[Archive-access-cvs] SF.net SVN: archive-access:[2639] trunk/archive-access/projects/wayback/ wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer

[<bra...@us...>]

Revision: 2639
          http://archive-access.svn.sourceforge.net/archive-access/?rev=2639&view=rev
Author:   bradtofel
Date:     2008-11-07 22:38:48 +0000 (Fri, 07 Nov 2008)

Log Message:
-----------
FEATURE: added -all option to warc-indexer command line tool, causing the tool to output records for request and metadata records as well as duplicate, capture, and dns records.

Modified Paths:
--------------
    trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WARCRecordToSearchResultAdapter.java
    trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WarcIndexer.java

Modified: trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WARCRecordToSearchResultAdapter.java
===================================================================
--- trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WARCRecordToSearchResultAdapter.java	2008-11-07 22:35:24 UTC (rev 2638)
+++ trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WARCRecordToSearchResultAdapter.java	2008-11-07 22:38:48 UTC (rev 2639)
@@ -2,7 +2,7 @@
 
 import java.io.File;
 import java.io.IOException;
-//import java.util.logging.Logger;
+import java.util.logging.Logger;
 
 import org.apache.commons.httpclient.Header;
 import org.apache.commons.httpclient.HttpParser;
@@ -33,14 +33,23 @@
  */
 public class WARCRecordToSearchResultAdapter
 implements Adapter<WARCRecord,CaptureSearchResult>{
+	private static final Logger LOGGER =
+        Logger.getLogger(WARCRecordToSearchResultAdapter.class.getName());
 	
 	private final static String DEFAULT_VALUE = "-"; 
 
-//	private static final Logger LOGGER = Logger.getLogger(
-//			WARCRecordToSearchResultAdapter.class.getName());
-
 	private UrlCanonicalizer canonicalizer = null;
+	
+	private boolean processAll = false;
 
+	public boolean isProcessAll() {
+		return processAll;
+	}
+
+	public void setProcessAll(boolean processAll) {
+		this.processAll = processAll;
+	}
+
 	public WARCRecordToSearchResultAdapter() {
 		canonicalizer = new AggressiveUrlCanonicalizer();
 	}
@@ -75,12 +84,19 @@
 		return output.toString();
 	}
 	
-	private static String transformHTTPMime(final String input) {
+	private static String escapeSpaces(final String input) {
+		if(input.contains(" ")) {
+			return input.replace(" ", "%20");
+		}
+		return input;
+	}
+	
+	private static String transformHTTPMime(String input) {
 		int semiIdx = input.indexOf(";");
 		if(semiIdx > 0) {
-			return input.substring(0,semiIdx).trim();
+			return escapeSpaces(input.substring(0,semiIdx).trim());
 		}
-		return input.trim();
+		return escapeSpaces(input.trim());
 	}
 
 	private String transformWarcFilename(String readerIdentifier) {
@@ -148,16 +164,21 @@
 		return result;
 	}
 
-	private CaptureSearchResult adaptRevisit(ArchiveRecordHeader header, WARCRecord rec) 
+	private CaptureSearchResult adaptGeneric(ArchiveRecordHeader header,
+			WARCRecord rec, String mime) 
 	throws IOException {
 
 		CaptureSearchResult result = getBlankSearchResult();
 
 		result.setCaptureTimestamp(transformDate(header.getDate()));
+		result.setFile(transformWarcFilename(header.getReaderIdentifier()));
+		result.setOffset(header.getOffset());
 		result.setDigest(transformDigest(header.getHeaderValue(
-						WARCRecord.HEADER_KEY_PAYLOAD_DIGEST)));
+				WARCRecord.HEADER_KEY_PAYLOAD_DIGEST)));
 		
 		addUrlDataToSearchResult(result,header.getUrl());
+		
+		result.setMimeType(mime);
 
 		return result;
 	}
@@ -243,7 +264,7 @@
 		}
 		return result;
 	}
-	
+
 	private CaptureSearchResult adaptInner(WARCRecord rec) throws IOException {
 		
 		CaptureSearchResult result = null;
@@ -257,7 +278,17 @@
 				result = adaptResponse(header,rec);
 			}
 		} else if(type.equals(WARCConstants.REVISIT)) {
-			result = adaptRevisit(header,rec);
+			result = adaptGeneric(header,rec,"warc/revisit");
+		} else if(type.equals(WARCConstants.REQUEST)) {
+			if(processAll) {
+				result = adaptGeneric(header,rec,"warc/request");
+			}
+		} else if(type.equals(WARCConstants.METADATA)) {
+			if(processAll) {
+				result = adaptGeneric(header,rec,"warc/metadata");
+			}
+		} else {
+			LOGGER.info("Skipping record type : " + type);
 		}
 
 		return result;

Modified: trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WarcIndexer.java
===================================================================
--- trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WarcIndexer.java	2008-11-07 22:35:24 UTC (rev 2638)
+++ trunk/archive-access/projects/wayback/wayback-core/src/main/java/org/archive/wayback/resourcestore/indexer/WarcIndexer.java	2008-11-07 22:38:48 UTC (rev 2639)
@@ -26,9 +26,19 @@
 	public final static String CDX_HEADER_MAGIC = " CDX N b h m s k r V g";
 
 	private UrlCanonicalizer canonicalizer = null;
+	private boolean processAll = false;
 	public WarcIndexer() {
 		canonicalizer = new AggressiveUrlCanonicalizer();
 	}
+
+	public boolean isProcessAll() {
+		return processAll;
+	}
+
+	public void setProcessAll(boolean processAll) {
+		this.processAll = processAll;
+	}
+
 	
 	/**
 	 * @param warc
@@ -61,6 +71,7 @@
 		WARCRecordToSearchResultAdapter adapter2 = 
 			new WARCRecordToSearchResultAdapter();
 		adapter2.setCanonicalizer(canonicalizer);
+		adapter2.setProcessAll(processAll);
 
 		ArchiveReaderCloseableIterator itr1 = 
 			new ArchiveReaderCloseableIterator(reader,reader.iterator());
@@ -82,11 +93,12 @@
 	private static void USAGE() {
 		System.err.println("USAGE:");
 		System.err.println("");
-		System.err.println("warc-indexer [-identity] WARCFILE");
-		System.err.println("warc-indexer [-identity] WARCFILE CDXFILE");
+		System.err.println("warc-indexer [-identity] [-all] WARCFILE");
+		System.err.println("warc-indexer [-identity] [-all] WARCFILE CDXFILE");
 		System.err.println("");
 		System.err.println("Create a CDX format index at CDXFILE or to STDOUT");
 		System.err.println("With -identity, perform no url canonicalization.");
+		System.err.println("With -all, output request and metadata records.");
 		System.exit(1);
 	}
 
@@ -96,8 +108,14 @@
 	public static void main(String[] args) {
 		WarcIndexer indexer = new WarcIndexer();
 		int idx = 0;
-		if(args[0] != null && args[0].equals("-identity")) {
-			indexer.setCanonicalizer(new IdentityUrlCanonicalizer());
+		while(args[idx] != null) {
+			if(args[idx].equals("-identity")) {
+				indexer.setCanonicalizer(new IdentityUrlCanonicalizer());
+			} else if(args[idx].equals("-all")) {
+				indexer.setProcessAll(true);
+			} else {
+				break;
+			}
 			idx++;
 		}
 		File arc = new File(args[idx]);


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.