static {
assert Codec.forName(Lucene.LATEST_CODEC)
.getClass()
.isAssignableFrom(PerFieldMappingPostingFormatCodec.class)
: "PerFieldMappingPostingFormatCodec must subclass the latest lucene codec: "
+ Lucene.LATEST_CODEC;
}
// we want an exception if its not found.
public void testBogusLookup() {
try {
Codec.forName("dskfdskfsdfksdfdsf");
fail();
} catch (IllegalArgumentException expected) {
}
}
private static Codec readCodec(DataInput input, boolean unsupportedAllowed) throws IOException {
final String name = input.readString();
try {
return Codec.forName(name);
} catch (IllegalArgumentException e) {
// give better error messages if we can, first check if this is a legacy codec
if (unsupportedCodecs.contains(name)) {
// We should only get here on pre-5.3 indices, but we can't test this until 7.0 when 5.x
// indices become too old:
assert unsupportedAllowed;
IOException newExc =
new IndexFormatTooOldException(input, "Codec '" + name + "' is too old");
newExc.initCause(e);
throw newExc;
}
// or maybe it's an old default codec that moved
if (name.startsWith("Lucene")) {
throw new IllegalArgumentException(
"Could not load codec '"
+ name
+ "'. Did you forget to add lucene-backward-codecs.jar?",
e);
}
throw e;
}
}
public void testLookup() {
Codec codec = Codec.forName("Lucene410");
assertEquals("Lucene410", codec.getName());
}
/**
* Encodes/decodes an inverted index segment.
*
* <p>Note, when extending this class, the name ({@link #getName}) is written into the index. In
* order for the segment to be read, the name must resolve to your implementation via {@link
* #forName(String)}. This method uses Java's {@link ServiceLoader Service Provider Interface} to
* resolve codec names.
*
* <p>
*
* @see ServiceLoader
*/
public abstract class Codec implements NamedSPILoader.NamedSPI {
private static final NamedSPILoader<Codec> loader = new NamedSPILoader<Codec>(Codec.class);
private final String name;
public Codec(String name) {
NamedSPILoader.checkServiceName(name);
this.name = name;
}
/** Returns this codec's name */
@Override
public final String getName() {
return name;
}
/** Encodes/decodes postings */
public abstract PostingsFormat postingsFormat();
/** Encodes/decodes docvalues */
public abstract DocValuesFormat docValuesFormat();
/** Encodes/decodes stored fields */
public abstract StoredFieldsFormat storedFieldsFormat();
/** Encodes/decodes term vectors */
public abstract TermVectorsFormat termVectorsFormat();
/** Encodes/decodes field infos file */
public abstract FieldInfosFormat fieldInfosFormat();
/** Encodes/decodes segment info file */
public abstract SegmentInfoFormat segmentInfoFormat();
/** Encodes/decodes document normalization values */
public abstract NormsFormat normsFormat();
/** Encodes/decodes live docs */
public abstract LiveDocsFormat liveDocsFormat();
/** looks up a codec by name */
public static Codec forName(String name) {
return loader.lookup(name);
}
/** returns a list of all available codec names */
public static Set<String> availableCodecs() {
return loader.availableServices();
}
private static Codec defaultCodec = Codec.forName("Lucene40");
/** expert: returns the default codec used for newly created {@link IndexWriterConfig}s. */
// TODO: should we use this, or maybe a system property is better?
public static Codec getDefault() {
return defaultCodec;
}
/** expert: sets the default codec used for newly created {@link IndexWriterConfig}s. */
public static void setDefault(Codec codec) {
defaultCodec = codec;
}
@Override
public String toString() {
return name;
}
}
@Test
public void testGetThatFieldProbabilityRatioIsReflectedInBoost() throws Exception {
ArgumentCaptor<Float> normalizeCaptor = ArgumentCaptor.forClass(Float.class);
DocumentFrequencyCorrection dfc = new DocumentFrequencyCorrection();
Directory directory = newDirectory();
Analyzer analyzer =
new Analyzer() {
protected TokenStreamComponents createComponents(String fieldName) {
Tokenizer source = new WhitespaceTokenizer();
TokenStream filter =
new WordDelimiterFilter(
source,
WordDelimiterFilter.GENERATE_WORD_PARTS
| WordDelimiterFilter.SPLIT_ON_CASE_CHANGE,
null);
filter = new LowerCaseFilter(filter);
return new TokenStreamComponents(source, filter);
}
};
IndexWriterConfig conf = new IndexWriterConfig(analyzer);
conf.setCodec(Codec.forName(TestUtil.LUCENE_CODEC));
IndexWriter indexWriter = new IndexWriter(directory, conf);
// Both fields f1 and f2 have 10 terms in total.
