New functionanlity - Selectable browser / ability to add extra browser connections (good for using "scraping browsers"/ etc) (#1943)

pull/1976/head
dgtlmoon 6 months ago committed by GitHub
parent 5a306aa78c
commit 5229094e44
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -30,7 +30,10 @@ jobs:
# Selenium+browserless
docker run --network changedet-network -d --hostname selenium -p 4444:4444 --rm --shm-size="2g" selenium/standalone-chrome:4
docker run --network changedet-network -d --hostname browserless -e "FUNCTION_BUILT_INS=[\"fs\",\"crypto\"]" -e "DEFAULT_LAUNCH_ARGS=[\"--window-size=1920,1080\"]" --rm -p 3000:3000 --shm-size="2g" browserless/chrome:1.60-chrome-stable
docker run --network changedet-network -d --name browserless --hostname browserless -e "FUNCTION_BUILT_INS=[\"fs\",\"crypto\"]" -e "DEFAULT_LAUNCH_ARGS=[\"--window-size=1920,1080\"]" --rm -p 3000:3000 --shm-size="2g" browserless/chrome:1.60-chrome-stable
# For accessing custom browser tests
docker run --network changedet-network -d --name browserless-custom-url --hostname browserless-custom-url -e "FUNCTION_BUILT_INS=[\"fs\",\"crypto\"]" -e "DEFAULT_LAUNCH_ARGS=[\"--window-size=1920,1080\"]" --rm --shm-size="2g" browserless/chrome:1.60-chrome-stable
- name: Build changedetection.io container for testing
run: |
@ -86,6 +89,12 @@ jobs:
# And again with PLAYWRIGHT_DRIVER_URL=..
cd ..
- name: Test custom browser URL
run: |
cd changedetectionio
./run_custom_browser_url_tests.sh
cd ..
- name: Test changedetection.io container starts+runs basically without error
run: |
docker run -p 5556:5000 -d test-changedetectionio

@ -614,6 +614,8 @@ def changedetection_app(config=None, datastore_o=None):
# For the form widget tag uuid lookup
form.tags.datastore = datastore # in _value
for p in datastore.extra_browsers:
form.fetch_backend.choices.append(p)
form.fetch_backend.choices.append(("system", 'System settings default'))
@ -714,7 +716,7 @@ def changedetection_app(config=None, datastore_o=None):
system_uses_webdriver = datastore.data['settings']['application']['fetch_backend'] == 'html_webdriver'
is_html_webdriver = False
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver':
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver' or watch.get('fetch_backend', '').startswith('extra_browser_'):
is_html_webdriver = True
# Only works reliably with Playwright
@ -987,7 +989,7 @@ def changedetection_app(config=None, datastore_o=None):
system_uses_webdriver = datastore.data['settings']['application']['fetch_backend'] == 'html_webdriver'
is_html_webdriver = False
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver':
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver' or watch.get('fetch_backend', '').startswith('extra_browser_'):
is_html_webdriver = True
password_enabled_and_share_is_off = False
@ -1041,7 +1043,7 @@ def changedetection_app(config=None, datastore_o=None):
is_html_webdriver = False
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver':
if (watch.get('fetch_backend') == 'system' and system_uses_webdriver) or watch.get('fetch_backend') == 'html_webdriver' or watch.get('fetch_backend', '').startswith('extra_browser_'):
is_html_webdriver = True
# Never requested successfully, but we detected a fetch error

