001/*
002 * Licensed to the Apache Software Foundation (ASF) under one or more
003 * contributor license agreements.  See the NOTICE file distributed with
004 * this work for additional information regarding copyright ownership.
005 * The ASF licenses this file to You under the Apache License, Version 2.0
006 * (the "License"); you may not use this file except in compliance with
007 * the License.  You may obtain a copy of the License at
008 *
009 *     http://www.apache.org/licenses/LICENSE-2.0
010 *
011 * Unless required by applicable law or agreed to in writing, software
012 * distributed under the License is distributed on an "AS IS" BASIS,
013 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
014 * See the License for the specific language governing permissions and
015 * limitations under the License.
016 */
017package org.apache.lucene.demo;
018
019import java.io.BufferedReader;
020import java.io.IOException;
021import java.io.InputStream;
022import java.io.InputStreamReader;
023import java.nio.charset.StandardCharsets;
024import java.nio.file.FileVisitResult;
025import java.nio.file.Files;
026import java.nio.file.Path;
027import java.nio.file.Paths;
028import java.nio.file.SimpleFileVisitor;
029import java.nio.file.attribute.BasicFileAttributes;
030import java.util.Date;
031import org.apache.lucene.analysis.Analyzer;
032import org.apache.lucene.analysis.standard.StandardAnalyzer;
033import org.apache.lucene.demo.knn.DemoEmbeddings;
034import org.apache.lucene.demo.knn.KnnVectorDict;
035import org.apache.lucene.document.Document;
036import org.apache.lucene.document.Field;
037import org.apache.lucene.document.KnnVectorField;
038import org.apache.lucene.document.LongPoint;
039import org.apache.lucene.document.StringField;
040import org.apache.lucene.document.TextField;
041import org.apache.lucene.index.DirectoryReader;
042import org.apache.lucene.index.IndexReader;
043import org.apache.lucene.index.IndexWriter;
044import org.apache.lucene.index.IndexWriterConfig;
045import org.apache.lucene.index.IndexWriterConfig.OpenMode;
046import org.apache.lucene.index.Term;
047import org.apache.lucene.index.VectorSimilarityFunction;
048import org.apache.lucene.store.Directory;
049import org.apache.lucene.store.FSDirectory;
050import org.apache.lucene.util.IOUtils;
051
052/**
053 * Index all text files under a directory.
054 *
055 * <p>This is a command-line application demonstrating simple Lucene indexing. Run it with no
056 * command-line arguments for usage information.
057 */
058public class IndexFiles implements AutoCloseable {
059  static final String KNN_DICT = "knn-dict";
060
061  // Calculates embedding vectors for KnnVector search
062  private final DemoEmbeddings demoEmbeddings;
063  private final KnnVectorDict vectorDict;
064
065  private IndexFiles(KnnVectorDict vectorDict) throws IOException {
066    if (vectorDict != null) {
067      this.vectorDict = vectorDict;
068      demoEmbeddings = new DemoEmbeddings(vectorDict);
069    } else {
070      this.vectorDict = null;
071      demoEmbeddings = null;
072    }
073  }
074
075  /** Index all text files under a directory. */
076  public static void main(String[] args) throws Exception {
077    String usage =
078        "java org.apache.lucene.demo.IndexFiles"
079            + " [-index INDEX_PATH] [-docs DOCS_PATH] [-update] [-knn_dict DICT_PATH]\n\n"
080            + "This indexes the documents in DOCS_PATH, creating a Lucene index"
081            + "in INDEX_PATH that can be searched with SearchFiles\n"
082            + "IF DICT_PATH contains a KnnVector dictionary, the index will also support KnnVector search";
083    String indexPath = "index";
084    String docsPath = null;
085    String vectorDictSource = null;
086    boolean create = true;
087    for (int i = 0; i < args.length; i++) {
088      switch (args[i]) {
089        case "-index":
090          indexPath = args[++i];
091          break;
092        case "-docs":
093          docsPath = args[++i];
094          break;
095        case "-knn_dict":
096          vectorDictSource = args[++i];
097          break;
098        case "-update":
099          create = false;
100          break;
101        case "-create":
102          create = true;
103          break;
104        default:
105          throw new IllegalArgumentException("unknown parameter " + args[i]);
106      }
107    }
108
109    if (docsPath == null) {
110      System.err.println("Usage: " + usage);
111      System.exit(1);
112    }
113
114    final Path docDir = Paths.get(docsPath);
115    if (!Files.isReadable(docDir)) {
116      System.out.println(
117          "Document directory '"
118              + docDir.toAbsolutePath()
119              + "' does not exist or is not readable, please check the path");
120      System.exit(1);
121    }
122
123    Date start = new Date();
124    try {
125      System.out.println("Indexing to directory '" + indexPath + "'...");
126
127      Directory dir = FSDirectory.open(Paths.get(indexPath));
128      Analyzer analyzer = new StandardAnalyzer();
129      IndexWriterConfig iwc = new IndexWriterConfig(analyzer);
130
131      if (create) {
132        // Create a new index in the directory, removing any
133        // previously indexed documents:
134        iwc.setOpenMode(OpenMode.CREATE);
135      } else {
136        // Add new documents to an existing index:
137        iwc.setOpenMode(OpenMode.CREATE_OR_APPEND);
138      }
139
140      // Optional: for better indexing performance, if you
141      // are indexing many documents, increase the RAM
142      // buffer.  But if you do this, increase the max heap
143      // size to the JVM (eg add -Xmx512m or -Xmx1g):
144      //
145      // iwc.setRAMBufferSizeMB(256.0);
146
147      KnnVectorDict vectorDictInstance = null;
148      long vectorDictSize = 0;
149      if (vectorDictSource != null) {
150        KnnVectorDict.build(Paths.get(vectorDictSource), dir, KNN_DICT);
151        vectorDictInstance = new KnnVectorDict(dir, KNN_DICT);
152        vectorDictSize = vectorDictInstance.ramBytesUsed();
