private void CreateRandomIndexes(int maxSegments)
{
dir = NewDirectory();
numDocs = AtLeast(150);
int numTerms = TestUtil.NextInt32(Random, 1, numDocs / 5);
ISet <string> randomTerms = new JCG.HashSet <string>();
while (randomTerms.size() < numTerms)
{
randomTerms.add(TestUtil.RandomSimpleString(Random));
}
terms = new JCG.List <string>(randomTerms);
long seed = Random.NextInt64();
IndexWriterConfig iwc = NewIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(new J2N.Randomizer(seed)));
iwc.SetMergePolicy(TestSortingMergePolicy.NewSortingMergePolicy(sort));
iw = new RandomIndexWriter(new J2N.Randomizer(seed), dir, iwc);
for (int i = 0; i < numDocs; ++i)
{
Document doc = RandomDocument();
iw.AddDocument(doc);
if (i == numDocs / 2 || (i != numDocs - 1 && Random.nextInt(8) == 0))
{
iw.Commit();
}
if (Random.nextInt(15) == 0)
{
string term = RandomPicks.RandomFrom(Random, terms);
iw.DeleteDocuments(new Term("s", term));
}
}
reader = iw.GetReader();
}
private void TestRandomWords(int maxNumWords, int numIter)
{
Random random = new Random(Random.Next());
for (int iter = 0; iter < numIter; iter++)
{
if (Verbose)
{
Console.WriteLine("\nTEST: iter " + iter);
}
for (int inputMode = 0; inputMode < 2; inputMode++)
{
int numWords = random.nextInt(maxNumWords + 1);
ISet <Int32sRef> termsSet = new JCG.HashSet <Int32sRef>();
Int32sRef[] terms = new Int32sRef[numWords];
while (termsSet.size() < numWords)
{
string term = FSTTester <object> .GetRandomString(random);
termsSet.Add(FSTTester <object> .ToInt32sRef(term, inputMode));
}
DoTest(inputMode, termsSet.ToArray());
}
}
}
public void TestWithContexts()
{
Directory dir = NewDirectory();
IndexWriterConfig iwc = NewIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(Random));
iwc.SetMergePolicy(NewLogMergePolicy());
RandomIndexWriter writer = new RandomIndexWriter(Random, dir, iwc);
KeyValuePair <List <string>, IDictionary <string, Document> > res = GenerateIndexDocuments(AtLeast(1000), true, true);
IDictionary <string, Document> docs = res.Value;
List <string> invalidDocTerms = res.Key;
foreach (Document doc in docs.Values)
{
writer.AddDocument(doc);
}
writer.Commit();
writer.Dispose();
IndexReader ir = DirectoryReader.Open(dir);
IDictionary dictionary = new DocumentDictionary(ir, FIELD_NAME, WEIGHT_FIELD_NAME, PAYLOAD_FIELD_NAME, CONTEXT_FIELD_NAME);
IInputIterator inputIterator = dictionary.GetEntryIterator();
BytesRef f;
while ((f = inputIterator.Next()) != null)
{
string field = f.Utf8ToString();
Document doc = docs[field];
docs.Remove(field);
//Document doc = docs.remove(f.utf8ToString());
assertTrue(f.equals(new BytesRef(doc.Get(FIELD_NAME))));
IIndexableField weightField = doc.GetField(WEIGHT_FIELD_NAME);
assertEquals(inputIterator.Weight, (weightField != null) ? weightField.GetInt64ValueOrDefault() : 0);
assertTrue(inputIterator.Payload.equals(doc.GetField(PAYLOAD_FIELD_NAME).GetBinaryValue()));
ISet <BytesRef> oriCtxs = new JCG.HashSet <BytesRef>();
IEnumerable <BytesRef> contextSet = inputIterator.Contexts;
foreach (IIndexableField ctxf in doc.GetFields(CONTEXT_FIELD_NAME))
{
oriCtxs.add(ctxf.GetBinaryValue());
}
assertEquals(oriCtxs.size(), contextSet.Count());
}
foreach (string invalidTerm in invalidDocTerms)
{
var invalid = docs[invalidTerm];
docs.Remove(invalidTerm);
assertNotNull(invalid);
}
assertTrue(!docs.Any());
ir.Dispose();
dir.Dispose();
}
public void TestMultiThreaded()
{
FileInfo file = new FileInfo(Path.Combine(getWorkDir().FullName, "one-line"));
PerfRunData runData = createPerfRunData(file, false, typeof(ThreadingDocMaker).AssemblyQualifiedName);
ThreadJob[] threads = new ThreadJob[10];
using (WriteLineDocTask wldt = new WriteLineDocTask(runData))
{
