From d923b2a420b77a3a29375516ed9929beeda0924a Mon Sep 17 00:00:00 2001 From: Thomas Pointhuber Date: Mon, 1 Jun 2015 09:15:50 +0200 Subject: [PATCH 1/2] [enh] implement swisscows engine, FIX #347 TODO: implement image engine --- searx/engines/swisscows.py | 86 +++++++++++++++++++++ searx/settings.yml | 4 + searx/tests/engines/test_swisscows.py | 104 ++++++++++++++++++++++++++ searx/tests/test_engines.py | 1 + 4 files changed, 195 insertions(+) create mode 100644 searx/engines/swisscows.py create mode 100644 searx/tests/engines/test_swisscows.py diff --git a/searx/engines/swisscows.py b/searx/engines/swisscows.py new file mode 100644 index 0000000..fe7a6ce --- /dev/null +++ b/searx/engines/swisscows.py @@ -0,0 +1,86 @@ +""" + Swisscows (Web) + + @website https://swisscows.ch + @provide-api no + + @using-api no + @results HTML (using search portal) + @stable no (HTML can change) + @parse url, title, content +""" + +from json import loads +from urllib import urlencode, unquote +import re + +# engine dependent config +categories = ['general'] +paging = True +language_support = True + +# search-url +base_url = 'https://swisscows.ch/' +search_string = '?{query}&page={page}' + +# regex +regex_json = re.compile('initialData: {"Request":(.|\n)*}\]},\s*environment') +regex_json_remove_start = re.compile('^initialData:\s*') +regex_json_remove_end = re.compile(',\s*environment$') +regex_img_url_remove_start = re.compile('^https?://i\.swisscows\.ch/\?link=') + + +# do search-request +def request(query, params): + if params['language'] == 'all': + ui_language = 'browser' + region = 'browser' + else: + region = params['language'].replace('_', '-') + ui_language = params['language'].split('_')[0] + + search_path = search_string.format( + query=urlencode({'query': query, + 'uiLanguage': ui_language, + 'region': region}), + page=params['pageno']) + + params['url'] = base_url + search_path + + return params + + +# get response from search-request +def response(resp): + results = [] + + json_regex = regex_json.search(resp.content) + + # check if results are returned + if not json_regex: + return [] + + json_raw = regex_json_remove_end.sub('', regex_json_remove_start.sub('', json_regex.group())) + json = loads(json_raw) + + # parse normal results + for result in json['Results'].get('items', []): + # append result + results.append({'url': result['Url'].replace(u'\uE000', '').replace(u'\uE001', ''), + 'title': result['Title'].replace(u'\uE000', '').replace(u'\uE001', ''), + 'content': result['Description'].replace(u'\uE000', '').replace(u'\uE001', '')}) + + # parse images + for result in json.get('Images', []): + # decode image url + img_url = unquote(regex_img_url_remove_start.sub('', result['Url'])) + + # append result + results.append({'url': result['SourceUrl'], + 'title': result['Title'], + 'content': '', + 'img_src': img_url, + 'template': 'images.html'}) + + # return results + return results diff --git a/searx/settings.yml b/searx/settings.yml index d35b137..ee8fc3d 100644 --- a/searx/settings.yml +++ b/searx/settings.yml @@ -213,6 +213,10 @@ engines: timeout : 6.0 disabled : True + - name : swisscows + engine : swisscows + shortcut : sw + - name : twitter engine : twitter shortcut : tw diff --git a/searx/tests/engines/test_swisscows.py b/searx/tests/engines/test_swisscows.py new file mode 100644 index 0000000..d529527 --- /dev/null +++ b/searx/tests/engines/test_swisscows.py @@ -0,0 +1,104 @@ +from collections import defaultdict +import mock +from searx.engines import swisscows +from searx.testing import SearxTestCase + + +class TestSwisscowsEngine(SearxTestCase): + + def test_request(self): + query = 'test_query' + dicto = defaultdict(dict) + dicto['pageno'] = 1 + dicto['language'] = 