source: trunk/gli/src/org/greenstone/gatherer/metadata/DocXMLFile.java@ 8123

Last change on this file since 8123 was 8123, checked in by mdewsnip, 20 years ago

More improvements to the new metadata code, including hfile writing.

  • Property svn:keywords set to Author Date Id Revision
File size: 9.1 KB
Line 
1package org.greenstone.gatherer.metadata;
2
3
4import java.io.*;
5import java.util.*;
6import org.greenstone.gatherer.util.Utility;
7import org.greenstone.gatherer.util.XMLTools;
8import org.w3c.dom.*;
9
10
11/** This class represents one doc.xml file */
12public class DocXMLFile
13 extends File
14{
15 static final private String ARCHIVE_ELEMENT = "Archive";
16 static final private String DESCRIPTION_ELEMENT = "Description";
17 static final private String METADATA_ELEMENT = "Metadata";
18 static final private String SECTION_ELEMENT = "Section";
19
20 private ArrayList files_in_doc_xml_file = new ArrayList();
21
22
23 public DocXMLFile(String doc_xml_file_path)
24 {
25 super(doc_xml_file_path);
26
27 // Parse the doc.xml file
28 System.err.println("Loading doc.xml file " + doc_xml_file_path + "...");
29 Document document = XMLTools.parseXMLFile(this);
30 if (document == null) {
31 System.err.println("Error: Could not parse doc.xml file " + getAbsolutePath());
32 return;
33 }
34
35 MetadataSet extracted_metadata_set = MetadataSetManager.getMetadataSet(MetadataSetManager.EXTRACTED_METADATA_NAMESPACE);
36
37 // Read all the Archive elements in the file
38 NodeList archive_elements_nodelist = document.getElementsByTagName(ARCHIVE_ELEMENT);
39 for (int i = 0; i < archive_elements_nodelist.getLength(); i++) {
40 Element current_archive_element = (Element) archive_elements_nodelist.item(i);
41
42 // Read the child Section elements of the archive (but not all descendants)
43 ArrayList child_section_elements = XMLTools.getChildElementsByTagName(current_archive_element, SECTION_ELEMENT);
44 for (int j = 0; j < child_section_elements.size(); j++) {
45 Element current_section_element = (Element) child_section_elements.get(j);
46
47 // Read the Description elements of this section only (not child sections as well)
48 ArrayList child_description_elements = XMLTools.getChildElementsByTagName(current_section_element, DESCRIPTION_ELEMENT);
49 for (int k = 0; k < child_description_elements.size(); k++) {
50 Element current_description_element = (Element) child_description_elements.get(k);
51
52 String gsdlsourcefilename_value = null;
53 boolean bibliographic_data = false;
54
55 // Read all the Metadata elements in this description element
56 NodeList metadata_elements_nodelist = current_description_element.getElementsByTagName(METADATA_ELEMENT);
57 for (int l = 0; l < metadata_elements_nodelist.getLength(); l++) {
58 Element current_metadata_element = (Element) metadata_elements_nodelist.item(l);
59 String metadata_element_name_full = current_metadata_element.getAttribute("name");
60
61 // If the metadata has a namespace it isn't extracted metadata, so we're not interested
62 String metadata_set_namespace = MetadataTools.getMetadataSetNamespace(metadata_element_name_full);
63 if (!metadata_set_namespace.equals("")) {
64 continue;
65 }
66
67 // Extracted metadata!
68 String metadata_element_name = metadata_element_name_full;
69
70 // Note which file this Section is for
71 if (metadata_element_name.equals("gsdlsourcefilename")) {
72 gsdlsourcefilename_value = XMLTools.getElementTextValue(current_metadata_element);
73
74 // We're only interested in the path relative to the import folder
75 int import_index = gsdlsourcefilename_value.indexOf("import");
76 if (import_index != -1) {
77 gsdlsourcefilename_value = gsdlsourcefilename_value.substring(import_index + "import".length());
78
79 boolean is_unix_path = gsdlsourcefilename_value.startsWith("/");
80 gsdlsourcefilename_value = gsdlsourcefilename_value.substring(1);
81
82 // Make sure the path matches the OS that is running
83 if (is_unix_path && Utility.isWindows()) {
84 // Convert path from Unix to Windows
85 gsdlsourcefilename_value = gsdlsourcefilename_value.replaceAll("/", File.separator);
86 }
87 if (!is_unix_path && !Utility.isWindows()) {
88 // Convert path from Windows to Unix
89 gsdlsourcefilename_value = gsdlsourcefilename_value.replaceAll("\\\\", File.separator);
90 }
91 }
92 else {
93 // We don't really know what is going on...
94 System.err.println("Warning: Could not understand gsdlsourcefilename " + gsdlsourcefilename_value);
95 }
96 }
97
98 // We don't do much with bibliographic data
99 if (metadata_element_name.equals("SourceSegment")) {
100 bibliographic_data = true;
101 }
102
103 // Ignore lower-case metadata elements (gsdlsourcefilename, gsdlassocfile etc.)
104 // and those starting with '/' (/srclink)
105 char first_character = metadata_element_name.charAt(0);
106 if (Character.isLowerCase(first_character) || first_character == '/') {