153      }
154
155      try (IndexWriter writer = new IndexWriter(dir, iwc);
156          IndexFiles indexFiles = new IndexFiles(vectorDictInstance)) {
157        indexFiles.indexDocs(writer, docDir);
158
159        // NOTE: if you want to maximize search performance,
160        // you can optionally call forceMerge here.  This can be
161        // a terribly costly operation, so generally it's only
162        // worth it when your index is relatively static (ie
163        // you're done adding documents to it):
164        //
165        // writer.forceMerge(1);
166      } finally {
167        IOUtils.close(vectorDictInstance);
168      }
169
170      Date end = new Date();
171      try (IndexReader reader = DirectoryReader.open(dir)) {
172        System.out.println(
173            "Indexed "
174                + reader.numDocs()
175                + " documents in "
176                + (end.getTime() - start.getTime())
177                + " milliseconds");
178        if (reader.numDocs() > 100
179            && vectorDictSize < 1_000_000
180            && System.getProperty("smoketester") == null) {
181          throw new RuntimeException(
182              "Are you (ab)using the toy vector dictionary? See the package javadocs to understand why you got this exception.");
183        }
184      }
185    } catch (IOException e) {
186      System.out.println(" caught a " + e.getClass() + "\n with message: " + e.getMessage());
187    }
188  }
189
190  /**
191   * Indexes the given file using the given writer, or if a directory is given, recurses over files
192   * and directories found under the given directory.
193   *
194   * <p>NOTE: This method indexes one document per input file. This is slow. For good throughput,
195   * put multiple documents into your input file(s). An example of this is in the benchmark module,
196   * which can create "line doc" files, one document per line, using the <a
197   * href="../../../../../contrib-benchmark/org/apache/lucene/benchmark/byTask/tasks/WriteLineDocTask.html"
198   * >WriteLineDocTask</a>.
199   *
200   * @param writer Writer to the index where the given file/dir info will be stored
201   * @param path The file to index, or the directory to recurse into to find files to index
202   * @throws IOException If there is a low-level I/O error
203   */
204  void indexDocs(final IndexWriter writer, Path path) throws IOException {
205    if (Files.isDirectory(path)) {
206      Files.walkFileTree(
207          path,
208          new SimpleFileVisitor<>() {
209            @Override
210            public FileVisitResult visitFile(Path file, BasicFileAttributes attrs) {
211              try {
212                indexDoc(writer, file, attrs.lastModifiedTime().toMillis());
213              } catch (
214                  @SuppressWarnings("unused")
215                  IOException ignore) {
216                ignore.printStackTrace(System.err);
217                // don't index files that can't be read.
218              }
219              return FileVisitResult.CONTINUE;
220            }
221          });
222    } else {
223      indexDoc(writer, path, Files.getLastModifiedTime(path).toMillis());
224    }
225  }
226
227  /** Indexes a single document */
228  void indexDoc(IndexWriter writer, Path file, long lastModified) throws IOException {
229    try (InputStream stream = Files.newInputStream(file)) {
230      // make a new, empty document
231      Document doc = new Document();
232
233      // Add the path of the file as a field named "path".  Use a
234      // field that is indexed (i.e. searchable), but don't tokenize
235      // the field into separate words and don't index term frequency
236      // or positional information:
237      Field pathField = new StringField("path", file.toString(), Field.Store.YES);
238      doc.add(pathField);
239
240      // Add the last modified date of the file a field named "modified".
241      // Use a LongPoint that is indexed (i.e. efficiently filterable with
242      // PointRangeQuery).  This indexes to milli-second resolution, which
243      // is often too fine.  You could instead create a number based on
244      // year/month/day/hour/minutes/seconds, down the resolution you require.
245      // For example the long value 2011021714 would mean
246      // February 17, 2011, 2-3 PM.
247      doc.add(new LongPoint("modified", lastModified));
248
249      // Add the contents of the file to a field named "contents".  Specify a Reader,
250      // so that the text of the file is tokenized and indexed, but not stored.
251      // Note that FileReader expects the file to be in UTF-8 encoding.
252      // If that's not the case searching for special characters will fail.
253      doc.add(
254          new TextField(
255              "contents",
256              new BufferedReader(new InputStreamReader(stream, StandardCharsets.UTF_8))));
257
258      if (demoEmbeddings != null) {
259        try (InputStream in = Files.newInputStream(file)) {
260          float[] vector =
261              demoEmbeddings.computeEmbedding(
262                  new BufferedReader(new InputStreamReader(in, StandardCharsets.UTF_8)));
263          doc.add(
264              new KnnVectorField("contents-vector", vector, VectorSimilarityFunction.DOT_PRODUCT));
265        }
266      }
267
268      if (writer.getConfig().getOpenMode() == OpenMode.CREATE) {
269        // New index, so we just add the document (no old document can be there):
270        System.out.println("adding " + file);
271        writer.addDocument(doc);
272      } else {
273        // Existing index (an old copy of this document may have been indexed) so
274        // we use updateDocument instead to replace the old one matching the exact
275        // path, if present:
276        System.out.println("updating " + file);
277        writer.updateDocument(new Term("path", file.toString()), doc);
278      }
279    }
280  }
281
282  @Override
283  public void close() throws IOException {
284    IOUtils.close(vectorDict);
285  }
286}