private void Open(Random random)
{
UninterruptableMonitor.Enter(syncLock);
try
{
Stream @is = null;
bool needSkip = true, isExternal = false;
long size = 0L, seekTo = 0L;
try
{
// LUCENENET: We have embedded the default file, so if that filename is passed,
// open the local resource instead of an external file.
if (path == LuceneTestCase.DEFAULT_LINE_DOCS_FILE)
{
@is = this.GetType().FindAndGetManifestResourceStream(path);
}
else
{
isExternal = true;
}
}
catch (Exception)
{
isExternal = true;
}
if (isExternal)
{
// if its not in classpath, we load it as absolute filesystem path (e.g. Hudson's home dir)
FileInfo file = new FileInfo(path);
size = file.Length;
if (path.EndsWith(".gz", StringComparison.Ordinal))
{
// if it is a gzip file, we need to use InputStream and slowly skipTo:
@is = new FileStream(file.FullName, FileMode.Open, FileAccess.Read, FileShare.Read);
}
else
{
// optimized seek using RandomAccessFile:
seekTo = RandomSeekPos(random, size);
if (LuceneTestCase.Verbose)
{
Console.WriteLine($"TEST: LineFileDocs: file seek to fp={seekTo} on open");
}
@is = new BufferedStream(new FileStream(path, FileMode.Open, FileAccess.Read, FileShare.Read)
{
Position = seekTo
});
needSkip = false;
}
}
else
{
// if the file comes from Classpath:
size = @is.Length;// available();
}
if (path.EndsWith(".gz", StringComparison.Ordinal))
{
@is = PrepareGZipStream(@is);
// guestimate:
size = (long)(size * 2.8);
}
// If we only have an InputStream, we need to seek now,
// but this seek is a scan, so very inefficient!!!
if (needSkip)
{
seekTo = RandomSeekPos(random, size);
if (LuceneTestCase.Verbose)
{
Console.WriteLine($"TEST: LineFileDocs: stream skip to fp={seekTo} on open");
}
@is.Position = seekTo;
}
// if we seeked somewhere, read until newline char
if (seekTo > 0L)
{
int b;
do
{
b = @is.ReadByte();
} while (b >= 0 && b != 13 && b != 10);
}
reader = new StreamReader(@is, Encoding.UTF8, detectEncodingFromByteOrderMarks: false, bufferSize: BUFFER_SIZE);
if (seekTo > 0L)
{
// read one more line, to make sure we are not inside a Windows linebreak (\r\n):
reader.ReadLine();
}
}
finally
{
UninterruptableMonitor.Exit(syncLock);
}
}
// LUCENENET: Called in one place, but since there is no implementation it is just wasted CPU
//internal void Recycle(DocumentsWriterPerThread dwpt)
//{
// // don't recycle DWPT by default
//}
// you cannot subclass this without being in o.a.l.index package anyway, so
// the class is already pkg-private... fix me: see LUCENE-4013
public ThreadState GetAndLock(/* Thread requestingThread, DocumentsWriter documentsWriter // LUCENENET: Not referenced */)
{
ThreadState threadState = null;
UninterruptableMonitor.Enter(this);
try
{
for (;;)
{
if (freeCount > 0)
{
// Important that we are LIFO here! This way if number of concurrent indexing threads was once high, but has now reduced, we only use a
// limited number of thread states:
threadState = freeList[freeCount - 1];
if (threadState.dwpt is null)
{
// This thread-state is not initialized, e.g. it
// was just flushed. See if we can instead find
// another free thread state that already has docs
// indexed. This way if incoming thread concurrency
// has decreased, we don't leave docs
// indefinitely buffered, tying up RAM. This
// will instead get those thread states flushed,
// freeing up RAM for larger segment flushes:
for (int i = 0; i < freeCount; i++)
{
if (freeList[i].dwpt != null)
{
// Use this one instead, and swap it with
// the un-initialized one:
ThreadState ts = freeList[i];
freeList[i] = threadState;
threadState = ts;
break;
}
}
}
freeCount--;
break;
}
else if (NumThreadStatesActive < threadStates.Length)
{
// ThreadState is already locked before return by this method:
return(NewThreadState());
}
else
{
// Wait until a thread state frees up:
try
{
UninterruptableMonitor.Wait(this);
}
catch (Exception ie) when(ie.IsInterruptedException())
{
throw new Util.ThreadInterruptedException(ie);
}
}
}
}
finally
{
UninterruptableMonitor.Exit(this);
}
// This could take time, e.g. if the threadState is [briefly] checked for flushing:
threadState.Lock();
return(threadState);
}
void ICollection.CopyTo(Array array, int index)
{
if (array is null)
{
throw new ArgumentNullException(nameof(array));
}
if (array.Rank != 1)
{
throw new ArgumentException("Only single dimensional arrays are supported for the requested action.", nameof(array));
}