// f1: the search terms (abc def) make 100% of all terms in f1
// f2: the search terms (abc def) make 50% of all terms in f2
// --> we expect that the sum of the boost factors for terms in bq(+f1:abc, +f1:def)
// equals 2 * sum of the boost factors for terms in bq(+f2:abc, +f2:def)
PRMSFieldBoostTest.addNumDocs("f1", "abc def", indexWriter, 2);
PRMSFieldBoostTest.addNumDocs("f1", "abc", indexWriter, 4);
PRMSFieldBoostTest.addNumDocs("f1", "def", indexWriter, 2);
PRMSFieldBoostTest.addNumDocs("f2", "abc def", indexWriter, 1);
PRMSFieldBoostTest.addNumDocs("f2", "abc", indexWriter, 2);
PRMSFieldBoostTest.addNumDocs("f2", "def", indexWriter, 1);
PRMSFieldBoostTest.addNumDocs("f2", "ghi", indexWriter, 5);
indexWriter.close();
IndexReader indexReader = DirectoryReader.open(directory);
IndexSearcher indexSearcher = new IndexSearcher(indexReader);
indexSearcher.setSimilarity(similarity);
Map<String, Float> fields = new HashMap<>();
fields.put("f1", 1f);
fields.put("f2", 1f);
SearchFieldsAndBoosting searchFieldsAndBoosting =
new SearchFieldsAndBoosting(FieldBoostModel.PRMS, fields, fields, 0.8f);
LuceneQueryBuilder queryBuilder =
new LuceneQueryBuilder(dfc, analyzer, searchFieldsAndBoosting, 0.01f, null);
WhiteSpaceQuerqyParser parser = new WhiteSpaceQuerqyParser();
Query query = queryBuilder.createQuery(parser.parse("AbcDef"));
dfc.finishedUserQuery();
assertTrue(query instanceof DisjunctionMaxQuery);
DisjunctionMaxQuery dmq = (DisjunctionMaxQuery) query;
List<Query> disjuncts = dmq.getDisjuncts();
assertEquals(2, disjuncts.size());
Query disjunct1 = disjuncts.get(0);
if (disjunct1 instanceof BoostQuery) {
disjunct1 = ((BoostQuery) disjunct1).getQuery();
}
assertTrue(disjunct1 instanceof BooleanQuery);
BooleanQuery bq1 = (BooleanQuery) disjunct1;
Query disjunct2 = disjuncts.get(1);
if (disjunct2 instanceof BoostQuery) {
disjunct2 = ((BoostQuery) disjunct2).getQuery();
}
assertTrue(disjunct2 instanceof BooleanQuery);
BooleanQuery bq2 = (BooleanQuery) disjunct2;
final Weight weight1 = bq1.createWeight(indexSearcher, true);
weight1.normalize(0.1f, 4f);
final Weight weight2 = bq2.createWeight(indexSearcher, true);
weight2.normalize(0.1f, 4f);
Mockito.verify(simWeight, times(4)).normalize(eq(0.1f), normalizeCaptor.capture());
final List<Float> capturedBoosts = normalizeCaptor.getAllValues();
// capturedBoosts = boosts of [bq1.term1, bq1.term2, bq2.term1, bq2.term2 ]
assertEquals(capturedBoosts.get(0), capturedBoosts.get(1), 0.00001);
assertEquals(capturedBoosts.get(2), capturedBoosts.get(3), 0.00001);
assertEquals(2f, capturedBoosts.get(0) / capturedBoosts.get(3), 0.00001);
indexReader.close();
directory.close();
analyzer.close();
}
@Override
protected void before() throws Exception {
// enable this by default, for IDE consistency with ant tests (as its the default from ant)
// TODO: really should be in solr base classes, but some extend LTC directly.