'de_DE' + params = swisscows.request(query, dicto) + self.assertTrue('url' in params) + self.assertTrue(query in params['url']) + self.assertTrue('swisscows.ch' in params['url']) + self.assertTrue('uiLanguage=de' in params['url']) + self.assertTrue('region=de-DE' in params['url']) + + dicto['language'] = 'all' + params = swisscows.request(query, dicto) + self.assertTrue('uiLanguage=browser' in params['url']) + self.assertTrue('region=browser' in params['url']) + + def test_response(self): + self.assertRaises(AttributeError, swisscows.response, None) + self.assertRaises(AttributeError, swisscows.response, []) + self.assertRaises(AttributeError, swisscows.response, '') + self.assertRaises(AttributeError, swisscows.response, '[]') + + response = mock.Mock(content='') + self.assertEqual(swisscows.response(response), []) + + response = mock.Mock(content='') + self.assertEqual(swisscows.response(response), []) + + html = u""" + + """ + response = mock.Mock(content=html) + results = swisscows.response(response) + self.assertEqual(type(results), list) + self.assertEqual(len(results), 2) + self.assertEqual(results[0]['title'], 'This should be the title') + self.assertEqual(results[0]['url'], 'http://this.should.be.the.link/') + self.assertEqual(results[0]['content'], 'This should be the content.') + self.assertEqual(results[1]['title'], 'Datei:This should.svg') + self.assertEqual(results[1]['url'], 'http://de.wikipedia.org/wiki/Datei:This should.svg') + self.assertEqual(results[1]['img_src'], 'http://ts2.mm.This/should.png') + self.assertEqual(results[1]['template'], 'images.html') diff --git a/searx/tests/test_engines.py b/searx/tests/test_engines.py index 5770458..90f0ce8 100644 --- a/searx/tests/test_engines.py +++ b/searx/tests/test_engines.py @@ -32,6 +32,7 @@ from searx.tests.engines.test_spotify import * # noqa from searx.tests.engines.test_stackoverflow import * # noqa from searx.tests.engines.test_startpage import * # noqa from searx.tests.engines.test_subtitleseeker import * # noqa +from searx.tests.engines.test_swisscows import * # noqa from searx.tests.engines.test_twitter import * # noqa from searx.tests.engines.test_vimeo import * # noqa from searx.tests.engines.test_www1x import * # noqa From 2d81079384706aab4c39efcc8a2c4267af4397a5 Mon Sep 17 00:00:00 2001 From: Thomas Pointhuber Date: Mon, 1 Jun 2015 12:30:07 +0200 Subject: [PATCH 2/2] [enh] implement image support for swisscows engine --- searx/engines/swisscows.py | 38 +++++++++++++++++++++------ searx/tests/engines/test_swisscows.py | 30 +++++++++++++++++---- 2 files changed, 55 insertions(+), 13 deletions(-) diff --git a/searx/engines/swisscows.py b/searx/engines/swisscows.py index fe7a6ce..2d31264 100644 --- a/searx/engines/swisscows.py +++ b/searx/engines/swisscows.py @@ -1,5 +1,5 @@ """ - Swisscows (Web) + Swisscows (Web, Images) @website https://swisscows.ch @provide-api no @@ -15,7 +15,7 @@ from urllib import urlencode, unquote import re # engine dependent config -categories = ['general'] +categories = ['general', 'images'] paging = True language_support = True @@ -24,7 +24,7 @@ base_url = 'https://swisscows.ch/' search_string = '?{query}&page={page}' # regex -regex_json = re.compile('initialData: {"Request":(.|\n)*}\]},\s*environment') +regex_json = re.compile('initialData: {"Request":(.|\n)*},\s*environment') regex_json_remove_start = re.compile('^initialData:\s*') regex_json_remove_end = re.compile(',\s*environment$') regex_img_url_remove_start = re.compile('^https?://i\.swisscows\.ch/\?link=') @@ -45,6 +45,10 @@ def request(query, params): 'region': region}), page=params['pageno']) + # image search query is something like 'image?