mirror of
https://github.com/searxng/searxng.git
synced 2024-11-05 12:50:11 +01:00
1540891561
Closes: https://github.com/searxng/searxng/issues/1449 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
226 lines
7.2 KiB
Python
226 lines
7.2 KiB
Python
# SPDX-License-Identifier: AGPL-3.0-or-later
|
|
# lint: pylint
|
|
"""This engine implements *Tineye - reverse image search*
|
|
|
|
Using TinEye, you can search by image or perform what we call a reverse image
|
|
search. You can do that by uploading an image or searching by URL. You can also
|
|
simply drag and drop your images to start your search. TinEye constantly crawls
|
|
the web and adds images to its index. Today, the TinEye index is over 50.2
|
|
billion images `[tineye.com] <https://tineye.com/how>`_.
|
|
|
|
.. hint::
|
|
|
|
This SearXNG engine only supports *'searching by URL'* and it does not use
|
|
the official API `[api.tineye.com] <https://api.tineye.com/python/docs/>`_.
|
|
|
|
"""
|
|
|
|
from urllib.parse import urlencode
|
|
from datetime import datetime
|
|
from flask_babel import gettext
|
|
|
|
about = {
|
|
"website": 'https://tineye.com',
|
|
"wikidata_id": 'Q2382535',
|
|
"official_api_documentation": 'https://api.tineye.com/python/docs/',
|
|
"use_official_api": False,
|
|
"require_api_key": False,
|
|
"results": 'JSON',
|
|
}
|
|
|
|
engine_type = 'online_url_search'
|
|
""":py:obj:`searx.search.processors.online_url_search`"""
|
|
|
|
categories = ['general']
|
|
paging = True
|
|
safesearch = False
|
|
base_url = 'https://tineye.com'
|
|
search_string = '/result_json/?page={page}&{query}'
|
|
|
|
FORMAT_NOT_SUPPORTED = gettext(
|
|
"Could not read that image url. This may be due to an unsupported file"
|
|
" format. TinEye only supports images that are JPEG, PNG, GIF, BMP, TIFF or WebP."
|
|
)
|
|
"""TinEye error message"""
|
|
|
|
NO_SIGNATURE_ERROR = gettext(
|
|
"The image is too simple to find matches. TinEye requires a basic level of"
|
|
" visual detail to successfully identify matches."
|
|
)
|
|
"""TinEye error message"""
|
|
|
|
DOWNLOAD_ERROR = gettext("The image could not be downloaded.")
|
|
"""TinEye error message"""
|
|
|
|
|
|
def request(query, params):
|
|
"""Build TinEye HTTP request using ``search_urls`` of a :py:obj:`engine_type`."""
|
|
|
|
params['raise_for_httperror'] = False
|
|
|
|
if params['search_urls']['data:image']:
|
|
query = params['search_urls']['data:image']
|
|
elif params['search_urls']['http']:
|
|
query = params['search_urls']['http']
|
|
|
|
logger.debug("query URL: %s", query)
|
|
query = urlencode({'url': query})
|
|
|
|
# see https://github.com/TinEye/pytineye/blob/main/pytineye/api.py
|
|
params['url'] = base_url + search_string.format(query=query, page=params['pageno'])
|
|
|
|
params['headers'].update(
|
|
{
|
|
'Connection': 'keep-alive',
|
|
'Accept-Encoding': 'gzip, defalte, br',
|
|
'Host': 'tineye.com',
|
|
'DNT': '1',
|
|
'TE': 'trailers',
|
|
}
|
|
)
|
|
return params
|
|
|
|
|
|
def parse_tineye_match(match_json):
|
|
"""Takes parsed JSON from the API server and turns it into a :py:obj:`dict`
|
|
object.
|
|
|
|
Attributes `(class Match) <https://github.com/TinEye/pytineye/blob/main/pytineye/api.py>`__
|
|
|
|
- `image_url`, link to the result image.
|
|
- `domain`, domain this result was found on.
|
|
- `score`, a number (0 to 100) that indicates how closely the images match.
|
|
- `width`, image width in pixels.
|
|
- `height`, image height in pixels.
|
|
- `size`, image area in pixels.
|
|
- `format`, image format.
|
|
- `filesize`, image size in bytes.
|
|
- `overlay`, overlay URL.
|
|
- `tags`, whether this match belongs to a collection or stock domain.
