1
0
mirror of https://github.com/mikf/gallery-dl.git synced 2024-11-24 03:32:33 +01:00
gallery-dl/gallery_dl/extractor/itaku.py
2022-11-24 11:09:00 +01:00

194 lines
6.6 KiB
Python

# -*- coding: utf-8 -*-
# Copyright 2022 Mike Fährmann
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License version 2 as
# published by the Free Software Foundation.
"""Extractors for https://itaku.ee/"""
from .common import Extractor, Message
from ..cache import memcache
from .. import text
BASE_PATTERN = r"(?:https?://)?itaku\.ee"
class ItakuExtractor(Extractor):
"""Base class for itaku extractors"""
category = "itaku"
root = "https://itaku.ee"
directory_fmt = ("{category}", "{owner_username}")
filename_fmt = ("{id}{title:? //}.{extension}")
archive_fmt = "{id}"
request_interval = (0.5, 1.5)
def __init__(self, match):
Extractor.__init__(self, match)
self.api = ItakuAPI(self)
self.item = match.group(1)
self.videos = self.config("videos", True)
def items(self):
for post in self.posts():
post["date"] = text.parse_datetime(
post["date_added"], "%Y-%m-%dT%H:%M:%S.%fZ")
for category, tags in post.pop("categorized_tags").items():
post["tags_" + category.lower()] = [t["name"] for t in tags]
post["tags"] = [t["name"] for t in post["tags"]]
sections = []
for s in post["sections"]:
group = s["group"]
if group:
sections.append(group["title"] + "/" + s["title"])
else:
sections.append(s["title"])
post["sections"] = sections
if post["video"] and self.videos:
url = post["video"]["video"]
else:
url = post["image"]
yield Message.Directory, post
yield Message.Url, url, text.nameext_from_url(url, post)
class ItakuGalleryExtractor(ItakuExtractor):
"""Extractor for posts from an itaku user gallery"""
subcategory = "gallery"
pattern = BASE_PATTERN + r"/profile/([^/?#]+)/gallery"
test = ("https://itaku.ee/profile/piku/gallery", {
"pattern": r"https://d1wmr8tlk3viaj\.cloudfront\.net/gallery_imgs"
r"/[^/?#]+\.(jpg|png|gif)",
"range": "1-10",
"count": 10,
})
def posts(self):
return self.api.galleries_images(self.item)
class ItakuImageExtractor(ItakuExtractor):
subcategory = "image"
pattern = BASE_PATTERN + r"/images/(\d+)"
test = (
("https://itaku.ee/images/100471", {
"pattern": r"https://d1wmr8tlk3viaj\.cloudfront\.net/gallery_imgs"
r"/220504_oUNIAFT\.png",
"count": 1,
"keyword": {
"already_pinned": None,
"blacklisted": {
"blacklisted_tags": [],
"is_blacklisted": False
},
"can_reshare": True,
"date": "dt:2022-05-05 19:21:17",
"date_added": "2022-05-05T19:21:17.674148Z",
"date_edited": "2022-05-25T14:37:46.220612Z",
"description": "sketch from drawpile",
"extension": "png",
"filename": "220504_oUNIAFT",
"hotness_score": float,
"id": 100471,
"image": "https://d1wmr8tlk3viaj.cloudfront.net/gallery_imgs"
"/220504_oUNIAFT.png",
"image_xl": "https://d1wmr8tlk3viaj.cloudfront.net"
"/gallery_imgs/220504_oUNIAFT/lg.jpg",
"liked_by_you": False,
"maturity_rating": "SFW",
"num_comments": int,
"num_likes": int,
"num_reshares": int,
"obj_tags": 136446,
"owner": 16775,
"owner_avatar": "https://d1wmr8tlk3viaj.cloudfront.net"
"/profile_pics/av2022r_vKYVywc/md.jpg",
"owner_displayname": "Piku",
"owner_username": "piku",
"reshared_by_you": False,
"sections": ["Fanart/Miku"],
"tags": list,
"tags_character": ["hatsune_miku"],
"tags_copyright": ["vocaloid"],
"tags_general" : ["twintails", "green_hair", "flag",
"gloves", "green_eyes", "female",
"racing_miku"],
"title": "Racing Miku 2022 Ver.",
"too_mature": False,
"uncompressed_filesize": "0.62",
"video": None,
"visibility": "PUBLIC",
},
}),
# video
("https://itaku.ee/images/19465", {
"pattern": r"https://d1wmr8tlk3viaj\.cloudfront\.net/gallery_vids"
r"/sleepy_af_OY5GHWw\.mp4",
}),
)
def posts(self):
return (self.api.image(self.item),)
class ItakuAPI():
def __init__(self, extractor):
self.extractor = extractor
self.root = extractor.root + "/api"
self.headers = {
"Accept": "application/json, text/plain, */*",
"Referer": extractor.root + "/",
}
def galleries_images(self, username, section=None):
endpoint = "/galleries/images/"
params = {
"cursor" : None,
"owner" : self.user(username)["owner"],
"section" : section,
"date_range": "",
"maturity_rating": ("SFW", "Questionable", "NSFW"),
"ordering" : "-date_added",
"page" : "1",
"page_size" : "30",
"visibility": ("PUBLIC", "PROFILE_ONLY"),
}
return self._pagination(endpoint, params, self.image)
def image(self, image_id):
endpoint = "/galleries/images/{}/".format(image_id)
return self._call(endpoint)
@memcache(keyarg=1)
def user(self, username):
return self._call("/user_profiles/{}/".format(username))
def _call(self, endpoint, params=None):
if not endpoint.startswith("http"):
endpoint = self.root + endpoint
response = self.extractor.request(
endpoint, params=params, headers=self.headers)
return response.json()
def _pagination(self, endpoint, params, extend):
data = self._call(endpoint, params)
while True:
if extend:
for result in data["results"]:
yield extend(result["id"])
else:
yield from data["results"]
url_next = data["links"].get("next")
if not url_next:
return
data = self._call(url_next)