source: trunk/gli/src/org/greenstone/gatherer/metadata/DocXMLFile.java@ 8142

package org.greenstone.gatherer.metadata;


import java.io.*;
import java.util.*;
import org.greenstone.gatherer.util.Utility;
import org.greenstone.gatherer.util.XMLTools;
import org.w3c.dom.*;


/** This class represents one doc.xml file */
public class DocXMLFile
    extends File
{
    static final private String ARCHIVE_ELEMENT = "Archive";
    static final private String DESCRIPTION_ELEMENT = "Description";
    static final private String METADATA_ELEMENT = "Metadata";
    static final private String SECTION_ELEMENT = "Section";

    private ArrayList files_in_doc_xml_file = new ArrayList();


    public DocXMLFile(String doc_xml_file_path)
    {
    super(doc_xml_file_path);
    }


    public ArrayList getMetadataExtractedFromFile(File file)
    {
    // Build up a list of metadata extracted from this file
    ArrayList metadata_values = new ArrayList();

    String file_relative_path = file.getAbsolutePath();
    int import_index = file_relative_path.indexOf("import");
    if (import_index != -1) {
        file_relative_path = file_relative_path.substring(import_index + "import".length() + 1);
    }

    // Check whether this doc.xml file contains extracted metadata for the specified file
    boolean contains_extracted_metadata_for_file = false;
    for (int i = 0; i < files_in_doc_xml_file.size(); i++) {
        if (file_relative_path.equals(files_in_doc_xml_file.get(i))) {
        contains_extracted_metadata_for_file = true;
        break;
        }
    }

    // ...it doesn't
    if (!contains_extracted_metadata_for_file) {
        return metadata_values;
    }

    // Parse the doc.xml file
    System.err.println("Applicable doc.xml file: " + this);
    Document document = XMLTools.parseXMLFile(this);
    if (document == null) {
        System.err.println("Error: Could not parse doc.xml file " + getAbsolutePath());
        return metadata_values;
    }

    MetadataSet extracted_metadata_set = MetadataSetManager.getMetadataSet(MetadataSetManager.EXTRACTED_METADATA_NAMESPACE);

    // Read all the Archive elements in the file
    NodeList archive_elements_nodelist = document.getElementsByTagName(ARCHIVE_ELEMENT);
    for (int i = 0; i < archive_elements_nodelist.getLength(); i++) {
        Element current_archive_element = (Element) archive_elements_nodelist.item(i);

        // Read the child Section elements of the archive (but not all descendants)
        ArrayList child_section_elements = XMLTools.getChildElementsByTagName(current_archive_element, SECTION_ELEMENT);
        for (int j = 0; j < child_section_elements.size(); j++) {
        Element current_section_element = (Element) child_section_elements.get(j);

        // Read the Description elements of this section only (not child sections as well)
        ArrayList child_description_elements = XMLTools.getChildElementsByTagName(current_section_element, DESCRIPTION_ELEMENT);
        for (int k = 0; k < child_description_elements.size(); k++) {
            Element current_description_element = (Element) child_description_elements.get(k);

            // Read all the Metadata elements in this description element
            NodeList metadata_elements_nodelist = current_description_element.getElementsByTagName(METADATA_ELEMENT);
            for (int l = 0; l < metadata_elements_nodelist.getLength(); l++) {
            Element current_metadata_element = (Element) metadata_elements_nodelist.item(l);
            String metadata_element_name_full = current_metadata_element.getAttribute("name");

            // If the metadata has a namespace it isn't extracted metadata, so we're not interested
            String metadata_set_namespace = MetadataTools.getMetadataSetNamespace(metadata_element_name_full);
            if (!metadata_set_namespace.equals("")) {
                continue;
            }

            // Extracted metadata!
            String metadata_element_name = metadata_element_name_full;

            // Ignore lower-case metadata elements (gsdlsourcefilename, gsdlassocfile etc.)
            //   and those starting with '/' (/srclink)
            char first_character = metadata_element_name.charAt(0);
            if (Character.isLowerCase(first_character) || first_character == '/') {
                continue;
            }

            MetadataElement metadata_element = extracted_metadata_set.getMetadataElement(metadata_element_name);

