calibre-web/cps/search.py

403 lines
18 KiB
Python

# This file is part of the Calibre-Web (https://github.com/janeczku/calibre-web)
# Copyright (C) 2022 OzzieIsaacs
#
# 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 json
from datetime import datetime
from flask import Blueprint, request, redirect, url_for, flash
from flask import session as flask_session
from flask_login import current_user
from flask_babel import format_date
from flask_babel import gettext as _
from sqlalchemy.sql.expression import func, not_, and_, or_, text, true
from sqlalchemy.sql.functions import coalesce
from . import logger, db, calibre_db, config, ub
from .usermanagement import login_required_if_no_ano
from .render_template import render_title_template
from .pagination import Pagination
search = Blueprint('search', __name__)
log = logger.create()
@search.route("/search", methods=["GET"])
@login_required_if_no_ano
def simple_search():
term = request.args.get("query")
if term:
return redirect(url_for('web.books_list', data="search", sort_param='stored', query=term.strip()))
else:
return render_title_template('search.html',
searchterm="",
result_count=0,
title=_("Search"),
page="search")
@search.route("/advsearch", methods=['POST'])
@login_required_if_no_ano
def advanced_search():
values = dict(request.form)
params = ['include_tag', 'exclude_tag', 'include_serie', 'exclude_serie', 'include_shelf', 'exclude_shelf',
'include_language', 'exclude_language', 'include_extension', 'exclude_extension']
for param in params:
values[param] = list(request.form.getlist(param))
flask_session['query'] = json.dumps(values)
return redirect(url_for('web.books_list', data="advsearch", sort_param='stored', query=""))
@search.route("/advsearch", methods=['GET'])
@login_required_if_no_ano
def advanced_search_form():
# Build custom columns names
cc = calibre_db.get_cc_columns(config, filter_config_custom_read=True)
return render_prepare_search_form(cc)
def adv_search_custom_columns(cc, term, q):
for c in cc:
if c.datatype == "datetime":
custom_start = term.get('custom_column_' + str(c.id) + '_start')
custom_end = term.get('custom_column_' + str(c.id) + '_end')
if custom_start:
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
func.datetime(db.cc_classes[c.id].value) >= func.datetime(custom_start)))
if custom_end:
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
func.datetime(db.cc_classes[c.id].value) <= func.datetime(custom_end)))
else:
custom_query = term.get('custom_column_' + str(c.id))
if custom_query != '' and custom_query is not None:
if c.datatype == 'bool':
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
db.cc_classes[c.id].value == (custom_query == "True")))
elif c.datatype == 'int' or c.datatype == 'float':
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
db.cc_classes[c.id].value == custom_query))
elif c.datatype == 'rating':
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
db.cc_classes[c.id].value == int(float(custom_query) * 2)))
else:
q = q.filter(getattr(db.Books, 'custom_column_' + str(c.id)).any(
func.lower(db.cc_classes[c.id].value).ilike("%" + custom_query + "%")))
return q
def adv_search_language(q, include_languages_inputs, exclude_languages_inputs):
if current_user.filter_language() != "all":
q = q.filter(db.Books.languages.any(db.Languages.lang_code == current_user.filter_language()))
else:
for language in include_languages_inputs:
q = q.filter(db.Books.languages.any(db.Languages.id == language))
for language in exclude_languages_inputs:
q = q.filter(not_(db.Books.series.any(db.Languages.id == language)))
return q
def adv_search_ratings(q, rating_high, rating_low):
if rating_high:
rating_high = int(rating_high) * 2
q = q.filter(db.Books.ratings.any(db.Ratings.rating <= rating_high))
if rating_low:
rating_low = int(rating_low) * 2
q = q.filter(db.Books.ratings.any(db.Ratings.rating >= rating_low))
return q
def adv_search_read_status(read_status):
if not config.config_read_column:
if read_status == "True":
db_filter = and_(ub.ReadBook.user_id == int(current_user.id),
ub.ReadBook.read_status == ub.ReadBook.STATUS_FINISHED)
else:
db_filter = coalesce(ub.ReadBook.read_status, 0) != ub.ReadBook.STATUS_FINISHED
else:
try:
if read_status == "True":
db_filter = db.cc_classes[config.config_read_column].value == True
else:
db_filter = coalesce(db.cc_classes[config.config_read_column].value, False) != True
except (KeyError, AttributeError, IndexError):
log.error("Custom Column No.{} does not exist in calibre database".format(config.config_read_column))