|
|
|
|
- `backlinks`, a list of Backlink objects pointing to the original websites
|
|
and image URLs. List items are instances of :py:obj:`dict`, (`Backlink
|
|
<https://github.com/TinEye/pytineye/blob/main/pytineye/api.py>`__):
|
|
|
|
- `url`, the image URL to the image.
|
|
- `backlink`, the original website URL.
|
|
- `crawl_date`, the date the image was crawled.
|
|
|
|
"""
|
|
|
|
# HINT: there exists an alternative backlink dict in the domains list / e.g.::
|
|
#
|
|
# match_json['domains'][0]['backlinks']
|
|
|
|
backlinks = []
|
|
if "backlinks" in match_json:
|
|
|
|
for backlink_json in match_json["backlinks"]:
|
|
if not isinstance(backlink_json, dict):
|
|
continue
|
|
|
|
crawl_date = backlink_json.get("crawl_date")
|
|
if crawl_date:
|
|
crawl_date = datetime.fromisoformat(crawl_date[:-3])
|
|
else:
|
|
crawl_date = datetime.min
|
|
|
|
backlinks.append(
|
|
{
|
|
'url': backlink_json.get("url"),
|
|
'backlink': backlink_json.get("backlink"),
|
|
'crawl_date': crawl_date,
|
|
'image_name': backlink_json.get("image_name"),
|
|
}
|
|
)
|
|
|
|
return {
|
|
'image_url': match_json.get("image_url"),
|
|
'domain': match_json.get("domain"),
|
|
'score': match_json.get("score"),
|
|
'width': match_json.get("width"),
|
|
'height': match_json.get("height"),
|
|
'size': match_json.get("size"),
|
|
'image_format': match_json.get("format"),
|
|
'filesize': match_json.get("filesize"),
|
|
'overlay': match_json.get("overlay"),
|
|
'tags': match_json.get("tags"),
|
|
'backlinks': backlinks,
|
|
}
|
|
|
|
|
|
def response(resp):
|
|
"""Parse HTTP response from TinEye."""
|
|
results = []
|
|
|
|
try:
|
|
json_data = resp.json()
|
|
except Exception as exc: # pylint: disable=broad-except
|
|
msg = "can't parse JSON response // %s" % exc
|
|
logger.error(msg)
|
|
json_data = {'error': msg}
|
|
|
|
# handle error codes from Tineye
|
|
|
|
if resp.is_error:
|
|
if resp.status_code in (400, 422):
|
|
|
|
message = 'HTTP status: %s' % resp.status_code
|
|
error = json_data.get('error')
|
|
s_key = json_data.get('suggestions', {}).get('key', '')
|
|
|
|
if error and s_key:
|
|
message = "%s (%s)" % (error, s_key)
|
|
elif error:
|
|
message = error
|
|
|
|
if s_key == "Invalid image URL":
|
|
# test https://docs.searxng.org/_static/searxng-wordmark.svg
|
|
message = FORMAT_NOT_SUPPORTED
|
|
elif s_key == 'NO_SIGNATURE_ERROR':
|
|
# test https://pngimg.com/uploads/dot/dot_PNG4.png
|
|
message = NO_SIGNATURE_ERROR
|
|
elif s_key == 'Download Error':
|
|
# test https://notexists
|
|
message = DOWNLOAD_ERROR
|
|
|
|
# see https://github.com/searxng/searxng/pull/1456#issuecomment-1193105023
|
|
# results.append({'answer': message})
|
|
logger.error(message)
|
|
|
|
return results
|
|
|
|
resp.raise_for_status()
|
|
|
|
# append results from matches
|
|
|
|
for match_json in json_data['matches']:
|
|
|
|
tineye_match = parse_tineye_match(match_json)
|
|
if not tineye_match['backlinks']:
|
|
continue
|
|
|
|
backlink = tineye_match['backlinks'][0]
|
|
results.append(
|
|
{
|
|
'template': 'images.html',
|
|
'url': backlink['backlink'],
|
|
'thumbnail_src': tineye_match['image_url'],
|
|
'source': backlink['url'],
|
|
'title': backlink['image_name'],
|
|
'img_src': backlink['url'],
|
|
'format': tineye_match['image_format'],
|
|
'widht': tineye_match['width'],
|
|
'height': tineye_match['height'],
|
|
'publishedDate': backlink['crawl_date'],
|
|
}
|
|
)
|
|
|
|
# append number of results
|
|
|
|
number_of_results = json_data.get('num_matches')
|
|
if number_of_results:
|
|
results.append({'number_of_results': number_of_results})
|
|
|
|
return results
|