// we do this in beforeClass, because some tests currently disable it
restoreProperties.put("solr.directoryFactory", System.getProperty("solr.directoryFactory"));
if (System.getProperty("solr.directoryFactory") == null) {
System.setProperty("solr.directoryFactory", "org.apache.solr.core.MockDirectoryFactory");
}
// Restore more Solr properties.
restoreProperties.put("solr.solr.home", System.getProperty("solr.solr.home"));
restoreProperties.put("solr.data.dir", System.getProperty("solr.data.dir"));
// if verbose: print some debugging stuff about which codecs are loaded.
if (VERBOSE) {
Set<String> codecs = Codec.availableCodecs();
for (String codec : codecs) {
System.out.println(
"Loaded codec: '" + codec + "': " + Codec.forName(codec).getClass().getName());
}
Set<String> postingsFormats = PostingsFormat.availablePostingsFormats();
for (String postingsFormat : postingsFormats) {
System.out.println(
"Loaded postingsFormat: '"
+ postingsFormat
+ "': "
+ PostingsFormat.forName(postingsFormat).getClass().getName());
}
}
savedInfoStream = InfoStream.getDefault();
final Random random = RandomizedContext.current().getRandom();
final boolean v = random.nextBoolean();
if (INFOSTREAM) {
InfoStream.setDefault(new ThreadNameFixingPrintStreamInfoStream(System.out));
} else if (v) {
InfoStream.setDefault(new NullInfoStream());
}
Class<?> targetClass = RandomizedContext.current().getTargetClass();
avoidCodecs = new HashSet<String>();
if (targetClass.isAnnotationPresent(SuppressCodecs.class)) {
SuppressCodecs a = targetClass.getAnnotation(SuppressCodecs.class);
avoidCodecs.addAll(Arrays.asList(a.value()));
}
// set back to default
LuceneTestCase.PREFLEX_IMPERSONATION_IS_ACTIVE = false;
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE = false;
savedCodec = Codec.getDefault();
int randomVal = random.nextInt(10);
if ("Lucene3x".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& "random".equals(TEST_POSTINGSFORMAT)
&& "random".equals(TEST_DOCVALUESFORMAT)
&& randomVal == 3
&& !shouldAvoidCodec("Lucene3x"))) { // preflex-only setup
codec = Codec.forName("Lucene3x");
assert (codec instanceof PreFlexRWCodec)
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
LuceneTestCase.PREFLEX_IMPERSONATION_IS_ACTIVE = true;
} else if ("Lucene40".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& "random".equals(TEST_POSTINGSFORMAT)
&& randomVal == 0
&& !shouldAvoidCodec("Lucene40"))) { // 4.0 setup
codec = Codec.forName("Lucene40");
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE = true;
assert codec instanceof Lucene40RWCodec
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
assert (PostingsFormat.forName("Lucene40") instanceof Lucene40RWPostingsFormat)
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
} else if ("Lucene41".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& "random".equals(TEST_POSTINGSFORMAT)
&& "random".equals(TEST_DOCVALUESFORMAT)
&& randomVal == 1
&& !shouldAvoidCodec("Lucene41"))) {
codec = Codec.forName("Lucene41");
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE = true;
assert codec instanceof Lucene41RWCodec
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
} else if ("Lucene42".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& "random".equals(TEST_POSTINGSFORMAT)
&& "random".equals(TEST_DOCVALUESFORMAT)
&& randomVal == 2
&& !shouldAvoidCodec("Lucene42"))) {
codec = Codec.forName("Lucene42");
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE = true;
assert codec instanceof Lucene42RWCodec
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
} else if ("Lucene45".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& "random".equals(TEST_POSTINGSFORMAT)
&& "random".equals(TEST_DOCVALUESFORMAT)
&& randomVal == 5
&& !shouldAvoidCodec("Lucene45"))) {
codec = Codec.forName("Lucene45");
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE = true;
assert codec instanceof Lucene45RWCodec
: "fix your classpath to have tests-framework.jar before lucene-core.jar";
} else if (("random".equals(TEST_POSTINGSFORMAT) == false)
|| ("random".equals(TEST_DOCVALUESFORMAT) == false)) {
// the user wired postings or DV: this is messy
// refactor into RandomCodec....