//throw new ArgumentException(SR.Arg_RankMultiDimNotSupported, nameof(array));
if (array.GetLowerBound(0) != 0)
{
throw new ArgumentException("The lower bound of target array must be zero.", nameof(array));
}
//throw new ArgumentException(SR.Arg_NonZeroLowerBound, nameof(array));
if (index < 0)
{
throw new ArgumentOutOfRangeException(nameof(index), index, "Non-negative number required.");
}
//throw new ArgumentOutOfRangeException(nameof(index), index, SR.ArgumentOutOfRange_NeedNonNegNum);
if (array.Length - index < Count)
{
throw new ArgumentException("Destination array is not long enough to copy all the items in the collection. Check array index and length.");
}
//throw new ArgumentException(SR.Arg_ArrayPlusOffTooSmall);
#pragma warning disable IDE0019 // Use pattern matching
T[]? tarray = array as T[];
#pragma warning restore IDE0019 // Use pattern matching
if (tarray != null)
{
CopyTo(tarray, index);
}
else
{
#pragma warning disable IDE0019 // Use pattern matching
object?[]? objects = array as object[];
#pragma warning restore IDE0019 // Use pattern matching
if (objects is null)
{
throw new ArgumentException("Target array type is not compatible with the type of items in the collection.", nameof(array));
//throw new ArgumentException(SR.Argument_InvalidArrayType, nameof(array));
}
try
{
UninterruptableMonitor.Enter(SyncRoot);
try
{
foreach (var item in set)
{
objects[index++] = item;
}
}
finally
{
UninterruptableMonitor.Exit(SyncRoot);
}
}
catch (ArrayTypeMismatchException)
{
throw new ArgumentException("Target array type is not compatible with the type of items in the collection.", nameof(array));
//throw new ArgumentException(SR.Argument_InvalidArrayType, nameof(array));
}
}
}
internal void MarkForFullFlush()
{
DocumentsWriterDeleteQueue flushingQueue;
UninterruptableMonitor.Enter(this);
try
{
if (Debugging.AssertsEnabled)
{
Debugging.Assert(!fullFlush, "called DWFC#markForFullFlush() while full flush is still running");
Debugging.Assert(fullFlushBuffer.Count == 0, "full flush buffer should be empty: {0}", fullFlushBuffer);
}
fullFlush = true;
flushingQueue = documentsWriter.deleteQueue;
// Set a new delete queue - all subsequent DWPT will use this queue until
// we do another full flush
DocumentsWriterDeleteQueue newQueue = new DocumentsWriterDeleteQueue(flushingQueue.generation + 1);
documentsWriter.deleteQueue = newQueue;
}
finally
{
UninterruptableMonitor.Exit(this);
}
int limit = perThreadPool.NumThreadStatesActive;
for (int i = 0; i < limit; i++)
{
ThreadState next = perThreadPool.GetThreadState(i);
next.@Lock();
try
{
if (!next.IsInitialized)
{
if (closed && next.IsActive)
{
DocumentsWriterPerThreadPool.DeactivateThreadState(next); // LUCENENET specific - made method static per CA1822
}
continue;
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(next.dwpt.deleteQueue == flushingQueue ||
next.dwpt.deleteQueue == documentsWriter.deleteQueue,
" flushingQueue: {0} currentqueue: {1} perThread queue: {2} numDocsInRam: {3}",
flushingQueue, documentsWriter.deleteQueue, next.dwpt.deleteQueue, next.dwpt.NumDocsInRAM);
}
if (next.dwpt.deleteQueue != flushingQueue)
{
// this one is already a new DWPT
continue;
}
AddFlushableState(next);
}
finally
{
next.Unlock();
}
}
UninterruptableMonitor.Enter(this);
try
{
/* make sure we move all DWPT that are where concurrently marked as
* pending and moved to blocked are moved over to the flushQueue. There is
* a chance that this happens since we marking DWPT for full flush without
* blocking indexing.*/
PruneBlockedQueue(flushingQueue);
if (Debugging.AssertsEnabled)
{
Debugging.Assert(AssertBlockedFlushes(documentsWriter.deleteQueue));
}
//FlushQueue.AddAll(FullFlushBuffer);
foreach (var dwpt in fullFlushBuffer)
{
flushQueue.Enqueue(dwpt);
}
fullFlushBuffer.Clear();
UpdateStallState();
}
finally
{
UninterruptableMonitor.Exit(this);
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(AssertActiveDeleteQueue(documentsWriter.deleteQueue));
}
}
private bool AddInternal(T item, int hashcode, bool acquireLock)
{
while (true)
{
var tables = _tables;
GetBucketAndLockNo(hashcode, out int bucketNo, out int lockNo, tables.Buckets.Length, tables.Locks.Length);
var resizeDesired = false;
var lockTaken = false;
try
{
if (acquireLock)
{
UninterruptableMonitor.Enter(tables.Locks[lockNo], ref lockTaken);
}
// If the table just got resized, we may not be holding the right lock, and must retry.