@ -96,6 +96,7 @@ class Fetcher():
content = None
error = None
fetcher_description = "No description"
browser_connection_url = None
headers = {}
status_code = None
webdriver_js_execute_code = None
@ -251,14 +252,16 @@ class base_html_playwright(Fetcher):
proxy = None
def __init__(self, proxy_override=None):
def __init__(self, proxy_override=None, browser_connection_url=None):
super().__init__()
# .strip('"') is going to save someone a lot of time when they accidently wrap the env value
self.browser_type = os.getenv("PLAYWRIGHT_BROWSER_TYPE", 'chromium').strip('"')
self.command_executor = os.getenv(
"PLAYWRIGHT_DRIVER_URL",
'ws://playwright-chrome:3000'
).strip('"')
# .strip('"') is going to save someone a lot of time when they accidently wrap the env value
if not browser_connection_url:
self.browser_connection_url = os.getenv("PLAYWRIGHT_DRIVER_URL", 'ws://playwright-chrome:3000').strip('"')
else:
self.browser_connection_url = browser_connection_url
# If any proxy settings are enabled, then we should setup the proxy object
proxy_args = {}
@ -444,7 +447,7 @@ class base_html_playwright(Fetcher):
# Seemed to cause a connection Exception even tho I can see it connect
# self.browser = browser_type.connect(self.command_executor, timeout=timeout*1000)
# 60,000 connection timeout only
browser = browser_type.connect_over_cdp(self.command_executor, timeout=60000)
browser = browser_type.connect_over_cdp(self.browser_connection_url, timeout=60000)
# SOCKS5 with authentication is not supported (yet)
# https://github.com/microsoft/playwright/issues/10567
@ -504,7 +507,11 @@ class base_html_playwright(Fetcher):
self.status_code = response.status
if self.status_code != 200 and not ignore_status_codes:
raise Non200ErrorCodeReceived(url=url, status_code=self.status_code)
screenshot=self.page.screenshot(type='jpeg', full_page=True,
quality=int(os.getenv("PLAYWRIGHT_SCREENSHOT_QUALITY", 72)))
raise Non200ErrorCodeReceived(url=url, status_code=self.status_code, screenshot=screenshot)
if len(self.page.content().strip()) == 0:
context.close()
@ -555,8 +562,6 @@ class base_html_webdriver(Fetcher):
else:
fetcher_description = "WebDriver Chrome/Javascript"
command_executor = ''
# Configs for Proxy setup
# In the ENV vars, is prefixed with "webdriver_", so it is for example "webdriver_sslProxy"
selenium_proxy_settings_mappings = ['proxyType', 'ftpProxy', 'httpProxy', 'noProxy',
@ -564,12 +569,15 @@ class base_html_webdriver(Fetcher):
'socksProxy', 'socksVersion', 'socksUsername', 'socksPassword']
proxy = None
def __init__(self, proxy_override=None):
def __init__(self, proxy_override=None, browser_connection_url=None):
super().__init__()
from selenium.webdriver.common.proxy import Proxy as SeleniumProxy
# .strip('"') is going to save someone a lot of time when they accidently wrap the env value
self.command_executor = os.getenv("WEBDRIVER_URL", 'http://browser-chrome:4444/wd/hub').strip('"')
if not browser_connection_url:
self.browser_connection_url = os.getenv("WEBDRIVER_URL", 'http://browser-chrome:4444/wd/hub').strip('"')
else:
self.browser_connection_url = browser_connection_url
# If any proxy settings are enabled, then we should setup the proxy object
proxy_args = {}
@ -611,7 +619,7 @@ class base_html_webdriver(Fetcher):
options.proxy = self.proxy
self.driver = webdriver.Remote(
command_executor=self.command_executor,
command_executor=self.browser_connection_url,
options=options)
try:
@ -666,9 +674,10 @@ class base_html_webdriver(Fetcher):
class html_requests(Fetcher):
fetcher_description = "Basic fast Plaintext/HTTP Client"
def __init__(self, proxy_override=None):
def __init__(self, proxy_override=None, browser_connection_url=None):
super().__init__()
self.proxy_override = proxy_override
# browser_connection_url is none because its always 'launched locally'
def run(self,
url,

