diff --git a/test.py b/test.py deleted file mode 100755 index 5b4a630..0000000 --- a/test.py +++ /dev/null @@ -1,99 +0,0 @@ -#! /usr/bin/env python3 -# -*- coding: utf-8 -*- - -from urllib.parse import urlparse, parse_qsl, urlencode, urlunparse -import requests - -def deredir_url(url): - """ - Given a URL, return the URL that the page really downloads from - :param url: url to be de-redirected - :return: direct url - """ - - ret = None - try: - # Download the page - ret = requests.get(url, timeout=5) - except: - # If anything goes wrong keep the URL intact - return url - - # Return the URL that the page was downloaded from - return ret.url - -def _remove_tracker_params(query_str): - """ - private function - Given a query string from a URL, strip out the known trackers - :param query_str: query to be cleaned - :return: query cleaned - """ - # Avalaible URL tracking parameters : - # UTM tags by Google Ads, M$ Ads, ... - # tag by TikTok - # tags by Snapchat - # tags by Facebook - params_to_remove = [ - "utm_source", "utm_medium", "utm_campaign", "utm_term", "utm_content", - "mkt_tok", - "campaign_name", "ad_set_name", "campaign_id", "ad_set_id", - "media", "interest_group_name", - "xtor" - ] - query_to_clean = dict(parse_qsl(query_str, keep_blank_values=True)) - query_cleaned = [(k, v) for k, v in query_to_clean.items() if not k in params_to_remove] - return urlencode(query_cleaned, safe='#', doseq=True) - - -def _remove_trackers_fragment(fragment_str): - """ - private function - Given a fragment string from a URL, strip out the known trackers - :param query_str: fragment to be cleaned - :return: cleaned fragment - """ - - # Not implemented - # Unclear what, if anything, can be done - # Need better understanding of fragment-based tracking - # https://builtvisible.com/one-weird-trick-to-avoid-utm-parameters/ - - return fragment_str - - -def clean_url(dirty_url): - """ - Given a URL, return it with the UTM parameters removed from query and fragment - :param dirty_url: url to be cleaned - :return: url cleaned - >>> clean_url('https://example.com/video/this-aerial-ropeway?utm_source=Twitter&utm_medium=video&utm_campaign=organic&utm_content=Nov13&a=aaa&b=1#mkt_tok=tik&mkt_tik=tok') - 'https://example.com/video/this-aerial-ropeway?a=aaa&b=1#mkt_tik=tok' - """ - - url_parsed = urlparse(dirty_url, allow_fragments=False) - - cleaned_url = urlunparse([ - url_parsed.scheme, - url_parsed.netloc, - url_parsed.path, - url_parsed.params, - _remove_tracker_params(url_parsed.query), - _remove_trackers_fragment(url_parsed.fragment) - ]) - - return cleaned_url - -def main(): - # url = 'https://example.com/video/this-aerial-ropeway?utm_source=Twitter&utm_medium=video&utm_campaign=organic&utm_content=Nov13&a=aaa&b=1#mkt_tok=tik&mkt_tik=tok' - # url = "https://docs.helix-editor.com/keymap.html#movement" - # url = "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7071508/#sec1-nutrients-12-00530title" - # url = "https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title42-section12208&num=0&edition=prelim" - url = "https://shorturl.at/qwP38" - print('Orig: ' + url) - direct_url = deredir_url(url) - print('dir : ' + direct_url) - print('to : ' + clean_url(direct_url)) - -if __name__=="__main__": - main()