From 6b19daf1d28a8f4848bfb537b2cb45bb36dc7ea7 Mon Sep 17 00:00:00 2001
From: Alex Osborne <aosborne@nla.gov.au>
Date: Tue, 17 Aug 2021 09:07:04 +0900
Subject: [PATCH] ExtractorChrome: Don't capture data: URIs

They are already captured as part of their containing document. There's
no need to record them separately.

#430
---
 .../java/org/archive/modules/extractor/ExtractorChrome.java    | 3 ++-
 .../org/archive/modules/extractor/ExtractorChromeTest.java     | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/contrib/src/main/java/org/archive/modules/extractor/ExtractorChrome.java b/contrib/src/main/java/org/archive/modules/extractor/ExtractorChrome.java
index 4d7c3a54..c1b38398 100644
--- a/contrib/src/main/java/org/archive/modules/extractor/ExtractorChrome.java
+++ b/contrib/src/main/java/org/archive/modules/extractor/ExtractorChrome.java
@@ -29,6 +29,7 @@ import org.archive.modules.ProcessorChain;
 import org.archive.net.chrome.*;
 import org.archive.spring.KeyedProperties;
 import org.archive.util.Recorder;
+import org.archive.util.UriUtils;
 import org.json.JSONArray;
 import org.springframework.context.ApplicationEventPublisher;
 
@@ -220,7 +221,7 @@ public class ExtractorChrome extends ContentExtractor {
     }
 
     private void handleCapturedRequest(CrawlURI via, ChromeRequest request) {
-        if (request.isResponseFulfilledByInterception()) {
+        if (request.isResponseFulfilledByInterception() || UriUtils.isDataUri(request.getUrl())) {
             return;
         }
 
diff --git a/contrib/src/test/java/org/archive/modules/extractor/ExtractorChromeTest.java b/contrib/src/test/java/org/archive/modules/extractor/ExtractorChromeTest.java
index 78c8e157..f3a39166 100644
--- a/contrib/src/test/java/org/archive/modules/extractor/ExtractorChromeTest.java
+++ b/contrib/src/test/java/org/archive/modules/extractor/ExtractorChromeTest.java
@@ -67,6 +67,7 @@ public class ExtractorChromeTest {
                         response.setContentType("text/html");
                         response.getWriter().write("<a href=http://example.org/page2.html>link</a>" +
                                 "<img src=/blue.png>" +
+                                "<img src='data:image/gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw=='>" +
                                 "<script>fetch('/post', {method: 'POST', body: 'hello'});</script>" +
                                 "<link rel=stylesheet href=style.css>");
                         baseRequest.setHandled(true);