@ -168,7 +168,9 @@ class ValidateContentFetcherIsReady(object):
def __call__(self, form, field):
import urllib3.exceptions
from changedetectionio import content_fetcher
return
# AttributeError: module 'changedetectionio.content_fetcher' has no attribute 'extra_browser_unlocked<>ASDF213r123r'
# Better would be a radiohandler that keeps a reference to each class
if field.data is not None and field.data != 'system':
klass = getattr(content_fetcher, field.data)
@ -496,6 +498,12 @@ class SingleExtraProxy(Form):
proxy_url = StringField('Proxy URL', [validators.Optional()], render_kw={"placeholder": "socks5:// or regular proxy http://user:pass@...:3128", "size":50})
# @todo do the validation here instead
class SingleExtraBrowser(Form):
browser_name = StringField('Name', [validators.Optional()], render_kw={"placeholder": "Name"})
browser_connection_url = StringField('Browser connection URL', [validators.Optional()], render_kw={"placeholder": "wss://brightdata... wss://oxylabs etc", "size":50})
# @todo do the validation here instead
# datastore.data['settings']['requests']..
class globalSettingsRequestForm(Form):
time_between_check = FormField(TimeBetweenCheckForm)
@ -504,6 +512,7 @@ class globalSettingsRequestForm(Form):
render_kw={"style": "width: 5em;"},
validators=[validators.NumberRange(min=0, message="Should contain zero or more seconds")])
extra_proxies = FieldList(FormField(SingleExtraProxy), min_entries=5)
extra_browsers = FieldList(FormField(SingleExtraBrowser), min_entries=5)
def validate_extra_proxies(self, extra_validators=None):
for e in self.data['extra_proxies']:

@ -16,6 +16,7 @@ class model(dict):
},
'requests': {
'extra_proxies': [], # Configurable extra proxies via the UI
'extra_browsers': [], # Configurable extra proxies via the UI
'jitter_seconds': 0,
'proxy': None, # Preferred proxy connection
'time_between_check': {'weeks': None, 'days': None, 'hours': 3, 'minutes': None, 'seconds': None},

@ -8,11 +8,12 @@ from distutils.util import strtobool
class difference_detection_processor():
browser_steps = None
datastore = None
fetcher = None
screenshot = None
watch = None
xpath_data = None
browser_steps = None
def __init__(self, *args, datastore, watch_uuid, **kwargs):
super().__init__(*args, **kwargs)
@ -40,6 +41,18 @@ class difference_detection_processor():
if not prefer_fetch_backend or prefer_fetch_backend == 'system':
prefer_fetch_backend = self.datastore.data['settings']['application'].get('fetch_backend')
# In the case that the preferred fetcher was a browser config with custom connection URL..
# @todo - on save watch, if its extra_browser_ then it should be obvious it will use playwright (like if its requests now..)
browser_connection_url = None
if prefer_fetch_backend.startswith('extra_browser_'):
(t, key) = prefer_fetch_backend.split('extra_browser_')
connection = list(
filter(lambda s: (s['browser_name'] == key), self.datastore.data['settings']['requests'].get('extra_browsers', [])))
if connection:
prefer_fetch_backend = 'base_html_playwright'
browser_connection_url = connection[0].get('browser_connection_url')
# Grab the right kind of 'fetcher', (playwright, requests, etc)
if hasattr(content_fetcher, prefer_fetch_backend):
fetcher_obj = getattr(content_fetcher, prefer_fetch_backend)
@ -54,8 +67,9 @@ class difference_detection_processor():
print(f"Using proxy Key: {preferred_proxy_id} as Proxy URL {proxy_url}")
# Now call the fetcher (playwright/requests/etc) with arguments that only a fetcher would need.
# When browser_connection_url is None, it method should default to working out whats the best defaults (os env vars etc)
self.fetcher = fetcher_obj(proxy_override=proxy_url,
#browser_url_extra/configurable browser url=...
browser_connection_url=browser_connection_url
)
if self.watch.has_browser_steps:

