mirror of
https://github.com/internetarchive/heritrix3.git
synced 2026-08-13 10:11:10 +00:00
Count url-agnostic content digest duplicates as duplicates in statistics
* CrawledBytesHistotable.java
accumulate() - count urls with warcRevisit:uriAgnosticDigest annotation as duplicates
* ContentDigestHistoryTest.java
testWarcDedupe() - check for presence of warcRevisit:uriAgnosticDigest annotation
This commit is contained in:
@@ -249,6 +249,7 @@ public class ContentDigestHistoryTest extends TmpDirTestCase {
|
||||
assertEquals(1, curi2.getContentDigestHistory().get(A_CONTENT_DIGEST_COUNT));
|
||||
|
||||
warcWriter.process(curi2);
|
||||
assertTrue(curi2.getAnnotations().contains("warcRevisit:uriAgnosticDigest"));
|
||||
assertEquals(curi1.getUURI().toString(), curi2.getContentDigestHistory().get(A_ORIGINAL_URL));
|
||||
assertNotSame(curi2.getUURI().toString(), curi2.getContentDigestHistory().get(A_ORIGINAL_URL));
|
||||
assertEquals(2, curi2.getContentDigestHistory().get(A_CONTENT_DIGEST_COUNT));
|
||||
|
||||
Reference in New Issue
Block a user