// This should be a rare occurrence.
if (tables != _tables)
{
continue;
}
// Try to find this item in the bucket
Node previous = null;
for (var current = tables.Buckets[bucketNo]; current != null; current = current.Next)
{
Debug.Assert(previous is null && current == tables.Buckets[bucketNo] || previous.Next == current);
if (hashcode == current.Hashcode && _comparer.Equals(current.Item, item))
{
return(false);
}
previous = current;
}
// The item was not found in the bucket. Insert the new item.
Volatile.Write(ref tables.Buckets[bucketNo], new Node(item, hashcode, tables.Buckets[bucketNo]));
checked
{
tables.CountPerLock[lockNo]++;
}
//
// If the number of elements guarded by this lock has exceeded the budget, resize the bucket table.
// It is also possible that GrowTable will increase the budget but won't resize the bucket table.
// That happens if the bucket table is found to be poorly utilized due to a bad hash function.
//
if (tables.CountPerLock[lockNo] > _budget)
{
resizeDesired = true;
}
}
finally
{
if (lockTaken)
{
UninterruptableMonitor.Exit(tables.Locks[lockNo]);
}
}
//
// The fact that we got here means that we just performed an insertion. If necessary, we will grow the table.
//
// Concurrency notes:
// - Notice that we are not holding any locks at when calling GrowTable. This is necessary to prevent deadlocks.
// - As a result, it is possible that GrowTable will be called unnecessarily. But, GrowTable will obtain lock 0
// and then verify that the table we passed to it as the argument is still the current table.
//
if (resizeDesired)
{
GrowTable(tables);
}
return(true);
}
}
public override void Run()
{
// TODO: maybe use private thread ticktock timer, in
// case clock shift messes up nanoTime?
// LUCENENET NOTE: Time.NanoTime() is not affected by clock changes.
long lastReopenStartNS = Time.NanoTime();
//System.out.println("reopen: start");
while (!finish)
{
// TODO: try to guestimate how long reopen might
// take based on past data?
// Loop until we've waiting long enough before the
// next reopen:
while (!finish)
{
try
{
// Need lock before finding out if has waiting
bool hasWaiting;
UninterruptableMonitor.Enter(this);
try
{
// True if we have someone waiting for reopened searcher:
hasWaiting = waitingGen > searchingGen;
}
finally
{
UninterruptableMonitor.Exit(this);
}
long nextReopenStartNS = lastReopenStartNS + (hasWaiting ? targetMinStaleNS : targetMaxStaleNS);
long sleepNS = nextReopenStartNS - Time.NanoTime();
if (sleepNS > 0)
{
reopenCond.WaitOne(TimeSpan.FromMilliseconds(sleepNS / Time.MillisecondsPerNanosecond));//Convert NS to MS
}
else
{
break;
}
}
catch (Exception ie) when(ie.IsInterruptedException())
{
Thread.CurrentThread.Interrupt();
return;
}
}
if (finish)
{
break;
}
lastReopenStartNS = Time.NanoTime();
// Save the gen as of when we started the reopen; the
// listener (HandleRefresh above) copies this to
// searchingGen once the reopen completes:
refreshStartGen.Value = writer.GetAndIncrementGeneration();
try
{
manager.MaybeRefreshBlocking();
}
catch (Exception ioe) when(ioe.IsIOException())
{
throw RuntimeException.Create(ioe);
}
}
}
private int InnerPurge(IndexWriter writer)
{
if (Debugging.AssertsEnabled)
{
Debugging.Assert(purgeLock.IsHeldByCurrentThread);
}
int numPurged = 0;
while (true)
{
FlushTicket head;
bool canPublish;
UninterruptableMonitor.Enter(this);
try
{
head = queue.Count <= 0 ? null : queue.Peek();
canPublish = head != null && head.CanPublish; // do this synced
}
finally
{
UninterruptableMonitor.Exit(this);
}
if (canPublish)
{
numPurged++;
try
{
/*
* if we block on publish -> lock IW -> lock BufferedDeletes we don't block
* concurrent segment flushes just because they want to append to the queue.
* the downside is that we need to force a purge on fullFlush since ther could
* be a ticket still in the queue.