107 continue;
108 }
109
110 MetadataElement metadata_element = extracted_metadata_set.getMetadataElement(metadata_element_name);
111 if (metadata_element == null) {
112 // This element isn't defined in ex.mds, so create it for this session
113 System.err.println("Extracted metadata element not defined: " + metadata_element_name);
114 extracted_metadata_set.addMetadataElementForThisSession(metadata_element_name);
115 }
116 }
117
118 // Remember this for quick access later
119 if (gsdlsourcefilename_value != null && !bibliographic_data) {
120 files_in_doc_xml_file.add(gsdlsourcefilename_value);
121 }
122 }
123 }
124 }
125 }
126
127
128 public ArrayList getMetadataExtractedFromFile(File file)
129 {
130 // Build up a list of metadata extracted from this file
131 ArrayList metadata_values = new ArrayList();
132
133 String file_relative_path = file.getAbsolutePath();
134 int import_index = file_relative_path.indexOf("import");
135 if (import_index != -1) {
136 file_relative_path = file_relative_path.substring(import_index + "import".length() + 1);
137 }
138
139 // Check whether this doc.xml file contains extracted metadata for the specified file
140 boolean contains_extracted_metadata_for_file = false;
141 for (int i = 0; i < files_in_doc_xml_file.size(); i++) {
142 if (file_relative_path.equals(files_in_doc_xml_file.get(i))) {
143 contains_extracted_metadata_for_file = true;
144 // System.err.println("Found extracted metadata in file " + getAbsolutePath());
145 break;
146 }
147 }
148
149 // ...it doesn't
150 if (!contains_extracted_metadata_for_file) {
151 return metadata_values;
152 }
153
154 // Parse the doc.xml file
155 Document document = XMLTools.parseXMLFile(this);
156 if (document == null) {
157 System.err.println("Error: Could not parse doc.xml file " + getAbsolutePath());
158 return metadata_values;
159 }
160
161 MetadataSet extracted_metadata_set = MetadataSetManager.getMetadataSet(MetadataSetManager.EXTRACTED_METADATA_NAMESPACE);
162
163 // Read all the Archive elements in the file
164 NodeList archive_elements_nodelist = document.getElementsByTagName(ARCHIVE_ELEMENT);
165 for (int i = 0; i < archive_elements_nodelist.getLength(); i++) {
166 Element current_archive_element = (Element) archive_elements_nodelist.item(i);
167
168 // Read the child Section elements of the archive (but not all descendants)
169 ArrayList child_section_elements = XMLTools.getChildElementsByTagName(current_archive_element, SECTION_ELEMENT);
170 for (int j = 0; j < child_section_elements.size(); j++) {
171 Element current_section_element = (Element) child_section_elements.get(j);
172
173 // Read the Description elements of this section only (not child sections as well)
174 ArrayList child_description_elements = XMLTools.getChildElementsByTagName(current_section_element, DESCRIPTION_ELEMENT);
175 for (int k = 0; k < child_description_elements.size(); k++) {
176 Element current_description_element = (Element) child_description_elements.get(k);
177
178 // Read all the Metadata elements in this description element
179 NodeList metadata_elements_nodelist = current_description_element.getElementsByTagName(METADATA_ELEMENT);
180 for (int l = 0; l < metadata_elements_nodelist.getLength(); l++) {
181 Element current_metadata_element = (Element) metadata_elements_nodelist.item(l);
182 String metadata_element_name_full = current_metadata_element.getAttribute("name");
183
184 // If the metadata has a namespace it isn't extracted metadata, so we're not interested
185 String metadata_set_namespace = MetadataTools.getMetadataSetNamespace(metadata_element_name_full);
186 if (!metadata_set_namespace.equals("")) {
187 continue;
188 }
189
190 // Extracted metadata!
191 String metadata_element_name = metadata_element_name_full;
192
193 // Ignore lower-case metadata elements (gsdlsourcefilename, gsdlassocfile etc.)
194 // and those starting with '/' (/srclink)
195 char first_character = metadata_element_name.charAt(0);
196 if (Character.isLowerCase(first_character) || first_character == '/') {
197 continue;
198 }
199
200 MetadataElement metadata_element = extracted_metadata_set.getMetadataElement(metadata_element_name);
201
202 // Value trees are not stored for extracted metadata, so create a new value tree node now
203 String current_metadata_element_value = XMLTools.getElementTextValue(current_metadata_element);
204 metadata_element.addMetadataValue(current_metadata_element_value);
205 MetadataValueTreeNode metadata_value_tree_node = metadata_element.getMetadataValueTreeNode(current_metadata_element_value);
206
207 // Add the new metadata value to the list
208 MetadataValue metadata_value = new MetadataValue(metadata_element, metadata_value_tree_node);
209 metadata_values.add(metadata_value);
210 }
211 }
212 }
213 }
214
215 return metadata_values;
216 }
217}
Note: See TracBrowser for help on using the repository browser.