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 java.util.Objects;
032import org.apache.lucene.analysis.Analyzer;
033import org.apache.lucene.analysis.standard.StandardAnalyzer;
034import org.apache.lucene.demo.knn.DemoEmbeddings;
035import org.apache.lucene.demo.knn.KnnVectorDict;
036import org.apache.lucene.document.Document;
037import org.apache.lucene.document.Field;
038import org.apache.lucene.document.KeywordField;
039import org.apache.lucene.document.KnnFloatVectorField;
040import org.apache.lucene.document.LongField;
041import org.apache.lucene.document.TextField;
042import org.apache.lucene.index.DirectoryReader;
043import org.apache.lucene.index.IndexReader;
044import org.apache.lucene.index.IndexWriter;
045import org.apache.lucene.index.IndexWriterConfig;
046import org.apache.lucene.index.IndexWriterConfig.OpenMode;
047import org.apache.lucene.index.Term;
048import org.apache.lucene.index.VectorSimilarityFunction;
049import org.apache.lucene.store.Directory;
050import org.apache.lucene.store.FSDirectory;
051import org.apache.lucene.util.IOUtils;
052
053/**
054 * Index all text files under a directory.
055 *
056 * <p>This is a command-line application demonstrating simple Lucene indexing. Run it with no
057 * command-line arguments for usage information.
058 */
059public class IndexFiles implements AutoCloseable {
060  static final String KNN_DICT = "knn-dict";
061
062  // Calculates embedding vectors for KnnVector search
063  private final DemoEmbeddings demoEmbeddings;
064  private final KnnVectorDict vectorDict;
065
066  private IndexFiles(KnnVectorDict vectorDict) throws IOException {
067    if (vectorDict != null) {
068      this.vectorDict = vectorDict;
069      demoEmbeddings = new DemoEmbeddings(vectorDict);
070    } else {
071      this.vectorDict = null;
072      demoEmbeddings = null;
073    }
074  }
075
076  /** Index all text files under a directory. */
077  public static void main(String[] args) throws Exception {
078    String usage =
079        "java org.apache.lucene.demo.IndexFiles"
080            + " [-index INDEX_PATH] [-docs DOCS_PATH] [-update] [-knn_dict DICT_PATH]\n\n"
081            + "This indexes the documents in DOCS_PATH, creating a Lucene index "
082            + "in INDEX_PATH that can be searched with SearchFiles\n"
083            + "IF DICT_PATH contains a KnnVector dictionary, the index will also support KnnVector search";
084    String indexPath = "index";
085    String docsPath = null;
086    String vectorDictSource = null;
087    boolean create = true;
088    for (int i = 0; i < args.length; i++) {
089      switch (args[i]) {
090        case "-index":
091          indexPath = args[++i];
092          break;
093        case "-docs":
094          docsPath = args[++i];
095          break;
096        case "-knn_dict":
097          vectorDictSource = args[++i];
098          break;
099        case "-update":
100          create = false;
101          break;
102        case "-create":
103          create = true;
104          break;
105        default:
106          throw new IllegalArgumentException("unknown parameter " + args[i]);
107      }
108    }
109
110    if (docsPath == null) {
111      System.err.println("Usage: " + usage);
112      System.exit(1);
113    }
114
115    final Path docDir = Paths.get(docsPath);
116    if (!Files.isReadable(docDir)) {
117      System.out.println(
118          "Document directory '"
119              + docDir.toAbsolutePath()
120              + "' does not exist or is not readable, please check the path");
121      System.exit(1);
122    }
123
124    Date start = new Date();
125    try {
126      System.out.println("Indexing to directory '" + indexPath + "'...");
127
128      Directory dir = FSDirectory.open(Paths.get(indexPath));
129      Analyzer analyzer = new StandardAnalyzer();
130      IndexWriterConfig iwc = new IndexWriterConfig(analyzer);
131
132      if (create) {
133        // Create a new index in the directory, removing any
134        // previously indexed documents:
135        iwc.setOpenMode(OpenMode.CREATE);
136      } else {
137        // Add new documents to an existing index:
138        iwc.setOpenMode(OpenMode.CREATE_OR_APPEND);
139      }
140
141      // Optional: for better indexing performance, if you
142      // are indexing many documents, increase the RAM
143      // buffer.  But if you do this, increase the max heap
144      // size to the JVM (e.g. add -Xmx512m or -Xmx1g):
145      //
146      // iwc.setRAMBufferSizeMB(256.0);
147
148      KnnVectorDict vectorDictInstance = null;
149      long vectorDictSize = 0;
150      if (vectorDictSource != null) {
151        KnnVectorDict.build(Paths.get(vectorDictSource), dir, KNN_DICT);
152        vectorDictInstance = new KnnVectorDict(dir, KNN_DICT);
153        vectorDictSize = vectorDictInstance.ramBytesUsed();
154      }
155
156      try (IndexWriter writer = new IndexWriter(dir, iwc);
157          IndexFiles indexFiles = new IndexFiles(vectorDictInstance)) {
158        indexFiles.indexDocs(writer, docDir);
159
160        // NOTE: if you want to maximize search performance,
161        // you can optionally call forceMerge here.  This can be
162        // a terribly costly operation, so generally it's only
163        // worth it when your index is relatively static (ie
164        // you're done adding documents to it):
165        //
166        // writer.forceMerge(1);
167      } finally {
168        IOUtils.close(vectorDictInstance);
169      }
170
171      Date end = new Date();
172      try (IndexReader reader = DirectoryReader.open(dir)) {
173        System.out.println(
174            "Indexed "
175                + reader.numDocs()
176                + " documents in "
177                + (end.getTime() - start.getTime())
178                + " ms");
179        if (Objects.isNull(vectorDictSource) == false
180            && reader.numDocs() > 100
181            && vectorDictSize < 1_000_000
182            && System.getProperty("smoketester") == null) {
183          throw new RuntimeException(
184              "Are you (ab)using the toy vector dictionary? See the package javadocs to understand why you got this exception.");
185        }
186      }
187    } catch (IOException e) {
188      System.out.println(" caught a " + e.getClass() + "\n with message: " + e.getMessage());
189    }
190  }
191
192  /**
193   * Indexes the given file using the given writer, or if a directory is given, recurses over files
194   * and directories found under the given directory.
