ method

in some places

i have 2 different git stashes that i keep checking things against

Eugh Git Stashes confuse me :P

IDK why, they make me panic I'm going to lose my stuff

Even though it's difficult to lose stuff on Git

i wouldn't have a problem redoing either tho at least

one is mostly adding b' EVERYWHERE

that made a ton of tests pass

but it's way over the top

Why is it we need to do that exactly?

Are strings unicode by default in 3?

yes

Ah, I see how that could break things

whati 'm trying to do is avoid any architectural refactorings

and all library paths and things in beets are bytestrings

but beets has a bunch of places where it works in py2 because bytes and strings are almost the same thing

Yeah that sounds a bit wobbly haha

so assertTrue(somebytestring, 'somesamestring') pass in py2 , but not 3