| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
wanted to do this before, but per comment:
the official docs led me to believe that importlib_resources was the only
way to do it, and that the thing i was trying to do wasn't supported in
some python versions
so i put it off to one side for later
later came (now), and i tried to implement it with importlib.resources
didn't work, i looked it up and found this stackoverflow post
https://stackoverflow.com/a/58941536
and the second answer, using pkgutil (not importlib.resources), just worked
and the syntax/api seems much nicer as well
have tested with pip yt-dlp + raw directory plugin,
and binary yt-dlp + .whl plugin
also on windows, with binary+whl
both worked, amazing success! so pkgutil seems like the way to go
i also updated the build thing to include .bin files in the extractor dir
|
| |
|
|
| |
parity with yt-dlp proper
|
| | |
|
| | |
|
| |
|
|
| |
now links to the actual issue page and prints info useful for debugging
|
| |
|
|
|
| |
it does not need to be cryptographically strong
only using token_hex for convenience
|
| |
|
|
|
|
|
|
|
|
| |
also fixed a rather embarrassing bug
the generated user agent was, for example:
Dalvik/2.1.0 (Linux; U; Android {'sdk': '27', 'builds': ['5794017', '6107733', '6037697']};SOV36/6107733)
because i put the wrong variable name in the fstring
thats alright its only been sending botched identification for
its entire existence(!)
|
| |
|
|
|
| |
makes it a bit easier to use in the cli since you don't need to 'quote'
for the #!
|
| |
|
|
| |
there is a live format that also doesn't work
|
| |
|
|
| |
closes #18
|
| |
|
|
|
|
|
|
|
|
|
| |
so now it's just datetime.datetime with custom input and a few new funcs
instead of actually having a datetime.datetime internally and jankily
kind-of-exposing it
doesn't really affect things currently, but it makes something i want to
do possible (and it's generally just the better way of doing it)
Co-Authored-By: 9382 <omnomaidan@gmail.com>
|
| | |
|
| |
|
|
|
| |
housou daigaku is going away soon, so switching to radio nikkei
https://www.ouj.ac.jp/news/2023/information/5-32-2.html
|
| |
|
|
| |
mostly whitespace changes
|
| | |
|
| |
|
|
|
|
| |
deduped code for working it out
also fixed bug in broadcast_day_end where it shifted forward by a day,
but didn't also set the time to 5am
|
| |
|
|
|
|
|
|
|
| |
the site will accept this so we have to as well
i should probably try and handle this in SiteTime
but cant be fucked right now
iirc the site just blanks out if the day is out of range
so the datetime exception matches the site's behaviour 👍
|
| |
|
|
|
| |
extractor shouldnt be executable
time thing should so i can test it easier
|
| |
|
|
| |
closes #17
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
| |
Station metadata changes extremely rarely, if ever, so it seems awfully
silly to keep downloading the exact same metadata every time we extract.
Wastes requests, wastes time (300ms round trip for me)
This commit makes it so station metadata is cached for a day, and cached
data is used in place of downloaded metadata when available.
Closes #16
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The old behaviour assumed programmes were deleted precisely a week after
their end time. This isn't actually the case though, as long as it's
within a week of the _broadcast day_, the site will let you.
This lead to the extractor bailing out (Programme is no longer available)
when in fact the programme was still playable on the site.
The fix uses the same logic as RadikoTime.broadcast_day to find the
broadcast day (TODO: one func for both?), then sets the time to 05:00:00
the next day - i.e. the start of a new broadcast day.
|
| | |
|
| | |
|
| |
|
|
|
| |
now only one thing gets passed around and it has most everything we need
closes #11
|
| | |
|
| |
|
|
| |
ie dont do the dumb noformats vars, just raise the message there and then
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
afaict nhk doesn't have timefree, and it says as much on the programme pages
The download doesn't fail though, you just get a loop of elevator music
and an apology that it's not available (lasting the duration of the programme)
from tsDetail.js
tsInNg = playable in the area
tsOutNg = playable outside the area
(maybe - just a guess really, but seems to line up)
outside station area + playable in area + not playable outside = not available in your area
outside station area + not playable in area + not playable outside area = not available on timefree
inside station area + not playable in area = not available on timefree
we'll always be in the station area, so we only have to check tsInNg
included a message about NHK Radiru because so far i've only seen NHK programmes
be unavailable
one day i should make the programme_meta func return a dict or something
instead of just adding stuff on the end ad infinitum
closes #15
|
| |
|
|
|
|
|
|
|
| |
fixes #14
new headers thing made b'bytes' into "b'bytes'", instead of just 'bytes'
so it sent eg `X-Radiko-Partialkey: b'9lOaAg+2TZRSF+IsiDnUJg=='`
likely a fix for this upstream soon, but will do the sensible thing and make it ascii myself
https://discord.com/channels/807245652072857610/1112613156934668338/1134879336512634880 (yt-dlp discord)
|
| | |
|
| |
|
|
|
| |
ERROR: Response.info() is deprecated, use Response.headers
see #14
|
| | |
|
| |
|
|
| |
fixes #13
|
| | |
|
| | |
|
| |
|
|
| |
so when embedding it gets tagged as the artist
|
| | |
|
| |
|
|
| |
closes #6
|
| |
|
|
| |
closes #4
|
| |
|
|
| |
not smort enough for traverse_obj
|
| | |
|
| |
|
|
| |
one of these days i should look in to a way to do this automatically
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
FM_OKINAWA, FM_OITA
|
| | |
|