*/
head.Publish(writer);
}
finally
{
UninterruptableMonitor.Enter(this);
try
{
// finally remove the published ticket from the queue
FlushTicket poll = queue.Dequeue();
ticketCount.DecrementAndGet();
if (Debugging.AssertsEnabled)
{
Debugging.Assert(poll == head);
}
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
}
else
{
break;
}
}
return(numPurged);
}
internal void Persist()
{
UninterruptableMonitor.Enter(this);
try
{
string fileName = SNAPSHOTS_PREFIX + nextWriteGen;
IndexOutput @out = dir.CreateOutput(fileName, IOContext.DEFAULT);
bool success = false;
try
{
CodecUtil.WriteHeader(@out, CODEC_NAME, VERSION_CURRENT);
@out.WriteVInt32(m_refCounts.Count);
foreach (KeyValuePair <long, int> ent in m_refCounts)
{
@out.WriteVInt64(ent.Key);
@out.WriteVInt32(ent.Value);
}
success = true;
}
finally
{
if (!success)
{
IOUtils.DisposeWhileHandlingException(@out);
try
{
dir.DeleteFile(fileName);
}
catch (Exception e) when(e.IsException())
{
// Suppress so we keep throwing original exception
}
}
else
{
IOUtils.Dispose(@out);
}
}
dir.Sync(/*Collections.singletonList(*/ new[] { fileName } /*)*/);
if (nextWriteGen > 0)
{
string lastSaveFile = SNAPSHOTS_PREFIX + (nextWriteGen - 1);
try
{
dir.DeleteFile(lastSaveFile);
}
catch (Exception ioe) when(ioe.IsIOException())
{
// OK: likely it didn't exist
}
}
nextWriteGen++;
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
public override void EndElement(string @namespace, string simple, string qualified)
{
int elemType = GetElementType(qualified);
switch (elemType)
{
case PAGE:
// the body must be null and we either are keeping image docs or the
// title does not start with Image:
if (body != null && (outerInstance.keepImages || !title.StartsWith("Image:", StringComparison.Ordinal)))
{
string[] tmpTuple = new string[LENGTH];
tmpTuple[TITLE] = title.Replace('\t', ' ');
tmpTuple[DATE] = time.Replace('\t', ' ');
tmpTuple[BODY] = Regex.Replace(body, "[\t\n]", " ");
tmpTuple[ID] = id;
UninterruptableMonitor.Enter(this);
try
{
while (tuple != null && !stopped)
{
try
{
UninterruptableMonitor.Wait(this); //wait();
}
catch (System.Threading.ThreadInterruptedException ie)
{
throw new Util.ThreadInterruptedException(ie);
}
}
tuple = tmpTuple;
UninterruptableMonitor.Pulse(this); //notify();
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
break;
case BODY:
body = contents.ToString();
//workaround that startswith doesn't have an ignore case option, get at least 10 chars.
string startsWith = body.Substring(0, Math.Min(10, contents.Length) - 0).ToLowerInvariant();
if (startsWith.StartsWith("#redirect", StringComparison.Ordinal))
{
body = null;
}
break;
case DATE:
time = Time(contents.ToString());
break;
case TITLE:
title = contents.ToString();
break;
case ID:
//the doc id is the first one in the page. All other ids after that one can be ignored according to the schema
if (id == null)
{
id = contents.ToString();
}
break;
default:
// this element should be discarded.
break;
}
}
public void Run()
{
try
{
Sax.IXMLReader reader = new TagSoup.Parser(); //XMLReaderFactory.createXMLReader();
reader.ContentHandler = this;
reader.ErrorHandler = this;
while (!stopped)
{
Stream localFileIS = outerInstance.@is;
if (localFileIS != null)
{ // null means fileIS was closed on us
try
{
// To work around a bug in XERCES (XERCESJ-1257), we assume the XML is always UTF8, so we simply provide reader.
reader.Parse(new InputSource(IOUtils.GetDecodingReader(localFileIS, Encoding.UTF8)));
}
catch (Exception ioe) when(ioe.IsIOException())
{
UninterruptableMonitor.Enter(outerInstance);
try
{
if (localFileIS != outerInstance.@is)
{
// fileIS was closed on us, so, just fall through
}
else
{
// Exception is real
throw; // LUCENENET: CA2200: Rethrow to preserve stack details (https://docs.microsoft.com/en-us/visualstudio/code-quality/ca2200-rethrow-to-preserve-stack-details)
}
}
finally
{
UninterruptableMonitor.Exit(outerInstance);
}
}
}
UninterruptableMonitor.Enter(this);
try
{
if (stopped || !outerInstance.m_forever)
{
nmde = new NoMoreDataException();
UninterruptableMonitor.Pulse(this); //notify();
return;
}
else if (localFileIS == outerInstance.@is)
{
// If file is not already re-opened then re-open it now
outerInstance.@is = outerInstance.OpenInputStream();
}
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
}
catch (SAXException sae)
{
throw RuntimeException.Create(sae);
}
catch (Exception ioe) when(ioe.IsIOException())
{
throw RuntimeException.Create(ioe);