for (int i = 0; i < threads.Length; i++)
{
threads[i] = new ThreadAnonymousHelper("t" + i, wldt);
}
foreach (ThreadJob t in threads)
{
t.Start();
}
foreach (ThreadJob t in threads)
{
t.Join();
}
} // wldt.Dispose();
ISet <String> ids = new JCG.HashSet <string>();
TextReader br = new StreamReader(new FileStream(file.FullName, FileMode.Open, FileAccess.Read, FileShare.None), Encoding.UTF8);
try
{
String line = br.ReadLine();
assertHeaderLine(line); // header line is written once, no matter how many threads there are
for (int i = 0; i < threads.Length; i++)
{
line = br.ReadLine();
assertNotNull($"line for index {i} is missing", line); // LUCENENET specific - ensure the line is there before splitting
String[] parts = line.Split(WriteLineDocTask.SEP).TrimEnd();
assertEquals(line, 3, parts.Length);
// check that all thread names written are the same in the same line
String tname = parts[0].Substring(parts[0].IndexOf('_'));
ids.add(tname);
assertEquals(tname, parts[1].Substring(parts[1].IndexOf('_')));
assertEquals(tname, parts[2].Substring(parts[2].IndexOf('_')));
}
// only threads.length lines should exist
assertNull(br.ReadLine());
assertEquals(threads.Length, ids.size());
}
finally
{
br.Dispose();
}
}
/**
* Makes a bunch of single-char tokens (the max # unique terms will at most be 26).
* puts the # unique terms into expected, to be checked against the norm.
*/
private string AddValue()
{
StringBuilder sb = new StringBuilder();
ISet <string> terms = new JCG.HashSet <string>();
int num = TestUtil.NextInt32(Random, 0, 255);
for (int i = 0; i < num; i++)
{
sb.append(' ');
char term = (char)TestUtil.NextInt32(Random, 'a', 'z');
sb.append(term);
terms.add("" + term);
}
expected.Add(terms.size());
return(sb.toString());
}
public void TestFastFilter()
{
DuplicateFilter df = new DuplicateFilter(KEY_FIELD);
df.ProcessingMode = (ProcessingMode.PM_FAST_INVALIDATION);
ISet <string> results = new JCG.HashSet <string>();
ScoreDoc[] hits = searcher.Search(tq, df, 1000).ScoreDocs;
assertTrue("Filtered searching should have found some matches", hits.Length > 0);
foreach (ScoreDoc hit in hits)
{
Document d = searcher.Doc(hit.Doc);
string url = d.Get(KEY_FIELD);
assertFalse("No duplicate urls should be returned", results.contains(url));
results.add(url);
}
assertEquals("Two urls found", 2, results.size());
}
public void TestRandom()
{
string[] terms = new string[TestUtil.NextInt32(Random, 2, 10)];
ISet <string> seen = new JCG.HashSet <string>();
while (seen.size() < terms.Length)
{
string token = TestUtil.RandomSimpleString(Random, 1, 5);
if (!seen.contains(token))
{
terms[seen.size()] = token;
seen.add(token);
}
}
Analyzer a = new MockAnalyzer(Random);
int numDocs = AtLeast(10);
long totTokens = 0;
string[][] docs = new string[numDocs][];
for (int i = 0; i < numDocs; i++)
{
docs[i] = new string[AtLeast(100)];
if (Verbose)
{
Console.Write(" doc " + i + ":");
}
for (int j = 0; j < docs[i].Length; j++)
{
docs[i][j] = GetZipfToken(terms);
if (Verbose)
{
Console.Write(" " + docs[i][j]);
}
}
if (Verbose)
{
Console.WriteLine();
}
totTokens += docs[i].Length;
}
int grams = TestUtil.NextInt32(Random, 1, 4);
if (Verbose)
{
Console.WriteLine("TEST: " + terms.Length + " terms; " + numDocs + " docs; " + grams + " grams");
}
// Build suggester model:
FreeTextSuggester sug = new FreeTextSuggester(a, a, grams, (byte)0x20);
sug.Build(new TestRandomInputIterator(this, docs));
// Build inefficient but hopefully correct model:
List <IDictionary <string, int?> > gramCounts = new List <IDictionary <string, int?> >(grams);
for (int gram = 0; gram < grams; gram++)
{
if (Verbose)
{
Console.WriteLine("TEST: build model for gram=" + gram);
}
IDictionary <string, int?> model = new JCG.Dictionary <string, int?>();