{query}&page={page}' + if params['category'] == 'images': + search_path = 'image' + search_path + params['url'] = base_url + search_path return params @@ -63,12 +67,30 @@ def response(resp): json_raw = regex_json_remove_end.sub('', regex_json_remove_start.sub('', json_regex.group())) json = loads(json_raw) - # parse normal results + # parse results for result in json['Results'].get('items', []): - # append result - results.append({'url': result['Url'].replace(u'\uE000', '').replace(u'\uE001', ''), - 'title': result['Title'].replace(u'\uE000', '').replace(u'\uE001', ''), - 'content': result['Description'].replace(u'\uE000', '').replace(u'\uE001', '')}) + result_title = result['Title'].replace(u'\uE000', '').replace(u'\uE001', '') + + # parse image results + if result.get('ContentType', '').startswith('image'): + img_url = unquote(regex_img_url_remove_start.sub('', result['Url'])) + + # append result + results.append({'url': result['SourceUrl'], + 'title': result['Title'], + 'content': '', + 'img_src': img_url, + 'template': 'images.html'}) + + # parse general results + else: + result_url = result['Url'].replace(u'\uE000', '').replace(u'\uE001', '') + result_content = result['Description'].replace(u'\uE000', '').replace(u'\uE001', '') + + # append result + results.append({'url': result_url, + 'title': result_title, + 'content': result_content}) # parse images for result in json.get('Images', []): diff --git a/searx/tests/engines/test_swisscows.py b/searx/tests/engines/test_swisscows.py index d529527..926ba88 100644 --- a/searx/tests/engines/test_swisscows.py +++ b/searx/tests/engines/test_swisscows.py @@ -51,7 +51,23 @@ class TestSwisscowsEngine(SearxTestCase): "Description":"\uE000This should\uE001 be the content.", "Url":"http://this.should.be.the.link/", "DisplayUrl":"www.\uE000this.should.be.the\uE001.link", - "Id":"782ef287-e439-451c-b380-6ebc14ba033d"} + "Id":"782ef287-e439-451c-b380-6ebc14ba033d"}, + {"Title":"Datei:This should1.svg", + "Url":"https://i.swisscows.ch/?link=http%3a%2f%2fts2.mm.This/should1.png", + "SourceUrl":"http://de.wikipedia.org/wiki/Datei:This should1.svg", + "DisplayUrl":"de.wikipedia.org/wiki/Datei:This should1.svg", + "Width":950, + "Height":534, + "FileSize":92100, + "ContentType":"image/jpeg", + "Thumbnail":{ + "Url":"https://i.swisscows.ch/?link=http%3a%2f%2fts2.mm.This/should1.png", + "ContentType":"image/jpeg", + "Width":300, + "Height":168, + "FileSize":9134}, + "Id":"6a97a542-8f65-425f-b7f6-1178c3aba7be" + } ],"TotalCount":55300, "Query":"This should " }, @@ -94,11 +110,15 @@ class TestSwisscowsEngine(SearxTestCase): response = mock.Mock(content=html) results = swisscows.response(response) self.assertEqual(type(results), list) - self.assertEqual(len(results), 2) + self.assertEqual(len(results), 3) self.assertEqual(results[0]['title'], 'This should be the title') self.assertEqual(results[0]['url'], 'http://this.should.be.the.link/') self.assertEqual(results[0]['content'], 'This should be the content.') - self.assertEqual(results[1]['title'], 'Datei:This should.svg') - self.assertEqual(results[1]['url'], 'http://de.wikipedia.org/wiki/Datei:This should.svg') - self.assertEqual(results[1]['img_src'], 'http://ts2.mm.This/should.png') + self.assertEqual(results[1]['title'], 'Datei:This should1.svg') + self.assertEqual(results[1]['url'], 'http://de.wikipedia.org/wiki/Datei:This should1.svg') + self.assertEqual(results[1]['img_src'], 'http://ts2.mm.This/should1.png') self.assertEqual(results[1]['template'], 'images.html') + self.assertEqual(results[2]['title'], 'Datei:This should.svg') + self.assertEqual(results[2]['url'], 'http://de.wikipedia.org/wiki/Datei:This should.svg') + self.assertEqual(results[2]['img_src'], 'http://ts2.mm.This/should.png') + self.assertEqual(results[2]['template'], 'images.html')