}
finally
{
UninterruptableMonitor.Enter(this);
try
{
threadDone = true;
UninterruptableMonitor.Pulse(this); //Notify();
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
}
public override NumericDocValues GetNumeric(FieldInfo field)
{
UninterruptableMonitor.Enter(this);
try
{
if (!numericInstances.TryGetValue(field.Number, out NumericDocValues instance))
{
string fileName = IndexFileNames.SegmentFileName(state.SegmentInfo.Name + "_" + Convert.ToString(field.Number, CultureInfo.InvariantCulture), segmentSuffix, "dat");
IndexInput input = dir.OpenInput(fileName, state.Context);
bool success = false;
try
{
var type = field.GetAttribute(legacyKey).ToLegacyDocValuesType();
//switch (Enum.Parse(typeof(LegacyDocValuesType), field.GetAttribute(LegacyKey)))
//{
if (type == LegacyDocValuesType.VAR_INTS)
{
instance = LoadVarInt32sField(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FIXED_INTS_8)
{
instance = LoadByteField(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FIXED_INTS_16)
{
instance = LoadInt16Field(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FIXED_INTS_32)
{
instance = LoadInt32Field(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FIXED_INTS_64)
{
instance = LoadInt64Field(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FLOAT_32)
{
instance = LoadSingleField(/* field, // LUCENENET: Never read */ input);
}
else if (type == LegacyDocValuesType.FLOAT_64)
{
instance = LoadDoubleField(/* field, // LUCENENET: Never read */ input);
}
else
{
throw AssertionError.Create();
}
CodecUtil.CheckEOF(input);
success = true;
}
finally
{
if (success)
{
IOUtils.Dispose(input);
}
else
{
IOUtils.DisposeWhileHandlingException(input);
}
}
numericInstances[field.Number] = instance;
}
return(instance);
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
protected override IndexSearcher GetCurrentSearcher()
{
if (Random.Next(10) == 7)
{
// NOTE: not best practice to call maybeReopen
// synchronous to your search threads, but still we
// test as apps will presumably do this for
// simplicity:
if (mgr.MaybeRefresh())
{
lifetimeMGR.Prune(pruner);
}
}
IndexSearcher s = null;
UninterruptableMonitor.Enter(pastSearchers);
try
{
while (pastSearchers.Count != 0 && Random.NextDouble() < 0.25)
{
// 1/4 of the time pull an old searcher, ie, simulate
// a user doing a follow-on action on a previous
// search (drilling down/up, clicking next/prev page,
// etc.)
long token = pastSearchers[Random.Next(pastSearchers.Count)];
s = lifetimeMGR.Acquire(token);
if (s is null)
{
// Searcher was pruned
pastSearchers.Remove(token);
}
else
{
break;
}
}
}
finally
{
UninterruptableMonitor.Exit(pastSearchers);
}
if (s is null)
{
s = mgr.Acquire();
if (s.IndexReader.NumDocs != 0)
{
long token = lifetimeMGR.Record(s);
UninterruptableMonitor.Enter(pastSearchers);
try
{
if (!pastSearchers.Contains(token))
{
pastSearchers.Add(token);
}
}
finally
{
UninterruptableMonitor.Exit(pastSearchers);
}
}
}
return(s);
}
/// <summary>
/// Reads the snapshots information from the given <see cref="Directory"/>. This
/// method can be used if the snapshots information is needed, however you
/// cannot instantiate the deletion policy (because e.g., some other process
/// keeps a lock on the snapshots directory).
/// </summary>
private void LoadPriorSnapshots()
{
UninterruptableMonitor.Enter(this);
try
{
long genLoaded = -1;
Exception ioe = null; // LUCENENET: No need to cast to IOExcpetion
IList <string> snapshotFiles = new JCG.List <string>();
foreach (string file in dir.ListAll())
{
if (file.StartsWith(SNAPSHOTS_PREFIX, StringComparison.Ordinal))
{
// LUCENENET: Optimized to not allocate a substring during the parse
long gen = Long.Parse(file, SNAPSHOTS_PREFIX.Length, file.Length - SNAPSHOTS_PREFIX.Length, radix: 10);
if (genLoaded == -1 || gen > genLoaded)
{
snapshotFiles.Add(file);
IDictionary <long, int> m = new Dictionary <long, int>();
IndexInput @in = dir.OpenInput(file, IOContext.DEFAULT);
try
{
CodecUtil.CheckHeader(@in, CODEC_NAME, VERSION_START, VERSION_START);
int count = @in.ReadVInt32();
for (int i = 0; i < count; i++)
{
long commitGen = @in.ReadVInt64();
int refCount = @in.ReadVInt32();
m[commitGen] = refCount;
}
}
catch (Exception ioe2) when(ioe2.IsIOException())
{
// Save first exception & throw in the end
if (ioe is null)
{
ioe = ioe2;
}
}
finally
{
@in.Dispose();
}
genLoaded = gen;
m_refCounts.Clear();
m_refCounts.PutAll(m);
}
}
}
if (genLoaded == -1)
{
// Nothing was loaded...