gramCounts.Add(model);
foreach (string[] doc in docs)
{
for (int i = 0; i < doc.Length - gram; i++)
{
StringBuilder b = new StringBuilder();
for (int j = i; j <= i + gram; j++)
{
if (j > i)
{
b.append(' ');
}
b.append(doc[j]);
}
string token = b.toString();
if (!model.TryGetValue(token, out int?curCount) || curCount == null)
{
model.Put(token, 1);
}
else
{
model.Put(token, 1 + curCount);
}
if (Verbose)
{
Console.WriteLine(" add '" + token + "' -> count=" + (model.TryGetValue(token, out int?count) ? (count.HasValue ? count.ToString() : "null") : ""));
}
}
}
}
int lookups = AtLeast(100);
for (int iter = 0; iter < lookups; iter++)
{
string[] tokens = new string[TestUtil.NextInt32(Random, 1, 5)];
for (int i = 0; i < tokens.Length; i++)
{
tokens[i] = GetZipfToken(terms);
}
// Maybe trim last token; be sure not to create the
// empty string:
int trimStart;
if (tokens.Length == 1)
{
trimStart = 1;
}
else
{
trimStart = 0;
}
int trimAt = TestUtil.NextInt32(Random, trimStart, tokens[tokens.Length - 1].Length);
tokens[tokens.Length - 1] = tokens[tokens.Length - 1].Substring(0, trimAt - 0);
int num = TestUtil.NextInt32(Random, 1, 100);
StringBuilder b = new StringBuilder();
foreach (string token in tokens)
{
b.append(' ');
b.append(token);
}
string query = b.toString();
query = query.Substring(1);
if (Verbose)
{
Console.WriteLine("\nTEST: iter=" + iter + " query='" + query + "' num=" + num);
}
// Expected:
List <Lookup.LookupResult> expected = new List <Lookup.LookupResult>();
double backoff = 1.0;
seen = new JCG.HashSet <string>();
if (Verbose)
{
Console.WriteLine(" compute expected");
}
for (int i = grams - 1; i >= 0; i--)
{
if (Verbose)
{
Console.WriteLine(" grams=" + i);
}
if (tokens.Length < i + 1)
{
// Don't have enough tokens to use this model
if (Verbose)
{
Console.WriteLine(" skip");
}
continue;
}
if (i == 0 && tokens[tokens.Length - 1].Length == 0)
{
// Never suggest unigrams from empty string:
if (Verbose)
{
Console.WriteLine(" skip unigram priors only");
}
continue;
}
// Build up "context" ngram:
b = new StringBuilder();
for (int j = tokens.Length - i - 1; j < tokens.Length - 1; j++)
{
b.append(' ');
b.append(tokens[j]);
}
string context = b.toString();
if (context.Length > 0)
{
context = context.Substring(1);
}
if (Verbose)
{
Console.WriteLine(" context='" + context + "'");
}
long contextCount;
if (context.Length == 0)
{
contextCount = totTokens;
}
else
{
//int? count = gramCounts.get(i - 1).get(context);
var gramCount = gramCounts[i - 1];
if (!gramCount.TryGetValue(context, out int?count) || count == null)
{
// We never saw this context:
backoff *= FreeTextSuggester.ALPHA;
if (Verbose)
{
Console.WriteLine(" skip: never saw context");
}
continue;
}
contextCount = count.GetValueOrDefault();
}
if (Verbose)
{
Console.WriteLine(" contextCount=" + contextCount);
}
IDictionary <string, int?> model = gramCounts[i];
// First pass, gather all predictions for this model:
if (Verbose)
{
Console.WriteLine(" find terms w/ prefix=" + tokens[tokens.Length - 1]);
}
List <Lookup.LookupResult> tmp = new List <Lookup.LookupResult>();
foreach (string term in terms)
{
if (term.StartsWith(tokens[tokens.Length - 1], StringComparison.Ordinal))
{
if (Verbose)
{
Console.WriteLine(" term=" + term);
}
if (seen.contains(term))
{
if (Verbose)
{
Console.WriteLine(" skip seen");
}
continue;
}
string ngram = (context + " " + term).Trim();
//Integer count = model.get(ngram);
if (model.TryGetValue(ngram, out int?count) && count != null)
{
// LUCENENET NOTE: We need to calculate this as decimal because when using double it can sometimes
// return numbers that are greater than long.MaxValue, which results in a negative long number.
