1
0
mirror of https://github.com/janeczku/calibre-web synced 2024-11-16 06:44:54 +00:00
calibre-web/cps/metadata_provider/douban.py
xlivevil 73567db4fb
Update douban metadata provider
Change search API
Change to Return an empty list when an error occurs
Change the way to get tags
Fix series and publisher perse error
Rename a variable with the same name as the built-in
2022-10-10 01:49:42 +08:00

259 lines
8.7 KiB
Python

# -*- coding: utf-8 -*-
# This file is part of the Calibre-Web (https://github.com/janeczku/calibre-web)
# Copyright (C) 2022 xlivevil
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
import re
from concurrent import futures
from typing import List, Optional
import requests
from html2text import HTML2Text
from lxml import etree
from cps import logger
from cps.services.Metadata import Metadata, MetaRecord, MetaSourceInfo
log = logger.create()
def html2text(html: str) -> str:
h2t = HTML2Text()
h2t.body_width = 0
h2t.single_line_break = True
h2t.emphasis_mark = "*"
return h2t.handle(html)
class Douban(Metadata):
__name__ = "豆瓣"
__id__ = "douban"
DESCRIPTION = "豆瓣"
META_URL = "https://book.douban.com/"
SEARCH_JSON_URL = "https://www.douban.com/j/search"
SEARCH_URL = "https://www.douban.com/search"
ID_PATTERN = re.compile(r"sid: (?P<id>\d+),")
AUTHORS_PATTERN = re.compile(r"作者|译者")
PUBLISHER_PATTERN = re.compile(r"出版社")
SUBTITLE_PATTERN = re.compile(r"副标题")
PUBLISHED_DATE_PATTERN = re.compile(r"出版年")
SERIES_PATTERN = re.compile(r"丛书")
IDENTIFIERS_PATTERN = re.compile(r"ISBN|统一书号")
CRITERIA_PATTERN = re.compile("criteria = '(.+)'")
TITTLE_XPATH = "//span[@property='v:itemreviewed']"
COVER_XPATH = "//a[@class='nbg']"
INFO_XPATH = "//*[@id='info']//span[@class='pl']"
TAGS_XPATH = "//a[contains(@class, 'tag')]"
DESCRIPTION_XPATH = "//div[@id='link-report']//div[@class='intro']"
RATING_XPATH = "//div[@class='rating_self clearfix']/strong"
session = requests.Session()
session.headers = {
'user-agent':
'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/98.0.4758.102 Safari/537.36 Edg/98.0.1108.56',
}
def search(self,
query: str,
generic_cover: str = "",
locale: str = "en") -> List[MetaRecord]:
val = []
if self.active:
log.debug(f"start searching {query} on douban")
if title_tokens := list(
self.get_title_tokens(query, strip_joiners=False)):
query = "+".join(title_tokens)
book_id_list = self._get_book_id_list_from_html(query)
if not book_id_list:
log.debug("No search results in Douban")
return []
with futures.ThreadPoolExecutor(
max_workers=5, thread_name_prefix='douban') as executor:
fut = [
executor.submit(self._parse_single_book, book_id,
generic_cover) for book_id in book_id_list
]
val = [
future.result() for future in futures.as_completed(fut)
if future.result()
]
return val
def _get_book_id_list_from_html(self, query: str) -> List[str]:
try:
r = self.session.get(self.SEARCH_URL,
params={
"cat": 1001,
"q": query
})
r.raise_for_status()
except Exception as e:
log.warning(e)
return []
html = etree.HTML(r.content.decode("utf8"))
result_list = html.xpath(self.COVER_XPATH)
return [
self.ID_PATTERN.search(item.get("onclick")).group("id")
for item in result_list[:10]
if self.ID_PATTERN.search(item.get("onclick"))
]
def _get_book_id_list_from_json(self, query: str) -> List[str]:
try:
r = self.session.get(self.SEARCH_JSON_URL,
params={
"cat": 1001,
"q": query
})
r.raise_for_status()
except Exception as e:
log.warning(e)
return []
results = r.json()
if results["total"] == 0:
return []
return [
self.ID_PATTERN.search(item).group("id")
for item in results["items"][:10] if self.ID_PATTERN.search(item)
]
def _parse_single_book(self,
id: str,
generic_cover: str = "") -> Optional[MetaRecord]:
url = f"https://book.douban.com/subject/{id}/"
log.debug(f"start parsing {url}")
try:
r = self.session.get(url)
r.raise_for_status()
except Exception as e:
log.warning(e)
return None
match = MetaRecord(
id=id,
title="",
authors=[],
url=url,
source=MetaSourceInfo(
id=self.__id__,
description=self.DESCRIPTION,
link=self.META_URL,
),
)
html = etree.HTML(r.content.decode("utf8"))
match.title = html.xpath(self.TITTLE_XPATH)[0].text
match.cover = html.xpath(
self.COVER_XPATH)[0].attrib["href"] or generic_cover
try:
rating_num = float(html.xpath(self.RATING_XPATH)[0].text.strip())
except Exception:
rating_num = 0
match.rating = int(-1 * rating_num // 2 * -1) if rating_num else 0
tag_elements = html.xpath(self.TAGS_XPATH)
if len(tag_elements):
match.tags = [tag_element.text for tag_element in tag_elements]
else:
match.tags = self._get_tags(html.text)
description_element = html.xpath(self.DESCRIPTION_XPATH)
if len(description_element):
match.description = html2text(
etree.tostring(description_element[-1]).decode("utf8"))
info = html.xpath(self.INFO_XPATH)
for element in info:
text = element.text
if self.AUTHORS_PATTERN.search(text):
next_element = element.getnext()
while next_element is not None and next_element.tag != "br":
match.authors.append(next_element.text)
next_element = next_element.getnext()
elif self.PUBLISHER_PATTERN.search(text):
if publisher := element.tail.strip():
match.publisher = publisher
else:
match.publisher = element.getnext().text
elif self.SUBTITLE_PATTERN.search(text):
match.title = f'{match.title}:{element.tail.strip()}'
elif self.PUBLISHED_DATE_PATTERN.search(text):
match.publishedDate = self._clean_date(element.tail.strip())
elif self.SERIES_PATTERN.search(text):
match.series = element.getnext().text
elif i_type := self.IDENTIFIERS_PATTERN.search(text):
match.identifiers[i_type.group()] = element.tail.strip()
return match
def _clean_date(self, date: str) -> str:
"""
Clean up the date string to be in the format YYYY-MM-DD
Examples of possible patterns:
'2014-7-16', '1988年4月', '1995-04', '2021-8', '2020-12-1', '1996年',
'1972', '2004/11/01', '1959年3月北京第1版第1印'
"""
year = date[:4]
moon = "01"
day = "01"
if len(date) > 5:
digit = []
ls = []
for i in range(5, len(date)):
if date[i].isdigit():
digit.append(date[i])
elif digit:
ls.append("".join(digit) if len(digit) ==
2 else f"0{digit[0]}")
digit = []
if digit:
ls.append("".join(digit) if len(digit) ==
2 else f"0{digit[0]}")
moon = ls[0]
if len(ls) > 1:
day = ls[1]
return f"{year}-{moon}-{day}"
def _get_tags(self, text: str) -> List[str]:
tags = []
if criteria := self.CRITERIA_PATTERN.search(text):
tags.extend(
item.replace('7:', '') for item in criteria.group().split('|')
if item.startswith('7:'))
return tags