if (ioe != null)
{
// ... not for lack of trying:
ExceptionDispatchInfo.Capture(ioe).Throw(); // LUCENENET: Rethrow to preserve stack details from the original throw
}
}
else
{
if (snapshotFiles.Count > 1)
{
// Remove any broken / old snapshot files:
string curFileName = SNAPSHOTS_PREFIX + genLoaded;
foreach (string file in snapshotFiles)
{
if (!curFileName.Equals(file, StringComparison.Ordinal))
{
dir.DeleteFile(file);
}
}
}
nextWriteGen = 1 + genLoaded;
}
}
finally
{
UninterruptableMonitor.Exit(this);
}
}
public override void Run()
{
if (Verbose)
{
Console.WriteLine(Thread.CurrentThread.Name + ": launch search thread");
}
while (J2N.Time.NanoTime() / J2N.Time.MillisecondsPerNanosecond < stopTimeMS) // LUCENENET: Use NanoTime() rather than CurrentTimeMilliseconds() for more accurate/reliable results
{
try
{
IndexSearcher s = outerInstance.GetCurrentSearcher();
try
{
// Verify 1) IW is correctly setting
// diagnostics, and 2) segment warming for
// merged segments is actually happening:
foreach (AtomicReaderContext sub in s.IndexReader.Leaves)
{
SegmentReader segReader = (SegmentReader)sub.Reader;
IDictionary <string, string> diagnostics = segReader.SegmentInfo.Info.Diagnostics;
assertNotNull(diagnostics);
diagnostics.TryGetValue("source", out string source);
assertNotNull(source);
if (source.Equals("merge", StringComparison.Ordinal))
{
#if !FEATURE_CONDITIONALWEAKTABLE_ADDORUPDATE
UninterruptableMonitor.Enter(outerInstance.warmedLock);
try
{
#endif
assertTrue("sub reader " + sub + " wasn't warmed: warmed=" + outerInstance.warmed + " diagnostics=" + diagnostics + " si=" + segReader.SegmentInfo,
// LUCENENET: ConditionalWeakTable doesn't have ContainsKey, so we normalize to TryGetValue
!outerInstance.m_assertMergedSegmentsWarmed || outerInstance.warmed.TryGetValue(segReader.core, out BooleanRef _));
#if !FEATURE_CONDITIONALWEAKTABLE_ADDORUPDATE
}
finally
{
UninterruptableMonitor.Exit(outerInstance.warmedLock);
}
#endif
}
}
if (s.IndexReader.NumDocs > 0)
{
outerInstance.SmokeTestSearcher(s);
Fields fields = MultiFields.GetFields(s.IndexReader);
if (fields is null)
{
continue;
}
Terms terms = fields.GetTerms("body");
if (terms is null)
{
continue;
}
TermsEnum termsEnum = terms.GetEnumerator();
int seenTermCount = 0;
int shift;
int trigger;
if (totTermCount < 30)
{
shift = 0;
trigger = 1;
}
else
{
trigger = totTermCount / 30;
shift = Random.Next(trigger);
}
while (J2N.Time.NanoTime() / J2N.Time.MillisecondsPerNanosecond < stopTimeMS) // LUCENENET: Use NanoTime() rather than CurrentTimeMilliseconds() for more accurate/reliable results
{
if (!termsEnum.MoveNext())
{
totTermCount.Value = seenTermCount;
break;
}
seenTermCount++;
// search 30 terms
if ((seenTermCount + shift) % trigger == 0)
{
//if (VERBOSE) {
//System.out.println(Thread.currentThread().getName() + " now search body:" + term.Utf8ToString());
//}
totHits.AddAndGet(outerInstance.RunQuery(s, new TermQuery(new Term("body", termsEnum.Term))));
}
}
//if (VERBOSE) {
//System.out.println(Thread.currentThread().getName() + ": search done");
//}
}
}
finally
{
outerInstance.ReleaseSearcher(s);
}
}
catch (Exception t) when(t.IsThrowable())
{
Console.WriteLine(Thread.CurrentThread.Name + ": hit exc");
outerInstance.m_failed.Value = (true);
Console.WriteLine(t.ToString());
throw RuntimeException.Create(t);
}
}
}
/*
* FlushAllThreads is synced by IW fullFlushLock. Flushing all threads is a
* two stage operation; the caller must ensure (in try/finally) that finishFlush
* is called after this method, to release the flush lock in DWFlushControl
*/
internal bool FlushAllThreads(IndexWriter indexWriter)
{
DocumentsWriterDeleteQueue flushingDeleteQueue;
if (infoStream.IsEnabled("DW"))
{
infoStream.Message("DW", "startFullFlush");
}
UninterruptableMonitor.Enter(this);
try
{
pendingChangesInCurrentFullFlush = AnyChanges();
flushingDeleteQueue = deleteQueue;
/* Cutover to a new delete queue. this must be synced on the flush control
* otherwise a new DWPT could sneak into the loop with an already flushing
* delete queue */
flushControl.MarkForFullFlush(); // swaps the delQueue synced on FlushControl
if (Debugging.AssertsEnabled)
{
Debugging.Assert(SetFlushingDeleteQueue(flushingDeleteQueue));
}
}
finally
{
UninterruptableMonitor.Exit(this);
