Changeset 6102

Timestamp:

2003-12-03T09:39:47+13:00 (20 years ago)

Author:

cs025

Message:

Added MetaXMLExtractor (at least temporarily), improved other extractors

Location:

trunk/gsdl3/src/java/org/greenstone/gsdl3/gs3build/extractor

Files:

• GMLExtractor.java (modified) (4 diffs)
• IndexExtractor.java (modified) (7 diffs)
• MetaXMLExtractor.java (added)

trunk/gsdl3/src/java/org/greenstone/gsdl3/gs3build/extractor/GMLExtractor.java

@@ -17 +17 @@
 public class GMLExtractor implements ExtractorInterface
 {
+  public static final String ACCUMULATE_MODE = "accumulate";
+
   /**
    *  An inner class to handle GML files
    */
   class GMLHandler extends DefaultHandler
-  { String file;
+  { List   files;
     String label;
     StringBuffer value;
     boolean inElement;
+    boolean accumulate;
 
     GMLHandler()
@@ -35 +38 @@
 
     public void startElement(String URI, String localName, String qName, Attributes attributes)
-    { if (localName.equals("Filename"))
+    { if (localName.equals("FileName"))
       { this.value = new StringBuffer();
+      }
+      else if (localName.equals("FileSet"))
+      { this.files = new ArrayList();
+      }
+      else if (localName.equals("Description"))
+      {
       }
       else if (localName.equals("Metadata"))
       { this.label = attributes.getValue("name");
         this.value = new StringBuffer();
+
+    String mode = attributes.getValue("mode");
+    this.accumulate = mode.equals(ACCUMULATE_MODE); 
       }
     }
 
     public void endElement(String URI, String localName, String qName)
-    { if (localName.equals("Filename"))
-      { this.file = this.value.toString();
+    { if (localName.equals("FileName"))
+      { String file = this.value.toString();
         this.value = null;
+    this.files.add(file);
+      }
+      else if (localName.equals("FileSet"))
+      { // post the existing files item...
+      }
+      else if (localName.equals("Description"))
+      {
       }
       else if (localName.equals("Metadata"))
-      { GMLExtractor.postMetadata(this.file, this.label, this.value.toString());
+      { GMLExtractor.postMetadata(this.files, this.label, this.value.toString());
         this.value = null;
     this.label = null;
@@ -57 +76 @@
 
     public void characters(char c[], int start, int length)
-    { if (this.label != null)
+    { if (this.value != null)
       { String string = new String(c, start, length);
         this.value.append(string);
@@ -123 +142 @@
   }
 
-  protected static void postMetadata(String file, String value, String label)
+  protected static void postMetadata(List files, String value, String label)
   {
+    
   }
 

trunk/gsdl3/src/java/org/greenstone/gsdl3/gs3build/extractor/IndexExtractor.java

@@ -8 +8 @@
 import java.util.ArrayList;
 import java.util.Iterator;
+import org.greenstone.gsdl3.gs3build.util.GS2TextFileHandler;
 
 import org.xml.sax.XMLReader;
@@ -33 +34 @@
    *  An inner class to handle GML files
    */
-  class IndexHandler
-  { String  content;
-    String  line;
-    int     pos;
-    boolean doneRow;
-    List    labels;
+  class IndexHandler extends GS2TextFileHandler
+  { List    labels;
     URL     base;
 
     IndexHandler(String content, URL url, DocumentList documentList) throws IndexHandlerException
-    { this.content = content;
-      this.doneRow = false;
+    { 
+      super(content);
+
       this.labels  = new ArrayList();
       this.base    = url;
@@ -69 +67 @@
       while (this.hasMore())
       { String label = this.getEntry(true);
-        if (label == null)
-      continue;
-    label.trim();
-
-    if (label.length() == 0) {
+        if (label == null || label.length() == 0) {
       continue;
     }
@@ -86 +80 @@
     // Get the file pattern itself
     String filePattern = this.getEntry(true);
-    if (filePattern == null) {
+    if (filePattern == null || filePattern.length() == 0) {
       continue;
     }
 