@ -0,0 +1,44 @@
#!/bin/bash
# run some tests and look if the 'custom-browser-search-string=1' connect string appeared in the correct containers
# enable debug
set -x
# A extra browser is configured, but we never chose to use it, so it should NOT show in the logs
docker run --rm -e "PLAYWRIGHT_DRIVER_URL=ws://browserless:3000" --network changedet-network test-changedetectionio bash -c 'cd changedetectionio;pytest tests/custom_browser_url/test_custom_browser_url.py::test_request_not_via_custom_browser_url'
docker logs browserless-custom-url &>log.txt
grep 'custom-browser-search-string=1' log.txt
if [ $? -ne 1 ]
then
echo "Saw a request in 'browserless-custom-url' container with 'custom-browser-search-string=1' when I should not"
exit 1
fi
docker logs browserless &>log.txt
grep 'custom-browser-search-string=1' log.txt
if [ $? -ne 1 ]
then
echo "Saw a request in 'browser' container with 'custom-browser-search-string=1' when I should not"
exit 1
fi
# Special connect string should appear in the custom-url container, but not in the 'default' one
docker run --rm -e "PLAYWRIGHT_DRIVER_URL=ws://browserless:3000" --network changedet-network test-changedetectionio bash -c 'cd changedetectionio;pytest tests/custom_browser_url/test_custom_browser_url.py::test_request_via_custom_browser_url'
docker logs browserless-custom-url &>log.txt
grep 'custom-browser-search-string=1' log.txt
if [ $? -ne 0 ]
then
echo "Did not see request in 'browserless-custom-url' container with 'custom-browser-search-string=1' when I should"
exit 1
fi
docker logs browserless &>log.txt
grep 'custom-browser-search-string=1' log.txt
if [ $? -ne 1 ]
then
echo "Saw a request in 'browser' container with 'custom-browser-search-string=1' when I should not"
exit 1
fi

@ -0,0 +1,24 @@
ul#requests-extra_browsers {
list-style: none;
/* tidy up the table to look more "inline" */
li {
> label {
display: none;
}
}
/* each proxy entry is a `table` */
table {
tr {
display: inline;
}
}
}
#extra-browsers-setting {
border: 1px solid var(--color-grey-800);
border-radius: 4px;
margin: 1em;
padding: 1em;
}

@ -60,3 +60,10 @@ body.proxy-check-active {
padding-bottom: 1em;
}
#extra-proxies-setting {
border: 1px solid var(--color-grey-800);
border-radius: 4px;
margin: 1em;
padding: 1em;
}

@ -5,6 +5,7 @@
@import "parts/_arrows";
@import "parts/_browser-steps";
@import "parts/_extra_proxies";
@import "parts/_extra_browsers";
@import "parts/_pagination";
@import "parts/_spinners";
@import "parts/_variables";

@ -128,6 +128,27 @@ body.proxy-check-active #request .proxy-timing {
border-radius: 4px;
padding: 1em; }
#extra-proxies-setting {
border: 1px solid var(--color-grey-800);
border-radius: 4px;
margin: 1em;
padding: 1em; }
ul#requests-extra_browsers {
list-style: none;
/* tidy up the table to look more "inline" */
/* each proxy entry is a `table` */ }
ul#requests-extra_browsers li > label {
display: none; }
ul#requests-extra_browsers table tr {
display: inline; }
#extra-browsers-setting {
border: 1px solid var(--color-grey-800);
border-radius: 4px;
margin: 1em;
padding: 1em; }
.pagination-page-info {
color: #fff;
font-size: 0.85rem;

@ -633,6 +633,18 @@ class ChangeDetectionStore:
return {}
@property
def extra_browsers(self):
res = []
p = list(filter(
lambda s: (s.get('browser_name') and s.get('browser_connection_url')),
self.__data['settings']['requests'].get('extra_browsers', [])))
if p:
for i in p:
res.append(("extra_browser_"+i['browser_name'], i['browser_name']))
return res
def tag_exists_by_name(self, tag_name):
return any(v.get('title', '').lower() == tag_name.lower() for k, v in self.__data['settings']['application']['tags'].items())

@ -230,11 +230,15 @@ nav
</p>
<p><strong>Tip</strong>: "Residential" and "Mobile" proxy type can be more successfull than "Data Center" for blocked websites.
<div class="pure-control-group">
<div class="pure-control-group" id="extra-proxies-setting">
{{ render_field(form.requests.form.extra_proxies) }}
<span class="pure-form-message-inline">"Name" will be used for selecting the proxy in the Watch Edit settings</span><br>
<span class="pure-form-message-inline">SOCKS5 proxies with authentication are only supported with 'plain requests' fetcher, for other fetchers you should whitelist the IP access instead</span>
</div>
<div class="pure-control-group" id="extra-browsers-setting">
<span class="pure-form-message-inline"><i>Extra Browsers</i> allow changedetection.io to communicate with a different web-browser.</span><br>
{{ render_field(form.requests.form.extra_browsers) }}
</div>
</div>
<div id="actions">
<div class="pure-control-group">