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(currentFullFlushDelQueue != null);
Debugging.Assert(currentFullFlushDelQueue != deleteQueue);
}
bool anythingFlushed = false;
try
{
DocumentsWriterPerThread flushingDWPT;
// Help out with flushing:
while ((flushingDWPT = flushControl.NextPendingFlush()) != null)
{
anythingFlushed |= DoFlush(flushingDWPT);
}
// If a concurrent flush is still in flight wait for it
flushControl.WaitForFlush();
if (!anythingFlushed && flushingDeleteQueue.AnyChanges()) // apply deletes if we did not flush any document
{
if (infoStream.IsEnabled("DW"))
{
infoStream.Message("DW", Thread.CurrentThread.Name + ": flush naked frozen global deletes");
}
ticketQueue.AddDeletes(flushingDeleteQueue);
}
ticketQueue.ForcePurge(indexWriter);
if (Debugging.AssertsEnabled)
{
Debugging.Assert(!flushingDeleteQueue.AnyChanges() && !ticketQueue.HasTickets);
}
}
finally
{
if (Debugging.AssertsEnabled)
{
Debugging.Assert(flushingDeleteQueue == currentFullFlushDelQueue);
}
}
return(anythingFlushed);
}
/// <summary>
/// Reads a sequence of bytes from a <see cref="Stream"/> to the given <see cref="ByteBuffer"/>, starting at the given position.
/// The <paramref name="stream"/> must be both seekable and readable.
/// </summary>
/// <param name="stream">The stream to read.</param>
/// <param name="destination">The <see cref="ByteBuffer"/> to write to.</param>
/// <param name="position">The file position at which the transfer is to begin; must be non-negative.</param>
/// <returns>The number of bytes read, possibly zero.</returns>
/// <exception cref="ArgumentNullException"><paramref name="stream"/> or <paramref name="destination"/> is <c>null</c></exception>
/// <exception cref="NotSupportedException">
/// <paramref name="stream"/> is not readable.
/// <para/>
/// -or-
/// <para/>
/// <paramref name="stream"/> is not seekable.
/// </exception>
/// <exception cref="ArgumentOutOfRangeException">
/// <paramref name="position"/> is less than 0.
/// <para/>
/// -or-
/// <para/>
/// <paramref name="position"/> is greater than the <see cref="Stream.Length"/> of the stream.
/// </exception>
/// <exception cref="IOException">An I/O error occurs.</exception>
/// <exception cref="ObjectDisposedException"><paramref name="stream"/> has already been disposed.</exception>
/// <remarks>
/// This method is atomic when used by itself, but does not synchronize with the rest of the stream methods.
/// </remarks>
public static int Read(this Stream stream, ByteBuffer destination, long position)
{
if (stream is null)
{
throw new ArgumentNullException(nameof(stream));
}
if (destination is null)
{
throw new ArgumentNullException(nameof(destination));
}
if (position < 0)
{
throw new ArgumentOutOfRangeException(nameof(position));
}
if (!stream.CanSeek)
{
throw new NotSupportedException("Stream does not support seeking.");
}
if (!stream.CanRead)
{
throw new NotSupportedException("Stream does not support reading.");
}
if (position > stream.Length)
{
return(0);
}
int read = 0;
UninterruptableMonitor.Enter(readLock);
try
{
long originalPosition = stream.Position;
stream.Seek(position, SeekOrigin.Begin);
if (destination.HasArray)
{
// If the buffer has an array, we can write to it directly and save
// an extra copy operation.
// Read from the stream
read = stream.Read(destination.Array, destination.Position, destination.Remaining);
destination.Position += read;
}
else
{
// If the buffer has no array, we must use a local buffer
byte[] buffer = new byte[destination.Remaining];
// Read from the stream
read = stream.Read(buffer, 0, buffer.Length);
// Write to the byte buffer
destination.Put(buffer, 0, read);
}
// Per Java's FileChannel.Read(), we don't want to alter the position
// of the stream, so we return it as it was originally.
stream.Seek(originalPosition, SeekOrigin.Begin);
}
finally
{
UninterruptableMonitor.Exit(readLock);
}
return(read);
}
public virtual void Collect(int doc)
{
//System.out.println("FP.collect doc=" + doc);
// If orderedGroups != null we already have collected N groups and
// can short circuit by comparing this document to the bottom group,
// without having to find what group this document belongs to.
// Even if this document belongs to a group in the top N, we'll know that
// we don't have to update that group.