// This is also the way it is being done in the FreeTextSuggester to work around the issue.
Lookup.LookupResult lr = new Lookup.LookupResult(ngram, (long)(long.MaxValue * ((decimal)backoff * (decimal)count / contextCount)));
tmp.Add(lr);
if (Verbose)
{
Console.WriteLine(" add tmp key='" + lr.Key + "' score=" + lr.Value);
}
}
}
}
// Second pass, trim to only top N, and fold those
// into overall suggestions:
tmp.Sort(byScoreThenKey);
if (tmp.size() > num)
{
//tmp.subList(num, tmp.size()).clear();
tmp.RemoveRange(num, tmp.size() - num);
}
foreach (Lookup.LookupResult result in tmp)
{
string key = result.Key.toString();
int idx = key.LastIndexOf(' ');
string lastToken;
if (idx != -1)
{
lastToken = key.Substring(idx + 1);
}
else
{
lastToken = key;
}
if (!seen.contains(lastToken))
{
seen.add(lastToken);
expected.Add(result);
if (Verbose)
{
Console.WriteLine(" keep key='" + result.Key + "' score=" + result.Value);
}
}
}
backoff *= FreeTextSuggester.ALPHA;
}
expected.Sort(byScoreThenKey);
if (expected.size() > num)
{
expected.RemoveRange(num, expected.size() - num);
}
// Actual:
IList <Lookup.LookupResult> actual = sug.DoLookup(query, num);
if (Verbose)
{
Console.WriteLine(" expected: " + expected);
Console.WriteLine(" actual: " + actual);
}
assertEquals(expected.ToString(), actual.ToString());
}
}
private void CreateRandomIndexes()
{
dir1 = NewDirectory();
dir2 = NewDirectory();
int numDocs = AtLeast(150);
int numTerms = TestUtil.NextInt32(Random, 1, numDocs / 5);
ISet <string> randomTerms = new JCG.HashSet <string>();
while (randomTerms.size() < numTerms)
{
randomTerms.add(TestUtil.RandomSimpleString(Random));
}
terms = new JCG.List <string>(randomTerms);
long seed = Random.NextInt64();
IndexWriterConfig iwc1 = NewIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(new Random((int)seed)));
IndexWriterConfig iwc2 = NewIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(new Random((int)seed)));
iwc2.SetMergePolicy(NewSortingMergePolicy(sort));
RandomIndexWriter iw1 = new RandomIndexWriter(new Random((int)seed), dir1, iwc1);
RandomIndexWriter iw2 = new RandomIndexWriter(new Random((int)seed), dir2, iwc2);
for (int i = 0; i < numDocs; ++i)
{
if (Random.nextInt(5) == 0 && i != numDocs - 1)
{
string term = RandomPicks.RandomFrom(Random, terms);
iw1.DeleteDocuments(new Term("s", term));
iw2.DeleteDocuments(new Term("s", term));
}
Document doc = randomDocument();
iw1.AddDocument(doc);
iw2.AddDocument(doc);
if (Random.nextInt(8) == 0)
{
iw1.Commit();
iw2.Commit();
}
}
// Make sure we have something to merge
iw1.Commit();
iw2.Commit();
Document doc2 = randomDocument();
// NOTE: don't use RIW.addDocument directly, since it sometimes commits
// which may trigger a merge, at which case forceMerge may not do anything.
// With field updates this is a problem, since the updates can go into the
// single segment in the index, and threefore the index won't be sorted.
// This hurts the assumption of the test later on, that the index is sorted
// by SortingMP.
iw1.IndexWriter.AddDocument(doc2);
iw2.IndexWriter.AddDocument(doc2);
if (DefaultCodecSupportsFieldUpdates)
{
// update NDV of docs belonging to one term (covers many documents)
long value = Random.NextInt64();
string term = RandomPicks.RandomFrom(Random, terms);
iw1.IndexWriter.UpdateNumericDocValue(new Term("s", term), "ndv", value);
iw2.IndexWriter.UpdateNumericDocValue(new Term("s", term), "ndv", value);
}
iw1.ForceMerge(1);
iw2.ForceMerge(1);
iw1.Dispose();
iw2.Dispose();
reader = DirectoryReader.Open(dir1);
sortedReader = DirectoryReader.Open(dir2);
}