@ -104,8 +104,9 @@
{% if watch.get_fetch_backend == "html_webdriver"
or ( watch.get_fetch_backend == "system" and system_default_fetcher == 'html_webdriver' )
or "extra_browser_" in watch.get_fetch_backend
%}
<img class="status-icon" src="{{url_for('static_content', group='images', filename='Google-Chrome-icon.png')}}" title="Using a chrome browser" >
<img class="status-icon" src="{{url_for('static_content', group='images', filename='Google-Chrome-icon.png')}}" title="Using a Chrome browser" >
{% endif %}
{%if watch.is_pdf %}<img class="status-icon" src="{{url_for('static_content', group='images', filename='pdf-icon.svg')}}" title="Converting PDF to text" >{% endif %}

@ -0,0 +1,89 @@
# !/usr/bin/python3
import os
from flask import url_for
from ..util import live_server_setup, wait_for_all_checks
def do_test(client, live_server, make_test_use_extra_browser=False):
# Grep for this string in the logs?
test_url = f"https://changedetection.io/ci-test.html"
custom_browser_name = 'custom browser URL'
# needs to be set and something like 'ws://127.0.0.1:3000?stealth=1&--disable-web-security=true'
assert os.getenv('PLAYWRIGHT_DRIVER_URL'), "Needs PLAYWRIGHT_DRIVER_URL set for this test"
#####################
res = client.post(
url_for("settings_page"),
data={"application-empty_pages_are_a_change": "",
"requests-time_between_check-minutes": 180,
'application-fetch_backend': "html_webdriver",
# browserless-custom-url is setup in .github/workflows/test-only.yml
# the test script run_custom_browser_url_test.sh will look for 'custom-browser-search-string' in the container logs
'requests-extra_browsers-0-browser_connection_url': 'ws://browserless-custom-url:3000?stealth=1&--disable-web-security=true&custom-browser-search-string=1',
'requests-extra_browsers-0-browser_name': custom_browser_name
},
follow_redirects=True
)
assert b"Settings updated." in res.data
# Add our URL to the import page
res = client.post(
url_for("import_page"),
data={"urls": test_url},
follow_redirects=True
)
assert b"1 Imported" in res.data
wait_for_all_checks(client)
if make_test_use_extra_browser:
# So the name should appear in the edit page under "Request" > "Fetch Method"
res = client.get(
url_for("edit_page", uuid="first"),
follow_redirects=True
)
assert b'custom browser URL' in res.data
res = client.post(
url_for("edit_page", uuid="first"),
data={
"url": test_url,
"tags": "",
"headers": "",
'fetch_backend': f"extra_browser_{custom_browser_name}",
'webdriver_js_execute_code': ''
},
follow_redirects=True
)
assert b"Updated watch." in res.data
wait_for_all_checks(client)
# Force recheck
res = client.get(url_for("form_watch_checknow"), follow_redirects=True)
assert b'1 watches queued for rechecking.' in res.data
wait_for_all_checks(client)
res = client.get(
url_for("preview_page", uuid="first"),
follow_redirects=True
)
assert b'cool it works' in res.data
# Requires playwright to be installed
def test_request_via_custom_browser_url(client, live_server):
live_server_setup(live_server)
# We do this so we can grep the logs of the custom container and see if the request actually went through that container
do_test(client, live_server, make_test_use_extra_browser=True)
def test_request_not_via_custom_browser_url(client, live_server):
live_server_setup(live_server)
# We do this so we can grep the logs of the custom container and see if the request actually went through that container
do_test(client, live_server, make_test_use_extra_browser=False)
Loading…
Cancel
Save