final PostingsFormat format;
if ("random".equals(TEST_POSTINGSFORMAT)) {
format = PostingsFormat.forName("Lucene41");
} else {
format = PostingsFormat.forName(TEST_POSTINGSFORMAT);
}
final DocValuesFormat dvFormat;
if ("random".equals(TEST_DOCVALUESFORMAT)) {
dvFormat = DocValuesFormat.forName("Lucene45");
} else {
dvFormat = DocValuesFormat.forName(TEST_DOCVALUESFORMAT);
}
codec =
new Lucene46Codec() {
@Override
public PostingsFormat getPostingsFormatForField(String field) {
return format;
}
@Override
public DocValuesFormat getDocValuesFormatForField(String field) {
return dvFormat;
}
@Override
public String toString() {
return super.toString() + ": " + format.toString() + ", " + dvFormat.toString();
}
};
} else if ("SimpleText".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& randomVal == 9
&& LuceneTestCase.rarely(random)
&& !shouldAvoidCodec("SimpleText"))) {
codec = new SimpleTextCodec();
} else if ("Appending".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC) && randomVal == 8 && !shouldAvoidCodec("Appending"))) {
codec = new AppendingRWCodec();
LuceneTestCase.OLD_FORMAT_IMPERSONATION_IS_ACTIVE =
true; // this is really just Lucene40 with some minor changes
} else if ("CheapBastard".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC)
&& randomVal == 8
&& !shouldAvoidCodec("CheapBastard")
&& !shouldAvoidCodec("Lucene41"))) {
// we also avoid this codec if Lucene41 is avoided, since thats the postings format it uses.
codec = new CheapBastardCodec();
} else if ("Asserting".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC) && randomVal == 6 && !shouldAvoidCodec("Asserting"))) {
codec = new AssertingCodec();
} else if ("Compressing".equals(TEST_CODEC)
|| ("random".equals(TEST_CODEC) && randomVal == 5 && !shouldAvoidCodec("Compressing"))) {
codec = CompressingCodec.randomInstance(random);
} else if (!"random".equals(TEST_CODEC)) {
codec = Codec.forName(TEST_CODEC);
} else if ("random".equals(TEST_POSTINGSFORMAT)) {
codec = new RandomCodec(random, avoidCodecs);
} else {
assert false;
}
Codec.setDefault(codec);
// Initialize locale/ timezone.
String testLocale = System.getProperty("tests.locale", "random");
String testTimeZone = System.getProperty("tests.timezone", "random");
// Always pick a random one for consistency (whether tests.locale was specified or not).
savedLocale = Locale.getDefault();
Locale randomLocale = randomLocale(random);
locale = testLocale.equals("random") ? randomLocale : localeForName(testLocale);
Locale.setDefault(locale);
// TimeZone.getDefault will set user.timezone to the default timezone of the user's locale.
// So store the original property value and restore it at end.
restoreProperties.put("user.timezone", System.getProperty("user.timezone"));
savedTimeZone = TimeZone.getDefault();
TimeZone randomTimeZone = randomTimeZone(random());
timeZone = testTimeZone.equals("random") ? randomTimeZone : TimeZone.getTimeZone(testTimeZone);
TimeZone.setDefault(timeZone);
similarity =
random().nextBoolean() ? new DefaultSimilarity() : new RandomSimilarityProvider(random());
// Check codec restrictions once at class level.
try {
checkCodecRestrictions(codec);
} catch (AssumptionViolatedException e) {
System.err.println(
"NOTE: "
+ e.getMessage()
+ " Suppressed codecs: "
+ Arrays.toString(avoidCodecs.toArray()));
throw e;
}
}
/**
* Encodes/decodes an inverted index segment.
*
* <p>Note, when extending this class, the name ({@link #getName}) is written into the index. In
* order for the segment to be read, the name must resolve to your implementation via {@link
* #forName(String)}. This method uses Java's {@link ServiceLoader Service Provider Interface} (SPI)
* to resolve codec names.
*
* <p>If you implement your own codec, make sure that it has a no-arg constructor so SPI can load
* it.
*
* @see ServiceLoader
*/
public abstract class Codec implements NamedSPILoader.NamedSPI {
private static final NamedSPILoader<Codec> loader = new NamedSPILoader<Codec>(Codec.class);
private final String name;
/**
* Creates a new codec.
*
* <p>The provided name will be written into the index segment: in order to for the segment to be
* read this class should be registered with Java's SPI mechanism (registered in META-INF/ of your
* jar file, etc).
*
* @param name must be all ascii alphanumeric, and less than 128 characters in length.