// Downside: if the number of unique groups is very low, this is
// wasted effort as we will most likely be updating an existing group.
if (m_orderedGroups != null)
{
for (int compIDX = 0; ; compIDX++)
{
int c = reversed[compIDX] * comparers[compIDX].CompareBottom(doc);
if (c < 0)
{
// Definitely not competitive. So don't even bother to continue
return;
}
else if (c > 0)
{
// Definitely competitive.
break;
}
else if (compIDX == compIDXEnd)
{
// Here c=0. If we're at the last comparer, this doc is not
// competitive, since docs are visited in doc Id order, which means
// this doc cannot compete with any other document in the queue.
return;
}
}
}
// TODO: should we add option to mean "ignore docs that
// don't have the group field" (instead of stuffing them
// under null group)?
TGroupValue groupValue = GetDocGroupValue(doc);
if (!groupMap.TryGetValue(groupValue, out CollectedSearchGroup <TGroupValue> group))
{
// First time we are seeing this group, or, we've seen
// it before but it fell out of the top N and is now
// coming back
if (groupMap.Count < topNGroups)
{
// Still in startup transient: we have not
// seen enough unique groups to start pruning them;
// just keep collecting them
// Add a new CollectedSearchGroup:
CollectedSearchGroup <TGroupValue> sg = new CollectedSearchGroup <TGroupValue>();
sg.GroupValue = CopyDocGroupValue(groupValue, default);
sg.ComparerSlot = groupMap.Count;
sg.TopDoc = docBase + doc;
foreach (FieldComparer fc in comparers)
{
fc.Copy(sg.ComparerSlot, doc);
}
groupMap[sg.GroupValue] = sg;
if (groupMap.Count == topNGroups)
{
// End of startup transient: we now have max
// number of groups; from here on we will drop
// bottom group when we insert new one:
BuildSortedSet();
}
return;
}
// We already tested that the document is competitive, so replace
// the bottom group with this new group.
//CollectedSearchGroup<TGroupValue> bottomGroup = orderedGroups.PollLast();
CollectedSearchGroup <TGroupValue> bottomGroup;
UninterruptableMonitor.Enter(m_orderedGroups);
try
{
bottomGroup = m_orderedGroups.Last();
m_orderedGroups.Remove(bottomGroup);
}
finally
{
UninterruptableMonitor.Exit(m_orderedGroups);
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(m_orderedGroups.Count == topNGroups - 1);
}
groupMap.Remove(bottomGroup.GroupValue);
// reuse the removed CollectedSearchGroup
bottomGroup.GroupValue = CopyDocGroupValue(groupValue, bottomGroup.GroupValue);
bottomGroup.TopDoc = docBase + doc;
foreach (FieldComparer fc in comparers)
{
fc.Copy(bottomGroup.ComparerSlot, doc);
}
groupMap[bottomGroup.GroupValue] = bottomGroup;
m_orderedGroups.Add(bottomGroup);
if (Debugging.AssertsEnabled)
{
Debugging.Assert(m_orderedGroups.Count == topNGroups);
}
int lastComparerSlot = m_orderedGroups.Last().ComparerSlot;
foreach (FieldComparer fc in comparers)
{
fc.SetBottom(lastComparerSlot);
}
return;
}
// Update existing group:
for (int compIDX = 0; ; compIDX++)
{
FieldComparer fc = comparers[compIDX];
fc.Copy(spareSlot, doc);
int c = reversed[compIDX] * fc.Compare(group.ComparerSlot, spareSlot);
if (c < 0)
{
// Definitely not competitive.
return;
}
else if (c > 0)
{
// Definitely competitive; set remaining comparers:
for (int compIDX2 = compIDX + 1; compIDX2 < comparers.Length; compIDX2++)
{
comparers[compIDX2].Copy(spareSlot, doc);
}
break;
}
else if (compIDX == compIDXEnd)
{
// Here c=0. If we're at the last comparer, this doc is not
// competitive, since docs are visited in doc Id order, which means
// this doc cannot compete with any other document in the queue.
return;
}
}
// Remove before updating the group since lookup is done via comparers
// TODO: optimize this
CollectedSearchGroup <TGroupValue> prevLast;
if (m_orderedGroups != null)
{
UninterruptableMonitor.Enter(m_orderedGroups);
try
{
prevLast = m_orderedGroups.Last();
m_orderedGroups.Remove(group);
}
finally
{
UninterruptableMonitor.Exit(m_orderedGroups);
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(m_orderedGroups.Count == topNGroups - 1);
}
}
else
{
prevLast = null;
}
group.TopDoc = docBase + doc;
// Swap slots
int tmp = spareSlot;
spareSlot = group.ComparerSlot;
group.ComparerSlot = tmp;
// Re-add the changed group
if (m_orderedGroups != null)
{
m_orderedGroups.Add(group);
if (Debugging.AssertsEnabled)
{
Debugging.Assert(m_orderedGroups.Count == topNGroups);
}
var newLast = m_orderedGroups.Last();
// If we changed the value of the last group, or changed which group was last, then update bottom:
if (group == newLast || prevLast != newLast)
{
foreach (FieldComparer fc in comparers)
{
fc.SetBottom(newLast.ComparerSlot);
}
}
}
}
