Add SOCKS Proxy support

.github/workflows/tests.yml

@@ -19,7 +19,11 @@ jobs:
         os: [ubuntu-latest]
         # https://github.com/actions/python-versions/blob/main/versions-manifest.json
         python-version: ["3.9", "3.11"]  # "3.13-dev"
-        env: [{ MINIMAL: "true" }, { MINIMAL: "false" }]
+        env:
+          - MINIMAL: "true"
+            PROXY_TEST: "false"
+          - MINIMAL: "false"
+            PROXY_TEST: "true"
         include:
           # custom python versions
           - os: ubuntu-20.04
@@ -36,6 +40,19 @@ jobs:
             python-version: "3.10"
           - os: ubuntu-latest
             python-version: "3.12"
+    services:
+      socks_proxy:
+        image: ${{ matrix.os == 'ubuntu-latest' && 'serjs/go-socks5-proxy' || '' }}
+        ports:
+          - 1080:1080
+      socks_proxy_auth:
+        image: ${{ matrix.os == 'ubuntu-latest' && 'serjs/go-socks5-proxy' || '' }}
+        env:
+          PROXY_USER: user
+          PROXY_PASSWORD: pass
+        ports:
+          - 1081:1080
+
     steps:
     # Python and pip setup
     - name: Set up Python ${{ matrix.python-version }}
@@ -97,6 +114,8 @@ jobs:
       run: |
         python -m pip install pytest pytest-cov
         pytest --cov=./ --cov-report=xml
+      env:
+        PROXY_TEST: ${{ matrix.env.PROXY_TEST }}
 
     # coverage
     - name: Upload coverage to Codecov

compose.yml

@@ -0,0 +1,18 @@
+services:
+  socks_proxy:
+    image: serjs/go-socks5-proxy
+    ports:
+      - 1080:1080
+  socks_proxy_auth:
+    image: serjs/go-socks5-proxy
+    ports:
+      - 1081:1080
+    environment:
+      PROXY_USER: user
+      PROXY_PASSWORD: pass
+#  tor_proxy:
+#    image: dperson/torproxy
+#    ports:
+#      - 9050:9050
+
+

setup.py

@@ -31,6 +31,7 @@ def get_long_description():
         "htmldate[speed] >= 1.8.1",
         "py3langid >= 0.2.2",
         "pycurl >= 7.45.3",
+        "urllib3[socks]",
         "zstandard >= 0.20.0",
     ],
     "gui": [

tests/downloads_tests.py

@@ -43,6 +43,7 @@
 from trafilatura.settings import DEFAULT_CONFIG, args_to_extractor, use_config
 from trafilatura.utils import decode_file, decode_response, handle_compressed_file, load_html
 
+
 logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
 
 ZERO_CONFIG = DEFAULT_CONFIG
@@ -59,6 +60,7 @@ def _reset_downloads_global_objects():
     """
     Force global objects to be re-created
     """
+    trafilatura.downloads.PROXY_URL = None
     trafilatura.downloads.HTTP_POOL = None
     trafilatura.downloads.NO_CERT_POOL = None
     trafilatura.downloads.RETRY_STRATEGY = None
@@ -154,6 +156,39 @@ def test_fetch():
     _reset_downloads_global_objects()
 
 
+IS_PROXY_TEST = os.environ.get("PROXY_TEST", "false") == "true"
+
+PROXY_URLS = (
+    ("socks5://localhost:1080", True),
+    ("socks5://user:pass@localhost:1081", True),
+    ("socks5://localhost:10/", False),
+    ("bogus://localhost:1080", False),
+)
+
+
+def proxied(f):
+    "Run the download using a potentially malformed proxy address."
+    for proxy_url, is_working in PROXY_URLS:
+        _reset_downloads_global_objects()
+        trafilatura.downloads.PROXY_URL = proxy_url
+        if is_working:
+            f()
+        else:
+            with pytest.raises(AssertionError):
+                f()
+    _reset_downloads_global_objects()
+
+
+@pytest.mark.skipif(not IS_PROXY_TEST, reason="proxy tests disabled")
+def test_proxied_is_live_page():
+    proxied(test_is_live_page)
+
+
+@pytest.mark.skipif(not IS_PROXY_TEST, reason="proxy tests disabled")
+def test_proxied_fetch():
+    proxied(test_fetch)
+
+
 def test_config():
     '''Test how configuration options are read and stored.'''
     # default config is none
@@ -241,6 +276,8 @@ def test_queue():
     test_response_object()
     test_is_live_page()
     test_fetch()
+    test_proxied_is_live_page()
+    test_proxied_fetch()
     test_config()
     test_decode()
     test_queue()

trafilatura/downloads.py

@@ -4,6 +4,7 @@
 """
 
 import logging
+import os
 import random
 
 from concurrent.futures import ThreadPoolExecutor, as_completed
@@ -16,9 +17,21 @@
 import certifi
 import urllib3
 