*/
protected Codec(String name) {
NamedSPILoader.checkServiceName(name);
this.name = name;
}
/** Returns this codec's name */
@Override
public final String getName() {
return name;
}
/** Encodes/decodes postings */
public abstract PostingsFormat postingsFormat();
/** Encodes/decodes docvalues */
public abstract DocValuesFormat docValuesFormat();
/** Encodes/decodes stored fields */
public abstract StoredFieldsFormat storedFieldsFormat();
/** Encodes/decodes term vectors */
public abstract TermVectorsFormat termVectorsFormat();
/** Encodes/decodes field infos file */
public abstract FieldInfosFormat fieldInfosFormat();
/** Encodes/decodes segment info file */
public abstract SegmentInfoFormat segmentInfoFormat();
/** Encodes/decodes document normalization values */
public abstract NormsFormat normsFormat();
/** Encodes/decodes live docs */
public abstract LiveDocsFormat liveDocsFormat();
/** looks up a codec by name */
public static Codec forName(String name) {
if (loader == null) {
throw new IllegalStateException(
"You called Codec.forName() before all Codecs could be initialized. "
+ "This likely happens if you call it from a Codec's ctor.");
}
return loader.lookup(name);
}
/** returns a list of all available codec names */
public static Set<String> availableCodecs() {
if (loader == null) {
throw new IllegalStateException(
"You called Codec.availableCodecs() before all Codecs could be initialized. "
+ "This likely happens if you call it from a Codec's ctor.");
}
return loader.availableServices();
}
/**
* Reloads the codec list from the given {@link ClassLoader}. Changes to the codecs are visible
* after the method ends, all iterators ({@link #availableCodecs()},...) stay consistent.
*
* <p><b>NOTE:</b> Only new codecs are added, existing ones are never removed or replaced.
*
* <p><em>This method is expensive and should only be called for discovery of new codecs on the
* given classpath/classloader!</em>
*/
public static void reloadCodecs(ClassLoader classloader) {
loader.reload(classloader);
}
private static Codec defaultCodec = Codec.forName("Lucene41");
/** expert: returns the default codec used for newly created {@link IndexWriterConfig}s. */
// TODO: should we use this, or maybe a system property is better?
public static Codec getDefault() {
return defaultCodec;
}
/** expert: sets the default codec used for newly created {@link IndexWriterConfig}s. */
public static void setDefault(Codec codec) {
defaultCodec = codec;
}
/**
* returns the codec's name. Subclasses can override to provide more detail (such as parameters).
*/
@Override
public String toString() {
return name;
}
}
public void testDocsStuckInRAMForever() throws Exception {
Directory dir = newDirectory();
IndexWriterConfig iwc = new IndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()));
iwc.setRAMBufferSizeMB(.2);
Codec codec = Codec.forName("Lucene49");
iwc.setCodec(codec);
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
final IndexWriter w = new IndexWriter(dir, iwc);
final CountDownLatch startingGun = new CountDownLatch(1);
Thread[] threads = new Thread[2];
for (int i = 0; i < threads.length; i++) {
final int threadID = i;
threads[i] =
new Thread() {
@Override
public void run() {
try {
startingGun.await();
for (int j = 0; j < 1000; j++) {
Document doc = new Document();
doc.add(newStringField("field", "threadID" + threadID, Field.Store.NO));
w.addDocument(doc);
}
} catch (Exception e) {
throw new RuntimeException(e);
}
}
};
threads[i].start();
}
startingGun.countDown();
for (Thread t : threads) {
t.join();
}
Set<String> segSeen = new HashSet<>();
int thread0Count = 0;
int thread1Count = 0;
// At this point the writer should have 2 thread states w/ docs; now we index with only 1 thread
// until we see all 1000 thread0 & thread1
// docs flushed. If the writer incorrectly holds onto previously indexed docs forever then this
// will run forever:
while (thread0Count < 1000 || thread1Count < 1000) {
Document doc = new Document();
doc.add(newStringField("field", "threadIDmain", Field.Store.NO));
w.addDocument(doc);
for (String fileName : dir.listAll()) {
if (fileName.endsWith(".si")) {
String segName = IndexFileNames.parseSegmentName(fileName);
if (segSeen.contains(segName) == false) {
segSeen.add(segName);
SegmentInfo si =
new Lucene46SegmentInfoFormat()
.getSegmentInfoReader()
.read(dir, segName, IOContext.DEFAULT);
si.setCodec(codec);
SegmentCommitInfo sci = new SegmentCommitInfo(si, 0, -1, -1, -1);
SegmentReader sr = new SegmentReader(sci, 1, IOContext.DEFAULT);
try {
thread0Count += sr.docFreq(new Term("field", "threadID0"));
thread1Count += sr.docFreq(new Term("field", "threadID1"));
} finally {
sr.close();
}
}
}
}
}
w.close();
dir.close();
}