-    filePattern.trim();
-    if (filePattern.length() == 0) {
-      continue;
-    }
-
     // get a list of documents that match the file pattern
-    List files = documentList.findDocumentIdsUsingFile(filePattern);
-    if (files != null) {
-      Iterator iterator = files.iterator();
+    List documentIds = documentList.findDocumentIdsUsingFile(filePattern);
+    if (documentIds != null) {
+      Iterator iterator = documentIds.iterator();
       while (iterator.hasNext()) {
         System.out.println("Matches file " + iterator.next().toString());
@@ -106 +95 @@
     // if no files match this data, then skip this row
     // TODO: raise a quality error message
-    if (files == null || files.size() == 0) {
+    if (documentIds == null || documentIds.size() == 0) {
       continue;
     }
 
-    // TODO: cache up the documents that match for speed?
+    // cache up the documents that match for speed improvements...
+    List documents = new ArrayList();
+    Iterator idIterator = documentIds.iterator();
+    while (idIterator.hasNext()) {
+      String docIdString = idIterator.next().toString();
+      System.out.println(docIdString);
+      DocumentID docId   = new DocumentID(docIdString);
+      DocumentInterface document = documentList.getDocument(docId);
+      if (document != null) {
+        documents.add(document);
+      }
+    }
 
     // Next, split the row into the separate metadata items
@@ -116 +116 @@
     while (this.hasMore()) {
       String item = this.getEntry(true);
-      if (item == null) {
-        entryNo ++;
-        continue;
-      }
-
-      item.trim();
-      if (item.length() == 0) {
+      if (item == null || item.length() == 0) {
         entryNo ++;
         continue;
@@ -154 +148 @@
       // in order to minimise rewrites...
       if (label != null) {
-        System.out.println("Posting " + label + "=" + item + " to " + parentDir + filePattern);
-        // Post to document
+        Iterator docIterator = documents.iterator();
+        while (docIterator.hasNext()) {
+          DocumentInterface document = (DocumentInterface) docIterator.next();
+
+          // Post to document
+          // TODO: tailor this to posting documents to *sections* as required...
+          document.addDocumentMetadata(label, item);
+          System.out.println("Posting " + label + "=" + item + " to " + parentDir + filePattern);
+        }
       }
       entryNo ++;
     }
+    
+    // write out the modified documents
+    // TODO: nicer/more generalised interface for this and related activity in
+    //       extractor manager (actually, enricher manager);
+    Iterator docIterator = documents.iterator();
+    while (docIterator.hasNext()) {
+      DocumentInterface document = (DocumentInterface) docIterator.next();
+
+      documentList.modifiedDocument(document);
+    }
       }
     }
 
-    private boolean hasMore()
-    { return this.line != null;
-    }
-
-    private boolean hasMoreLines()
-    { return this.content != null;
-    }
-
-    private String getEntry()
-    { return this.getEntry(false);
-    }
-
-    private String getEntry(boolean breakSpace)
-    { String reply;
-      int start, tab = 0;
-      boolean quoted = false;
-
-      start = 0;
-      while (start < this.line.length() &&
-         this.line.charAt(start) == ' ') {
-    start ++;
-      }
-
-      if (start == this.line.length()) {
-    this.line = null;
-    return null;
-      }
-
-      if (this.line.charAt(start) == '"') {
-    quoted = true;
-    breakSpace = false;
-    start ++;
-      }
-      tab = start;
-
-      while (tab != this.line.length() &&
-         this.line.charAt(tab) != '\t' &&
-         !(quoted && this.line.charAt(tab) == '"') &&
-             !(this.line.charAt(tab) == ' ' && breakSpace))
-      { tab ++;
-      }
-
-      if (start > 0) {
-    this.line = this.line.substring(start);
-    tab -= start;
-      }
-
-      if (tab == this.line.length()) {
-    reply = this.line;
-    this.line = null;
-      }
-      else {
-    reply = this.line.substring(0, tab);
-    this.line = this.line.substring(tab+1);
-      }
-
-      return reply;
-    }
-
-    private String getLine()
-    { if (this.content == null) {
-    this.line = null;
-    return null;
-      }
-    
-      do {
-        int eol = this.content.indexOf('\n');
-    if (eol < 0) {
-      this.line = this.content;
-      this.content = null;
-    }
-    else {
-      this.line = this.content.substring(0, eol);
-      this.content = this.content.substring(eol+1);
-      while (this.content.length() > 0 &&
-         this.content.charAt(0) < ' ')
-      { this.content = this.content.substring(1);
-      }
-    }
-
-    if (this.line != null) {
-      this.line.trim();
-    }
-      } while (this.content != null && this.line != null && this.line.length() == 0);
-
-      return this.line;
-    }
   }