From 90a623bd902e9abd9ce700be99d71ff7be9d00a1 Mon Sep 17 00:00:00 2001
From: Bryan Newbold <bnewbold@archive.org>
Date: Mon, 24 Aug 2020 12:02:27 -0700
Subject: html: handle embed with mangled 'src' attribute

---
 python/sandcrawler/html.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'python')

diff --git a/python/sandcrawler/html.py b/python/sandcrawler/html.py
index 6236a3b..acf1522 100644
--- a/python/sandcrawler/html.py
+++ b/python/sandcrawler/html.py
@@ -105,7 +105,7 @@ def extract_fulltext_url(html_url, html_body):
     # http://www.jasstudies.com/DergiTamDetay.aspx?ID=3401
     # <embed src="/files/jass_makaleler/1359848334_33-Okt.%20Yasemin%20KARADEM%C4%B0R.pdf" type="application/pdf" />
     embed = soup.find('embed', attrs={"type": "application/pdf"})
-    if embed:
+    if embed and embed.get('src'):
         url = embed['src'].strip()
         if url.startswith('/'):
             url = host_prefix+url
-- 
cgit v1.2.3