[simplyhentai] add gallery extractor (#89)

pull/133/head
Mike Fährmann 6 years ago
parent ae9a37a528
commit 55b0913412
No known key found for this signature in database
GPG Key ID: 5680CA389D365A88

@ -67,6 +67,7 @@ Sankaku Channel https://chan.sankakucomplex.com/ Pools, Posts, Tag-Searc
Sea Otter Scans https://reader.seaotterscans.com/ Chapters, Manga
Sen Manga http://raw.senmanga.com/ Chapters
Sense-Scans http://sensescans.com/ Chapters, Manga
Simply Hentai https://www.simply-hentai.com/ Galleries
SlideShare https://www.slideshare.net/ Presentations
SmugMug https://www.smugmug.com/ |Albums, individ-5| Optional (OAuth)
Subapics https://subapics.com/ Chapters, Manga

@ -74,6 +74,7 @@ modules = [
"seiga",
"senmanga",
"sensescans",
"simplyhentai",
"slideshare",
"smugmug",
"subapics",

@ -0,0 +1,74 @@
# -*- coding: utf-8 -*-
# Copyright 2018 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.
"""Extract hentai-manga from https://www.simply-hentai.com/"""
from .common import ChapterExtractor
from .. import text, util, exception
class SimplyhentaiGalleryExtractor(ChapterExtractor):
"""Extractor for image galleries from simply-hentai.com"""
category = "simplyhentai"
subcategory = "gallery"
directory_fmt = ["{category}", "{gallery_id} {title}"]
filename_fmt = "{category}_{gallery_id}_{page:>03}.{extension}"
archive_fmt = "{image_id}"
pattern = [r"(?:https?://)?(?!videos)[^.]+\.simply-hentai\.com"
r"(?:/(?!page|series|album|all-pages)[^/?&#]+)+"]
test = [
(("https://original-work.simply-hentai.com"
"/amazon-no-hiyaku-amazon-elixir"), {
"url": "35f3843d0ea83e6a618df7afaebd2b03f3628db9",
"keyword": "1e22ccbe66412eab844f135ad9cd3424b8b064e8",
}),
("https://www.simply-hentai.com/notfound", {
"exception": exception.GalleryDLException,
}),
# custom subdomain
("https://pokemon.simply-hentai.com/mao-friends-9bc39", None),
# www subdomain, two path segments
("https://www.simply-hentai.com/vocaloid/black-magnet", None),
]
def __init__(self, match):
ChapterExtractor.__init__(self, match.group(0))
self.session.headers["Referer"] = match.group(0)
def get_metadata(self, page):
extr = text.extract
title , pos = extr(page, '<meta property="og:title" content="', '"')
if not title:
raise exception.NotFoundError("gallery")
gid , pos = extr(page, '/Album/', '/', pos)
series, pos = extr(page, 'box-title">Series</div>', '</div>', pos)
lang , pos = extr(page, 'box-title">Language</div>', '</div>', pos)
chars , pos = extr(page, 'box-title">Characters</div>', '</div>', pos)
tags , pos = extr(page, 'box-title">Tags</div>', '</div>', pos)
artist, pos = extr(page, 'box-title">Artists</div>', '</div>', pos)
date , pos = extr(page, 'Uploaded', '</div>', pos)
lang = text.remove_html(lang) if lang else None
return {
"gallery_id": text.parse_int(gid),
"title": text.unescape(title),
"series": text.remove_html(series),
"characters": ", ".join(text.split_html(chars)),
"tags": ", ".join(text.split_html(tags)),
"artist": ", ".join(text.split_html(artist)),
"lang": util.language_to_code(lang),
"language": lang,
"date": text.remove_html(date),
}
def get_images(self, _):
images = self.request(self.url + "/all-pages.json").json()
return [
(urls["full"], {"image_id": text.parse_int(image_id)})
for image_id, urls in sorted(images.items())
]

@ -52,6 +52,7 @@ CATEGORY_MAP = {
"seiga" : "Niconico Seiga",
"senmanga" : "Sen Manga",
"sensescans" : "Sense-Scans",
"simplyhentai" : "Simply Hentai",
"slideshare" : "SlideShare",
"smugmug" : "SmugMug",
"thebarchive" : "The /b/ Archive",

Loading…
Cancel
Save