Handle wget WARC-Target-URI formatting.

- Resolves #514 - Regex replacement for getUrl - Add test - Add test fixture
archivesunleashed · May 11, 2021 · 2d6b2e0 · 2d6b2e0
1 parent bfa7109
commit 2d6b2e0
Show file tree

Hide file tree

Showing 3 changed files with 636 additions and 1 deletion.
diff --git a/src/main/scala/io/archivesunleashed/ArchiveRecord.scala b/src/main/scala/io/archivesunleashed/ArchiveRecord.scala
@@ -155,7 +155,11 @@ class ArchiveRecordImpl(r: SerializableWritable[ArchiveRecordWritable])
     if (r.t.getFormat == ArchiveRecordWritable.ArchiveFormat.ARC) {
       r.t.getRecord.asInstanceOf[ARCRecord].getMetaData.getUrl
     } else {
-      r.t.getRecord.asInstanceOf[WARCRecord].getHeader.getUrl
+      r.t.getRecord
+        .asInstanceOf[WARCRecord]
+        .getHeader
+        .getUrl
+        .replaceAll("<|>", "")
     }
   }