mirror of
https://github.com/mikf/gallery-dl.git
synced 2024-11-25 12:12:34 +01:00
93 lines
3.4 KiB
Python
93 lines
3.4 KiB
Python
# -*- coding: utf-8 -*-
|
|
|
|
# Copyright 2020 Jake Mannens
|
|
# Copyright 2021-2023 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://mangakakalot.tv/"""
|
|
|
|
from .common import ChapterExtractor, MangaExtractor
|
|
from .. import text
|
|
import re
|
|
|
|
BASE_PATTERN = r"(?:https?://)?(?:ww[\dw]?\.)?mangakakalot\.tv"
|
|
|
|
|
|
class MangakakalotBase():
|
|
"""Base class for mangakakalot extractors"""
|
|
category = "mangakakalot"
|
|
root = "https://ww8.mangakakalot.tv"
|
|
|
|
|
|
class MangakakalotChapterExtractor(MangakakalotBase, ChapterExtractor):
|
|
"""Extractor for manga chapters from mangakakalot.tv"""
|
|
pattern = BASE_PATTERN + r"(/chapter/[^/?#]+/chapter[_-][^/?#]+)"
|
|
example = "https://ww6.mangakakalot.tv/chapter/manga-ID/chapter-01"
|
|
|
|
def __init__(self, match):
|
|
self.path = match.group(1)
|
|
ChapterExtractor.__init__(self, match, self.root + self.path)
|
|
|
|
def metadata(self, page):
|
|
_ , pos = text.extract(page, '<span itemprop="title">', '<')
|
|
manga , pos = text.extract(page, '<span itemprop="title">', '<', pos)
|
|
info , pos = text.extract(page, '<span itemprop="title">', '<', pos)
|
|
author, pos = text.extract(page, '. Author:', ' already has ', pos)
|
|
|
|
match = re.match(
|
|
r"(?:[Vv]ol\. *(\d+) )?"
|
|
r"[Cc]hapter *([^:]*)"
|
|
r"(?:: *(.+))?", info or "")
|
|
volume, chapter, title = match.groups() if match else ("", "", info)
|
|
chapter, sep, minor = chapter.partition(".")
|
|
|
|
return {
|
|
"manga" : text.unescape(manga),
|
|
"title" : text.unescape(title) if title else "",
|
|
"author" : text.unescape(author).strip() if author else "",
|
|
"volume" : text.parse_int(volume),
|
|
"chapter" : text.parse_int(chapter),
|
|
"chapter_minor": sep + minor,
|
|
"lang" : "en",
|
|
"language" : "English",
|
|
}
|
|
|
|
def images(self, page):
|
|
return [
|
|
(url, None)
|
|
for url in text.extract_iter(page, '<img data-src="', '"')
|
|
]
|
|
|
|
|
|
class MangakakalotMangaExtractor(MangakakalotBase, MangaExtractor):
|
|
"""Extractor for manga from mangakakalot.tv"""
|
|
chapterclass = MangakakalotChapterExtractor
|
|
pattern = BASE_PATTERN + r"(/manga/[^/?#]+)"
|
|
example = "https://ww6.mangakakalot.tv/manga/manga-ID"
|
|
|
|
def chapters(self, page):
|
|
data = {"lang": "en", "language": "English"}
|
|
data["manga"], pos = text.extract(page, "<h1>", "<")
|
|
author, pos = text.extract(page, "<li>Author(s) :", "</a>", pos)
|
|
data["author"] = text.remove_html(author)
|
|
|
|
results = []
|
|
for chapter in text.extract_iter(page, '<div class="row">', '</div>'):
|
|
url, pos = text.extract(chapter, '<a href="', '"')
|
|
title, pos = text.extract(chapter, '>', '</a>', pos)
|
|
data["title"] = title.partition(": ")[2]
|
|
data["date"] , pos = text.extract(
|
|
chapter, '<span title=" ', '"', pos)
|
|
|
|
chapter, sep, minor = url.rpartition("/chapter-")[2].partition(".")
|
|
data["chapter"] = text.parse_int(chapter)
|
|
data["chapter_minor"] = sep + minor
|
|
|
|
if url[0] == "/":
|
|
url = self.root + url
|
|
results.append((url, data.copy()))
|
|
return results
|