mirror of
https://github.com/searxng/searxng.git
synced 2024-11-30 06:31:04 +00:00
3d96a9839a
This patch was generated by black [1]:: make format.python [1] https://github.com/psf/black Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
99 lines
2.7 KiB
Python
99 lines
2.7 KiB
Python
# SPDX-License-Identifier: AGPL-3.0-or-later
|
|
"""
|
|
Recoll (local search engine)
|
|
"""
|
|
|
|
from datetime import date, timedelta
|
|
from json import loads
|
|
from urllib.parse import urlencode, quote
|
|
|
|
# about
|
|
about = {
|
|
"website": None,
|
|
"wikidata_id": 'Q15735774',
|
|
"official_api_documentation": 'https://www.lesbonscomptes.com/recoll/',
|
|
"use_official_api": True,
|
|
"require_api_key": False,
|
|
"results": 'JSON',
|
|
}
|
|
|
|
# engine dependent config
|
|
paging = True
|
|
time_range_support = True
|
|
|
|
# parameters from settings.yml
|
|
base_url = None
|
|
search_dir = ''
|
|
mount_prefix = None
|
|
dl_prefix = None
|
|
|
|
# embedded
|
|
embedded_url = '<{ttype} controls height="166px" ' + 'src="{url}" type="{mtype}"></{ttype}>'
|
|
|
|
|
|
# helper functions
|
|
def get_time_range(time_range):
|
|
sw = {'day': 1, 'week': 7, 'month': 30, 'year': 365}
|
|
|
|
offset = sw.get(time_range, 0)
|
|
if not offset:
|
|
return ''
|
|
|
|
return (date.today() - timedelta(days=offset)).isoformat()
|
|
|
|
|
|
# do search-request
|
|
def request(query, params):
|
|
search_after = get_time_range(params['time_range'])
|
|
search_url = base_url + 'json?{query}&highlight=0'
|
|
params['url'] = search_url.format(
|
|
query=urlencode({'query': query, 'page': params['pageno'], 'after': search_after, 'dir': search_dir})
|
|
)
|
|
|
|
return params
|
|
|
|
|
|
# get response from search-request
|
|
def response(resp):
|
|
results = []
|
|
|
|
response_json = loads(resp.text)
|
|
|
|
if not response_json:
|
|
return []
|
|
|
|
for result in response_json.get('results', []):
|
|
title = result['label']
|
|
url = result['url'].replace('file://' + mount_prefix, dl_prefix)
|
|
content = '{}'.format(result['snippet'])
|
|
|
|
# append result
|
|
item = {'url': url, 'title': title, 'content': content, 'template': 'files.html'}
|
|
|
|
if result['size']:
|
|
item['size'] = int(result['size'])
|
|
|
|
for parameter in ['filename', 'abstract', 'author', 'mtype', 'time']:
|
|
if result[parameter]:
|
|
item[parameter] = result[parameter]
|
|
|
|
# facilitate preview support for known mime types
|
|
if 'mtype' in result and '/' in result['mtype']:
|
|
(mtype, subtype) = result['mtype'].split('/')
|
|
item['mtype'] = mtype
|
|
item['subtype'] = subtype
|
|
|
|
if mtype in ['audio', 'video']:
|
|
item['embedded'] = embedded_url.format(
|
|
ttype=mtype, url=quote(url.encode('utf8'), '/:'), mtype=result['mtype']
|
|
)
|
|
|
|
if mtype in ['image'] and subtype in ['bmp', 'gif', 'jpeg', 'png']:
|
|
item['img_src'] = url
|
|
|
|
results.append(item)
|
|
|
|
if 'nres' in response_json:
|
|
results.append({'number_of_results': response_json['nres']})
|
|
|
|
return results
|