flash(_("Custom Column No.%(column)d does not exist in calibre database",
column=config.config_read_column),
category="error")
return true()
return db_filter
def adv_search_extension(q, include_extension_inputs, exclude_extension_inputs):
for extension in include_extension_inputs:
q = q.filter(db.Books.data.any(db.Data.format == extension))
for extension in exclude_extension_inputs:
q = q.filter(not_(db.Books.data.any(db.Data.format == extension)))
return q
def adv_search_tag(q, include_tag_inputs, exclude_tag_inputs):
for tag in include_tag_inputs:
q = q.filter(db.Books.tags.any(db.Tags.id == tag))
for tag in exclude_tag_inputs:
q = q.filter(not_(db.Books.tags.any(db.Tags.id == tag)))
return q
def adv_search_serie(q, include_series_inputs, exclude_series_inputs):
for serie in include_series_inputs:
q = q.filter(db.Books.series.any(db.Series.id == serie))
for serie in exclude_series_inputs:
q = q.filter(not_(db.Books.series.any(db.Series.id == serie)))
return q
def adv_search_shelf(q, include_shelf_inputs, exclude_shelf_inputs):
q = q.outerjoin(ub.BookShelf, db.Books.id == ub.BookShelf.book_id)\
.filter(or_(ub.BookShelf.shelf == None, ub.BookShelf.shelf.notin_(exclude_shelf_inputs)))
if len(include_shelf_inputs) > 0:
q = q.filter(ub.BookShelf.shelf.in_(include_shelf_inputs))
return q
def extend_search_term(searchterm,
author_name,
book_title,
publisher,
pub_start,
pub_end,
tags,
rating_high,
rating_low,
read_status,
):
searchterm.extend((author_name.replace('|', ','), book_title, publisher))
if pub_start:
try:
searchterm.extend([_("Published after ") +
format_date(datetime.strptime(pub_start, "%Y-%m-%d"),
format='medium')])
except ValueError:
pub_start = ""
if pub_end:
try:
searchterm.extend([_("Published before ") +
format_date(datetime.strptime(pub_end, "%Y-%m-%d"),
format='medium')])
except ValueError:
pub_end = ""
elements = {'tag': db.Tags, 'serie':db.Series, 'shelf':ub.Shelf}
for key, db_element in elements.items():
tag_names = calibre_db.session.query(db_element).filter(db_element.id.in_(tags['include_' + key])).all()
searchterm.extend(tag.name for tag in tag_names)
tag_names = calibre_db.session.query(db_element).filter(db_element.id.in_(tags['exclude_' + key])).all()
searchterm.extend(tag.name for tag in tag_names)
language_names = calibre_db.session.query(db.Languages). \
filter(db.Languages.id.in_(tags['include_language'])).all()
if language_names:
language_names = calibre_db.speaking_language(language_names)
searchterm.extend(language.name for language in language_names)
language_names = calibre_db.session.query(db.Languages). \
filter(db.Languages.id.in_(tags['exclude_language'])).all()
if language_names:
language_names = calibre_db.speaking_language(language_names)
searchterm.extend(language.name for language in language_names)
if rating_high:
searchterm.extend([_("Rating <= %(rating)s", rating=rating_high)])
if rating_low:
searchterm.extend([_("Rating >= %(rating)s", rating=rating_low)])
if read_status:
searchterm.extend([_("Read Status = %(status)s", status=read_status)])
searchterm.extend(ext for ext in tags['include_extension'])
searchterm.extend(ext for ext in tags['exclude_extension'])
# handle custom columns
searchterm = " + ".join(filter(None, searchterm))
return searchterm, pub_start, pub_end
def render_adv_search_results(term, offset=None, order=None, limit=None):
sort = order[0] if order else [db.Books.sort]
pagination = None
cc = calibre_db.get_cc_columns(config, filter_config_custom_read=True)
calibre_db.session.connection().connection.connection.create_function("lower", 1, db.lcase)
query = calibre_db.generate_linked_query(config.config_read_column, db.Books)
q = query.outerjoin(db.books_series_link, db.Books.id == db.books_series_link.c.book)\
.outerjoin(db.Series)\
.filter(calibre_db.common_filters(True))
# parse multi selects to a complete dict
tags = dict()
elements = ['tag', 'serie', 'shelf', 'language', 'extension']
for element in elements:
tags['include_' + element] = term.get('include_' + element)
tags['exclude_' + element] = term.get('exclude_' + element)
author_name = term.get("author_name")
book_title = term.get("book_title")
publisher = term.get("publisher")
pub_start = term.get("publishstart")
pub_end = term.get("publishend")
rating_low = term.get("ratinghigh")
rating_high = term.get("ratinglow")
description = term.get("comment")
read_status = term.get("read_status")
if author_name:
author_name = author_name.strip().lower().replace(',', '|')
if book_title:
book_title = book_title.strip().lower()
if publisher:
publisher = publisher.strip().lower()
search_term = []