            // Value trees are not stored for extracted metadata, so create a new value tree node now
            String current_metadata_element_value = XMLTools.getElementTextValue(current_metadata_element);
            metadata_element.addMetadataValue(current_metadata_element_value);
            MetadataValueTreeNode metadata_value_tree_node = metadata_element.getMetadataValueTreeNode(current_metadata_element_value);

            // Add the new metadata value to the list
            MetadataValue metadata_value = new MetadataValue(metadata_element, metadata_value_tree_node);
            metadata_values.add(metadata_value);
            }
        }
        }
    }

    return metadata_values;
    }


    /**
     * Every doc.xml file must be skimmed when a collection is opened, for two reasons:
     *   - To build a mapping from source file to its corresponding doc.xml file
     *   - To get a complete list of all extracted metadata elements
     */
    public void skimFile()
    {
    MetadataSet extracted_metadata_set = MetadataSetManager.getMetadataSet(MetadataSetManager.EXTRACTED_METADATA_NAMESPACE);

    // Skim the doc.xml file as quickly as possible (don't parse as XML), looking at the Metadata elements
    System.err.println("Skimming doc.xml file " + this + "...");
    try {
        BufferedReader buffered_reader = new BufferedReader(new FileReader(this));
        String line = null;
        while ((line = buffered_reader.readLine()) != null) {
        // This line doesn't contain a metadata element
        if (line.indexOf("<Metadata ") == -1) {
            continue;
        }

        // Extract the metadata element name
        int name_index = line.indexOf(" name=\"") + " name=\"".length();
        String metadata_element_name_full = line.substring(name_index, line.indexOf("\"", name_index));

        // If the metadata has a namespace it isn't extracted metadata, so we're not interested
        String metadata_set_namespace = MetadataTools.getMetadataSetNamespace(metadata_element_name_full);
        if (!metadata_set_namespace.equals("")) {
            continue;
        }

        // Extracted metadata!
        String metadata_element_name = metadata_element_name_full;

        // Note which file this doc.xml is for
        if (metadata_element_name.equals("gsdlsourcefilename")) {
            // Extract the gsdlsourcefilename element value
            int value_index = line.indexOf(">", name_index) + ">".length();
            String gsdlsourcefilename_value = line.substring(value_index, line.indexOf("<", value_index));

            // We're only interested in the path relative to the import folder
            int import_index = gsdlsourcefilename_value.indexOf("import");
            if (import_index != -1) {
            gsdlsourcefilename_value = gsdlsourcefilename_value.substring(import_index + "import".length());

            boolean is_unix_path = gsdlsourcefilename_value.startsWith("/");
            gsdlsourcefilename_value = gsdlsourcefilename_value.substring(1);

            // Make sure the path matches the OS that is running
            if (is_unix_path && Utility.isWindows()) {
                // Convert path from Unix to Windows
                gsdlsourcefilename_value = gsdlsourcefilename_value.replaceAll("/", File.separator);
            }
            if (!is_unix_path && !Utility.isWindows()) {
                // Convert path from Windows to Unix
                gsdlsourcefilename_value = gsdlsourcefilename_value.replaceAll("\\\\", File.separator);
            }

            // Remember this for quick access later
            if (gsdlsourcefilename_value != null) {
                files_in_doc_xml_file.add(gsdlsourcefilename_value);
            }
            }
            else {
            // We don't really know what is going on...
            System.err.println("Warning: Could not understand gsdlsourcefilename " + gsdlsourcefilename_value);
            }
        }

        // Ignore lower-case metadata elements (gsdlsourcefilename, gsdlassocfile etc.)
        //   and those starting with '/' (/srclink)
        char first_character = metadata_element_name.charAt(0);
        if (Character.isLowerCase(first_character) || first_character == '/') {
            continue;
        }

        MetadataElement metadata_element = extracted_metadata_set.getMetadataElement(metadata_element_name);
        if (metadata_element == null) {
            // This element isn't defined in ex.mds, so create it for this session
            System.err.println("Extracted metadata element not defined: " + metadata_element_name);
            extracted_metadata_set.addMetadataElementForThisSession(metadata_element_name);
        }
        }
    }
    catch (Exception ex) {
        System.err.println("Exception: " + ex);
        ex.printStackTrace();
    }
    }
}