195   *
196   * <p>NOTE: This method indexes one document per input file. This is slow. For good throughput,
197   * put multiple documents into your input file(s). An example of this is in the benchmark module,
198   * which can create "line doc" files, one document per line, using the <a
199   * href="../../../../../contrib-benchmark/org/apache/lucene/benchmark/byTask/tasks/WriteLineDocTask.html"
200   * >WriteLineDocTask</a>.
201   *
202   * @param writer Writer to the index where the given file/dir info will be stored
203   * @param path The file to index, or the directory to recurse into to find files to index
204   * @throws IOException If there is a low-level I/O error
205   */
206  void indexDocs(final IndexWriter writer, Path path) throws IOException {
207    if (Files.isDirectory(path)) {
208      Files.walkFileTree(
209          path,
210          new SimpleFileVisitor<>() {
211            @Override
212            public FileVisitResult visitFile(Path file, BasicFileAttributes attrs) {
213              try {
214                indexDoc(writer, file, attrs.lastModifiedTime().toMillis());
215              } catch (
216                  @SuppressWarnings("unused")
217                  IOException ignore) {
218                ignore.printStackTrace(System.err);
219                // don't index files that can't be read.
220              }
221              return FileVisitResult.CONTINUE;
222            }
223          });
224    } else {
225      indexDoc(writer, path, Files.getLastModifiedTime(path).toMillis());
226    }
227  }
228
229  /** Indexes a single document */
230  void indexDoc(IndexWriter writer, Path file, long lastModified) throws IOException {
231    try (InputStream stream = Files.newInputStream(file)) {
232      // make a new, empty document
233      Document doc = new Document();
234
235      // Add the path of the file as a field named "path".  Use a
236      // field that is indexed (i.e. searchable), but don't tokenize
237      // the field into separate words and don't index term frequency
238      // or positional information:
239      doc.add(new KeywordField("path", file.toString(), Field.Store.YES));
240
241      // Add the last modified date of the file a field named "modified".
242      // Use a LongField that is indexed with points and doc values, and is efficient
243      // for both filtering (LongField#newRangeQuery) and sorting
244      // (LongField#newSortField).  This indexes to millisecond resolution, which
245      // is often too fine.  You could instead create a number based on
246      // year/month/day/hour/minutes/seconds, down the resolution you require.
247      // For example the long value 2011021714 would mean
248      // February 17, 2011, 2-3 PM.
249      doc.add(new LongField("modified", lastModified, Field.Store.NO));
250
251      // Add the contents of the file to a field named "contents".  Specify a Reader,
252      // so that the text of the file is tokenized and indexed, but not stored.
253      // Note that FileReader expects the file to be in UTF-8 encoding.
254      // If that's not the case searching for special characters will fail.
255      doc.add(
256          new TextField(
257              "contents",
258              new BufferedReader(new InputStreamReader(stream, StandardCharsets.UTF_8))));
259
260      if (demoEmbeddings != null) {
261        try (InputStream in = Files.newInputStream(file)) {
262          float[] vector =
263              demoEmbeddings.computeEmbedding(
264                  new BufferedReader(new InputStreamReader(in, StandardCharsets.UTF_8)));
265          doc.add(
266              new KnnFloatVectorField(
267                  "contents-vector", vector, VectorSimilarityFunction.DOT_PRODUCT));
268        }
269      }
270
271      if (writer.getConfig().getOpenMode() == OpenMode.CREATE) {
272        // New index, so we just add the document (no old document can be there):
273        System.out.println("adding " + file);
274        writer.addDocument(doc);
275      } else {
276        // Existing index (an old copy of this document may have been indexed) so
277        // we use updateDocument instead to replace the old one matching the exact
278        // path, if present:
279        System.out.println("updating " + file);
280        writer.updateDocument(new Term("path", file.toString()), doc);
281      }
282    }
283  }
284
285  @Override
286  public void close() throws IOException {
287    IOUtils.close(vectorDict);
288  }
289}