From 2c8e5a3a36d45c14c285d8d010c22ab34eccc932 Mon Sep 17 00:00:00 2001
From: Kovid Goyal <kovid@kovidgoyal.net>
Date: Thu, 24 Mar 2022 17:35:48 +0530
Subject: [PATCH] Start working on using Chromiums network stack to download
 HTML

A bunch of websites have started using TLS fingerprinting to deny
access. Bloody morons.
---
 src/calibre/__init__.py         | 10 ++-----
 src/calibre/scraper/__init__.py |  0
 src/calibre/scraper/simple.py   | 50 +++++++++++++++++++++++++++++++++
 src/calibre/utils/random_ua.py  | 18 ++++++++++++
 4 files changed, 71 insertions(+), 7 deletions(-)
 create mode 100644 src/calibre/scraper/__init__.py
 create mode 100644 src/calibre/scraper/simple.py

diff --git a/src/calibre/__init__.py b/src/calibre/__init__.py
index e03ada8cc5..bea87b67ef 100644
--- a/src/calibre/__init__.py
+++ b/src/calibre/__init__.py
@@ -3,7 +3,7 @@ __license__   = 'GPL v3'
 __copyright__ = '2008, Kovid Goyal <kovid@kovidgoyal.net>'
 __docformat__ = 'restructuredtext en'
 
-import sys, os, re, time, random, warnings
+import sys, os, re, time, warnings
 from polyglot.builtins import codepoint_to_chr, hasenv, native_string_type
 from math import floor
 from functools import partial
@@ -290,18 +290,14 @@ def is_mobile_ua(ua):
 
 
 def random_user_agent(choose=None, allow_ie=True):
-    from calibre.utils.random_ua import common_user_agents, user_agents_popularity_map
+    from calibre.utils.random_ua import common_user_agents, choose_randomly_by_popularity
     ua_list = common_user_agents()
     ua_list = tuple(x for x in ua_list if not is_mobile_ua(x))
     if not allow_ie:
         ua_list = tuple(x for x in ua_list if 'Trident/' not in x)
     if choose is not None:
         return ua_list[choose]
-    pm = user_agents_popularity_map()
-    weights = None
-    if pm:
-        weights = tuple(map(pm.__getitem__, ua_list))
-    return random.choices(ua_list, weights=weights)[0]
+    return choose_randomly_by_popularity(ua_list)
 
 
 def browser(honor_time=True, max_time=2, user_agent=None, verify_ssl_certificates=True, handle_refresh=True, **kw):
diff --git a/src/calibre/scraper/__init__.py b/src/calibre/scraper/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/src/calibre/scraper/simple.py b/src/calibre/scraper/simple.py
new file mode 100644
index 0000000000..f45ba8d89e
--- /dev/null
+++ b/src/calibre/scraper/simple.py
@@ -0,0 +1,50 @@
+#!/usr/bin/env python
+# vim:fileencoding=utf-8
+# License: GPL v3 Copyright: 2022, Kovid Goyal <kovid at kovidgoyal.net>
+
+
+import os
+from functools import lru_cache
+from qt.core import QApplication
+from qt.webengine import QWebEnginePage, QWebEngineProfile, QWebEngineSettings
+
+from calibre.constants import cache_dir
+
+
+@lru_cache(maxsize=4)
+def create_profile(cache_name='simple', allow_js=False):
+    from calibre.utils.random_ua import random_common_chrome_user_agent
+    ans = QWebEngineProfile(cache_name, QApplication.instance())
+    ans.setHttpUserAgent(random_common_chrome_user_agent())
+    ans.setHttpCacheMaximumSize(0)  # managed by webengine
+    ans.setCachePath(os.path.join(cache_dir(), 'scraper', cache_name))
+    s = ans.settings()
+    a = s.setAttribute
+    a(QWebEngineSettings.WebAttribute.PluginsEnabled, False)
+    a(QWebEngineSettings.WebAttribute.JavascriptEnabled, allow_js)
+    s.setUnknownUrlSchemePolicy(QWebEngineSettings.UnknownUrlSchemePolicy.DisallowUnknownUrlSchemes)
+    a(QWebEngineSettings.WebAttribute.JavascriptCanOpenWindows, False)
+    a(QWebEngineSettings.WebAttribute.JavascriptCanAccessClipboard, False)
+    # ensure javascript cannot read from local files
+    a(QWebEngineSettings.WebAttribute.LocalContentCanAccessFileUrls, False)
+    a(QWebEngineSettings.WebAttribute.AllowWindowActivationFromJavaScript, False)
+    return s
+
+
+class SimpleScraper(QWebEnginePage):
+
+    def __init__(self, source, parent=None):
+        super().__init__(create_profile(source), parent=parent)
+        self.setAudioMuted(True)
+
+    def javaScriptAlert(self, url, msg):
+        pass
+
+    def javaScriptConfirm(self, url, msg):
+        return True
+
+    def javaScriptPrompt(self, url, msg, defval):
+        return True, defval
+
+    def javaScriptConsoleMessage(self, level, message, line_num, source_id):
+        pass
diff --git a/src/calibre/utils/random_ua.py b/src/calibre/utils/random_ua.py
index d6e76bcaf1..96f5bea650 100644
--- a/src/calibre/utils/random_ua.py
+++ b/src/calibre/utils/random_ua.py
@@ -25,6 +25,24 @@ def common_user_agents():
     return user_agent_data()['common_user_agents']
 
 
+def common_chrome_user_agents():
+    for x in user_agent_data()['common_user_agents']:
+        if 'Chrome/' in x:
+            yield x
+
+
+def choose_randomly_by_popularity(ua_list):
+    pm = user_agents_popularity_map()
+    weights = None
+    if pm:
+        weights = tuple(map(pm.__getitem__, ua_list))
+    return random.choices(ua_list, weights=weights)[0]
+
+
+def random_common_chrome_user_agent():
+    return choose_randomly_by_popularity(tuple(common_chrome_user_agents()))
+
+
 def user_agents_popularity_map():
     return user_agent_data().get('user_agents_popularity', {})