From 3ea11f5d5e94a8b1b0abef80cd1499459008ad23 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Mike=20F=C3=A4hrmann?= Date: Wed, 27 Feb 2019 16:54:25 +0100 Subject: [PATCH] [nhentai] rewrite - use GalleryExtractor as base class - extract a lot more metadata (artist, tags, etc.) --- gallery_dl/extractor/nhentai.py | 129 +++++++++++++++++++------------- 1 file changed, 77 insertions(+), 52 deletions(-) diff --git a/gallery_dl/extractor/nhentai.py b/gallery_dl/extractor/nhentai.py index 372531ed..bd23d9fd 100644 --- a/gallery_dl/extractor/nhentai.py +++ b/gallery_dl/extractor/nhentai.py @@ -8,73 +8,100 @@ """Extract images from https://nhentai.net/""" -from .common import Extractor, Message -from .. import text +from .common import GalleryExtractor, Extractor, Message +from .. import text, util +import collections import json -class NHentaiExtractor(Extractor): +class NhentaiBase(): """Base class for nhentai extractors""" category = "nhentai" root = "https://nhentai.net" media_url = "https://i.nhentai.net" - @staticmethod - def transform_to_metadata(ginfo): - """Transform an nhentai API response into compatible metadata""" - title_en = ginfo["title"].get("english", "") - title_ja = ginfo["title"].get("japanese", "") - return { - "gallery_id": ginfo["id"], - "upload_date": ginfo["upload_date"], - "media_id": ginfo["media_id"], - "scanlator": ginfo["scanlator"], - "count": ginfo["num_pages"], - "title": title_en or title_ja, - "title_en": title_en, - "title_ja": title_ja, - } - -class NhentaiGalleryExtractor(NHentaiExtractor): +class NhentaiGalleryExtractor(NhentaiBase, GalleryExtractor): """Extractor for image galleries from nhentai.net""" - subcategory = "gallery" - directory_fmt = ("{category}", "{gallery_id} {title}") - filename_fmt = "{category}_{gallery_id}_{num:>03}.{extension}" - archive_fmt = "{gallery_id}_{num}" - pattern = r"(?:https?://)?nhentai\.net/g/(\d+)" + pattern = r"(?:https?://)?nhentai\.net(/g/(\d+))" test = ("https://nhentai.net/g/147850/", { "url": "5179dbf0f96af44005a0ff705a0ad64ac26547d0", - "keyword": "2f94976e657f3043a89997e22f4de8e1b22d9175", + "keyword": { + "title" : r"re:\[Morris\] Amazon no Hiyaku \| Amazon Elixir", + "title_en" : str, + "title_ja" : str, + "gallery_id": 147850, + "media_id" : 867789, + "date" : 1446050915, + "scanlator" : "", + "artist" : ["morris"], + "group" : list, + "parody" : list, + "characters": list, + "tags" : list, + "type" : "manga", + "language" : ["translated", "english"], + "lang" : "en", + "width" : int, + "height" : int, + }, }) def __init__(self, match): - NHentaiExtractor.__init__(self, match) - self.gid = match.group(1) + GalleryExtractor.__init__(self, match) + self.gallery_id = match.group(2) + self.data = None - def items(self): - ginfo = self.get_gallery_info(self.gid) - data = self.transform_to_metadata(ginfo) - urlfmt = "{}/galleries/{}/{{}}.{{}}".format( - self.media_url, data["media_id"]) + def metadata(self, page): + data = json.loads(text.extract(page, "N.gallery(", ");")[0]) + self.data = data + + title_en = data["title"].get("english", "") + title_ja = data["title"].get("japanese", "") + + info = collections.defaultdict(list) + for tag in data["tags"]: + info[tag["type"]].append(tag["name"]) + + for language in info["language"]: + if language != "translated": + lang = util.language_to_code(language) + break + else: + lang = "" + + return { + "title" : title_en or title_ja, + "title_en" : title_en, + "title_ja" : title_ja, + "gallery_id": data["id"], + "media_id" : text.parse_int(data["media_id"]), + "date" : data["upload_date"], + "scanlator" : data["scanlator"], + "artist" : info["artist"], + "group" : info["group"], + "parody" : info["parody"], + "characters": info["character"], + "tags" : info["tag"], + "type" : info["category"][0] if info["category"] else "", + "language" : info["language"], + "lang" : lang, + } + + def images(self, _): + ufmt = "{}/galleries/{}/{{}}.{{}}".format( + self.media_url, self.data["media_id"]) extdict = {"j": "jpg", "p": "png", "g": "gif"} - yield Message.Version, 1 - yield Message.Directory, data - for data["num"], image in enumerate(ginfo["images"]["pages"], 1): - ext = extdict.get(image["t"], "jpg") - data["width"] = image["w"] - data["height"] = image["h"] - data["extension"] = ext - yield Message.Url, urlfmt.format(data["num"], ext), data - def get_gallery_info(self, gallery_id): - """Extract and return info about a gallery by ID""" - url = "{}/g/{}".format(self.root, gallery_id) - page = self.request(url).text - return json.loads(text.extract(page, "N.gallery(", ");")[0]) + return [ + (ufmt.format(num, extdict.get(img["t"], "jpg")), { + "width": img["w"], "height": img["h"], + }) + for num, img in enumerate(self.data["images"]["pages"], 1) + ] -class NhentaiSearchExtractor(NHentaiExtractor): +class NhentaiSearchExtractor(NhentaiBase, Extractor): """Extractor for nhentai search results""" category = "nhentai" subcategory = "search" @@ -86,14 +113,14 @@ class NhentaiSearchExtractor(NHentaiExtractor): }) def __init__(self, match): - NHentaiExtractor.__init__(self, match) + Extractor.__init__(self, match) self.params = text.parse_query(match.group(1)) def items(self): yield Message.Version, 1 data = {"_extractor": NhentaiGalleryExtractor} - for gid in self._pagination(self.params): - url = "{}/g/{}/".format(self.root, gid) + for gallery_id in self._pagination(self.params): + url = "{}/g/{}/".format(self.root, gallery_id) yield Message.Queue, url, data def _pagination(self, params): @@ -102,9 +129,7 @@ class NhentaiSearchExtractor(NHentaiExtractor): while True: page = self.request(url, params=params).text - yield from text.extract_iter(page, 'href="/g/', '/') - if 'class="next"' not in page: return params["page"] += 1