+from courlan import UrlStore
+from courlan.network import redirection_test
+
+from .settings import DEFAULT_CONFIG, Extractor
+from .utils import URL_BLACKLIST_REGEX, decode_file, is_acceptable_length, make_chunks
+
+
 try:
-    import pycurl
+    from urllib3.contrib.socks import SOCKSProxyManager
+    PROXY_URL = os.environ.get("http_proxy")
+except ImportError:
+    PROXY_URL = None
 
+try:
+    import pycurl
     CURL_SHARE = pycurl.CurlShare()
     # available options:
     # https://curl.se/libcurl/c/curl_share_setopt.html
@@ -30,27 +43,28 @@
 except ImportError:
     HAS_PYCURL = False
 
-from courlan import UrlStore
-from courlan.network import redirection_test
-
 try:  # Python 3.8+
     from importlib.metadata import version
 except ImportError:
     from importlib_metadata import version
 
-from .settings import DEFAULT_CONFIG, Extractor
-from .utils import URL_BLACKLIST_REGEX, decode_file, is_acceptable_length, make_chunks
-
 
 LOGGER = logging.getLogger(__name__)
 
-NUM_CONNECTIONS = 50
-
 urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
 HTTP_POOL = None
 NO_CERT_POOL = None
 RETRY_STRATEGY = None
 
+
+def create_pool(**args):
+    "Configure urllib3 download pool according to user-defined settings."
+    manager_class = SOCKSProxyManager if PROXY_URL else urllib3.PoolManager
+    manager_args = {"proxy_url": PROXY_URL} if PROXY_URL else {}
+    manager_args["num_pools"] = 50
+    return manager_class(**manager_args, **args)
+
+
 DEFAULT_HEADERS = urllib3.util.make_headers(accept_encoding=True)
 USER_AGENT = (
     "trafilatura/" + version("trafilatura") + " (+https://github.com/adbar/trafilatura)"
@@ -161,20 +175,18 @@ def _send_urllib_request(
     try:
         if no_ssl is False:
             if not HTTP_POOL:
-                HTTP_POOL = urllib3.PoolManager(
+                HTTP_POOL = create_pool(
                     retries=RETRY_STRATEGY,
                     timeout=config.getint("DEFAULT", "DOWNLOAD_TIMEOUT"),
-                    ca_certs=certifi.where(),
-                    num_pools=NUM_CONNECTIONS,
+                    ca_certs=certifi.where()
                 )  # cert_reqs='CERT_REQUIRED'
             pool_manager = HTTP_POOL
         else:
             if not NO_CERT_POOL:
-                NO_CERT_POOL = urllib3.PoolManager(
+                NO_CERT_POOL = create_pool(
                     retries=RETRY_STRATEGY,
                     timeout=config.getint("DEFAULT", "DOWNLOAD_TIMEOUT"),
-                    cert_reqs="CERT_NONE",
-                    num_pools=NUM_CONNECTIONS,
+                    cert_reqs="CERT_NONE"
                 )
             pool_manager = NO_CERT_POOL
         # execute request
@@ -288,6 +300,8 @@ def _pycurl_is_live_page(url: str) -> bool:
     curl.setopt(pycurl.SSL_VERIFYHOST, 0)
     # Set option to avoid getting the response body
     curl.setopt(curl.NOBODY, True)
+    if PROXY_URL:
+        curl.setopt(pycurl.PRE_PROXY, PROXY_URL)
     # Perform the request
     try:
         curl.perform()
@@ -410,6 +424,9 @@ def _send_pycurl_request(
         headerbytes = BytesIO()
         curl.setopt(pycurl.HEADERFUNCTION, headerbytes.write)
 
+    if PROXY_URL:
+        curl.setopt(pycurl.PRE_PROXY, PROXY_URL)
+
     # TCP_FASTOPEN
     # curl.setopt(pycurl.FAILONERROR, 1)
     # curl.setopt(pycurl.ACCEPT_ENCODING, '')