From 7e8ff96fb90ddd1c853418a6c405d97afbc45355 Mon Sep 17 00:00:00 2001
From: Bryan Newbold <bnewbold@archive.org>
Date: Tue, 11 Aug 2020 17:22:10 -0700
Subject: check for simple URL patterns that are usually paywalls or loginwalls

---
 python/sandcrawler/ingest.py | 11 +++++++++++
 python/tests/test_ingest.py  | 18 ++++++++++++++++++
 2 files changed, 29 insertions(+)

(limited to 'python')

diff --git a/python/sandcrawler/ingest.py b/python/sandcrawler/ingest.py
index 1f693dc..918a832 100644
--- a/python/sandcrawler/ingest.py
+++ b/python/sandcrawler/ingest.py
@@ -102,6 +102,11 @@ class IngestFileWorker(SandcrawlerWorker):
             "digital.ucd.ie/",      # ireland national historical
         ]
 
+        self.wall_blocklist = [
+            # loginwall
+            "://profile.thieme.de/HTML/sso/ejournals/login.htm",
+        ]
+
         # these are special-case web domains for which we want SPN2 to not run
         # a headless browser (brozzler), but instead simply run wget.
         # the motivation could be to work around browser issues, or in the
@@ -330,6 +335,12 @@ class IngestFileWorker(SandcrawlerWorker):
                     result['status'] = "skip-url-blocklist"
                     return result
 
+            # check against known loginwall URLs
+            for block in self.wall_blocklist:
+                if block in next_url:
+                    result['status'] = "skip-wall"
+                    return result
+
             # check for popular cookie blocking URL patterns. On successful SPN
             # crawls, shouldn't see these redirect URLs
             if '/cookieAbsent' in next_url or 'cookieSet=1' in next_url:
diff --git a/python/tests/test_ingest.py b/python/tests/test_ingest.py
index c2d6266..46346b7 100644
--- a/python/tests/test_ingest.py
+++ b/python/tests/test_ingest.py
@@ -173,3 +173,21 @@ def test_ingest_blocklist(ingest_worker):
     assert resp['status'] == "skip-url-blocklist"
     assert resp['request'] == request
 
+
+@responses.activate
+def test_ingest_wall_blocklist(ingest_worker):
+
+    ingest_worker.wall_blocklist = [
+        '://test.fatcat.wiki/',
+    ]
+    request = {
+        'ingest_type': 'pdf',
+        'base_url': "https://test.fatcat.wiki/asdfasdf.pdf",
+    }
+
+    resp = ingest_worker.process(request)
+
+    assert resp['hit'] == False
+    assert resp['status'] == "skip-wall"
+    assert resp['request'] == request
+
-- 
cgit v1.2.3