cc_present = False
for c in cc:
if c.datatype == "datetime":
column_start = term.get('custom_column_' + str(c.id) + '_start')
column_end = term.get('custom_column_' + str(c.id) + '_end')
if column_start:
search_term.extend(["{} >= {}".format(c.name,
format_date(datetime.strptime(column_start, "%Y-%m-%d").date(),
format='medium')
)])
cc_present = True
if column_end:
search_term.extend(["{} <= {}".format(c.name,
format_date(datetime.strptime(column_end, "%Y-%m-%d").date(),
format='medium')
)])
cc_present = True
elif term.get('custom_column_' + str(c.id)):
search_term.extend([("{}: {}".format(c.name, term.get('custom_column_' + str(c.id))))])
cc_present = True
if any(tags.values()) or author_name or book_title or publisher or pub_start or pub_end or rating_low \
or rating_high or description or cc_present or read_status:
search_term, pub_start, pub_end = extend_search_term(search_term,
author_name,
book_title,
publisher,
pub_start,
pub_end,
tags,
rating_high,
rating_low,
read_status)
if author_name:
q = q.filter(db.Books.authors.any(func.lower(db.Authors.name).ilike("%" + author_name + "%")))
if book_title:
q = q.filter(func.lower(db.Books.title).ilike("%" + book_title + "%"))
if pub_start:
q = q.filter(func.datetime(db.Books.pubdate) > func.datetime(pub_start))
if pub_end:
q = q.filter(func.datetime(db.Books.pubdate) < func.datetime(pub_end))
q = q.filter(adv_search_read_status(read_status))
if publisher:
q = q.filter(db.Books.publishers.any(func.lower(db.Publishers.name).ilike("%" + publisher + "%")))
q = adv_search_tag(q, tags['include_tag'], tags['exclude_tag'])
q = adv_search_serie(q, tags['include_serie'], tags['exclude_serie'])
q = adv_search_shelf(q, tags['include_shelf'], tags['exclude_shelf'])
q = adv_search_extension(q, tags['include_extension'], tags['exclude_extension'])
q = adv_search_language(q, tags['include_language'], tags['exclude_language'])
q = adv_search_ratings(q, rating_high, rating_low)
if description:
q = q.filter(db.Books.comments.any(func.lower(db.Comments.text).ilike("%" + description + "%")))
# search custom columns
try:
q = adv_search_custom_columns(cc, term, q)
except AttributeError as ex:
log.debug_or_exception(ex)
flash(_("Error on search for custom columns, please restart Calibre-Web"), category="error")
q = q.order_by(*sort).all()
flask_session['query'] = json.dumps(term)
ub.store_combo_ids(q)
result_count = len(q)
if offset is not None and limit is not None:
offset = int(offset)
limit_all = offset + int(limit)
pagination = Pagination((offset / (int(limit)) + 1), limit, result_count)
else:
offset = 0
limit_all = result_count
entries = calibre_db.order_authors(q[offset:limit_all], list_return=True, combined=True)
return render_title_template('search.html',
adv_searchterm=search_term,
pagination=pagination,
entries=entries,
result_count=result_count,
title=_("Advanced Search"), page="advsearch",
order=order[1])
def render_prepare_search_form(cc):
# prepare data for search-form
tags = calibre_db.session.query(db.Tags)\
.join(db.books_tags_link)\
.join(db.Books)\
.filter(calibre_db.common_filters()) \
.group_by(text('books_tags_link.tag'))\
.order_by(db.Tags.name).all()
series = calibre_db.session.query(db.Series)\
.join(db.books_series_link)\
.join(db.Books)\
.filter(calibre_db.common_filters()) \
.group_by(text('books_series_link.series'))\
.order_by(db.Series.name)\
.filter(calibre_db.common_filters()).all()
shelves = ub.session.query(ub.Shelf)\
.filter(or_(ub.Shelf.is_public == 1, ub.Shelf.user_id == int(current_user.id)))\
.order_by(ub.Shelf.name).all()
extensions = calibre_db.session.query(db.Data)\
.join(db.Books)\
.filter(calibre_db.common_filters()) \
.group_by(db.Data.format)\
.order_by(db.Data.format).all()
if current_user.filter_language() == "all":
languages = calibre_db.speaking_language()
else:
languages = None
return render_title_template('search_form.html', tags=tags, languages=languages, extensions=extensions,
series=series,shelves=shelves, title=_("Advanced Search"), cc=cc, page="advsearch")
def render_search_results(term, offset=None, order=None, limit=None):
if term:
join = db.books_series_link, db.Books.id == db.books_series_link.c.book, db.Series
entries, result_count, pagination = calibre_db.get_search_results(term,
config,
offset,
order,
limit,
*join)
else:
entries = list()
order = [None, None]
pagination = result_count = None
return render_title_template('search.html',
searchterm=term,
pagination=pagination,
query=term,
adv_searchterm=term,
entries=entries,
result_count=result_count,
title=_("Search"),
page="search",
order=order[1])