# -*- coding: utf-8 -*- # Copyright (C) 2004-2008 Tristan Seligmann and Jonathan Jacobs # Copyright (C) 2012-2014 Bastian Kleineidam # Copyright (C) 2015-2020 Tobias Gruetzmacher # Copyright (C) 2019-2020 Daniel Ring from __future__ import absolute_import, division, print_function from re import compile from six.moves.urllib.parse import urljoin from ..helpers import bounceStarter, xpath_class from ..scraper import _BasicScraper, _ParserScraper from ..util import tagre from .common import _WordPressScraper, _WPWebcomic class RalfTheDestroyer(_WordPressScraper): url = 'http://ralfthedestroyer.com/' class RaynaOnTheRiver(_WordPressScraper): url = 'http://www.catomix.com/rayna/' firstStripUrl = url + 'archives/comic/teaser-poster' class RealLife(_WordPressScraper): url = 'https://reallifecomics.com/' stripUrl = url + 'comic.php?comic=%s' firstStripUrl = stripUrl % 'title-1' help = 'Index format: monthname-dd-yyyy' def namer(self, imageUrl, pageUrl): # Fix inconsisntent filenames filename = imageUrl.rsplit('/', 1)[-1] if pageUrl.rsplit('=', 1)[-1] == 'may-27-2014': filename = filename.replace('20140219_3121', '20140527') filename = filename.replace('5-Finished', '20140623_3161') filename = filename.replace('520140722', '20140722') filename = filename.replace('520140724', '20140724') return filename def getPrevUrl(self, url, data): # "Parse" JavaScript prevtag = data.find_class('comic-nav-previous') if not prevtag: return None target = prevtag[0].get('onclick').split("'")[1] return urljoin(url, target) class RealmOfAtland(_BasicScraper): url = 'http://www.realmofatland.com/' stripUrl = url + '?p=%s' firstStripUrl = stripUrl % '1' prevSearch = compile(tagre("a", "href", r'(\?p=\d+)', after="cg_back")) imageSearch = compile(tagre("img", "src", r'(images/strips/atland\d+.[^"]+)')) help = 'Index format: nnn' class RedMeat(_ParserScraper): url = 'http://www.redmeat.com/max-cannon/FreshMeat' imageSearch = '//div[@class="comicStrip"]//img' prevSearch = '//a[@class="prev"]' def namer(self, image_url, page_url): parts = image_url.rsplit('/', 2) return '_'.join(parts[1:3]) class Replay(_ParserScraper): url = 'http://replaycomic.com/' stripUrl = url + 'comic/%s/' url = stripUrl % 'trying-it-out' firstStripUrl = stripUrl % 'red-desert' imageSearch = '//div[@id="comic"]//img' prevSearch = '//a[contains(@class, "comic-nav-previous")]' nextSearch = '//a[contains(@class, "comic-nav-next")]' def starter(self): # Retrieve archive page to identify chapters archivePage = self.getPage(self.url + 'archive') archive = archivePage.xpath('//div[@class="comic-archive-chapter-wrap"]') self.chapter = len(archive) - 1 self.startOfChapter = [] for archiveChapter in archive: self.startOfChapter.append(archiveChapter.xpath('.//a')[0].get('href')) return bounceStarter(self) def namer(self, imageUrl, pageUrl): # Name pages based on chapter, index, and post title name = pageUrl.rstrip('/').rsplit('/', 1)[-1] page = imageUrl.rsplit('/', 1)[-1].rsplit('.', 1) # Fix inconsistent page number formatting if page[0].isdigit() and len(page[0]) > 2 and self.chapter == 1 and name != 'through-the-woods': page[0] = page[0][:2] + '-' + page[0][2:] name = '%d-%s-%s.%s' % (self.chapter, page[0], name, page[1]) if pageUrl in self.startOfChapter: self.chapter -= 1 return name class RiversideExtras(_WPWebcomic): url = 'https://riversidecomics.com/' class RomanticallyApocalyptic(_ParserScraper): url = 'http://romanticallyapocalyptic.com/' stripUrl = url + '%s' firstStripUrl = stripUrl % '0' imageSearch = '//div[%s]/center//img' % xpath_class('comicpanel') prevSearch = '//a[@accesskey="p"]' help = 'Index format: n' adult = True class Roza(_ParserScraper): url = 'http://www.junglestudio.com/roza/index.php' stripUrl = url + '?date=%s' firstStripUrl = stripUrl % '2007-05-01' imageSearch = '//img[contains(@src, "pages/")]' prevSearch = '//a[img[contains(@src, "navtable_01.gif")]]' help = 'Index format: yyyy-mm-dd' class Ruthe(_BasicScraper): url = 'http://ruthe.de/' stripUrl = url + 'cartoon/%s/datum/asc/' firstStripUrl = stripUrl % '1' lang = 'de' imageSearch = compile(tagre("img", "src", r'(/?cartoons/strip_\d+[^"]+)')) prevSearch = compile(tagre("a", "href", r'(/cartoon/\d+/datum/asc/)') + 'vorheriger') help = 'Index format: number' class Ryugou(_WPWebcomic): url = 'http://ryugou.swashbuckledcomics.com/' stripUrl = url + 'comic/%s/' firstStripUrl = 'ryugou-chapter-1-cover' starter = bounceStarter def namer(self, imageUrl, pageUrl): title = pageUrl.rstrip('/').rsplit('/', 1)[-1] ext = imageUrl.rsplit('.', 1)[-1] return title + '.' + ext def fetchUrls(self, url, data, urlSearch): imageUrls = super(Ryugou, self).fetchUrls(url, data, urlSearch) if url == self.stripUrl % '1-3': imageUrls = [imageUrls[1]] return imageUrls