org.apache.lucene.index

Class IndexWriter

  • All Implemented Interfaces:
    Closeable, AutoCloseable, TwoPhaseCommit, Accountable
    Direct Known Subclasses:
    SolrIndexWriter


    public class IndexWriter
    extends Object
    implements Closeable, TwoPhaseCommit, Accountable
    An IndexWriter creates and maintains an index.

    The IndexWriterConfig.OpenMode option on IndexWriterConfig.setOpenMode(OpenMode) determines whether a new index is created, or whether an existing index is opened. Note that you can open an index with IndexWriterConfig.OpenMode.CREATE even while readers are using the index. The old readers will continue to search the "point in time" snapshot they had opened, and won't see the newly created index until they re-open. If IndexWriterConfig.OpenMode.CREATE_OR_APPEND is used IndexWriter will create a new index if there is not already an index at the provided path and otherwise open the existing index.

    In either case, documents are added with addDocument and removed with deleteDocuments(Term...) or deleteDocuments(Query...). A document can be updated with updateDocument (which just deletes and then adds the entire document). When finished adding, deleting and updating documents, close should be called.

    Each method that changes the index returns a long sequence number, which expresses the effective order in which each change was applied. commit() also returns a sequence number, describing which changes are in the commit point and which are not. Sequence numbers are transient (not saved into the index in any way) and only valid within a single IndexWriter instance.

    These changes are buffered in memory and periodically flushed to the Directory (during the above method calls). A flush is triggered when there are enough added documents since the last flush. Flushing is triggered either by RAM usage of the documents (see IndexWriterConfig.setRAMBufferSizeMB(double)) or the number of added documents (see IndexWriterConfig.setMaxBufferedDocs(int)). The default is to flush when RAM usage hits IndexWriterConfig.DEFAULT_RAM_BUFFER_SIZE_MB MB. For best indexing speed you should flush by RAM usage with a large RAM buffer. Additionally, if IndexWriter reaches the configured number of buffered deletes (see IndexWriterConfig.setMaxBufferedDeleteTerms(int)) the deleted terms and queries are flushed and applied to existing segments. In contrast to the other flush options IndexWriterConfig.setRAMBufferSizeMB(double) and IndexWriterConfig.setMaxBufferedDocs(int), deleted terms won't trigger a segment flush. Note that flushing just moves the internal buffered state in IndexWriter into the index, but these changes are not visible to IndexReader until either commit() or close() is called. A flush may also trigger one or more segment merges which by default run with a background thread so as not to block the addDocument calls (see below for changing the MergeScheduler).

    Opening an IndexWriter creates a lock file for the directory in use. Trying to open another IndexWriter on the same directory will lead to a LockObtainFailedException.

    Expert: IndexWriter allows an optional IndexDeletionPolicy implementation to be specified. You can use this to control when prior commits are deleted from the index. The default policy is KeepOnlyLastCommitDeletionPolicy which removes all prior commits as soon as a new commit is done. Creating your own policy can allow you to explicitly keep previous "point in time" commits alive in the index for some time, either because this is useful for your application, or to give readers enough time to refresh to the new commit without having the old commit deleted out from under them. The latter is necessary when multiple computers take turns opening their own IndexWriter and IndexReaders against a single shared index mounted via remote filesystems like NFS which do not support "delete on last close" semantics. A single computer accessing an index via NFS is fine with the default deletion policy since NFS clients emulate "delete on last close" locally. That said, accessing an index via NFS will likely result in poor performance compared to a local IO device.

    Expert: IndexWriter allows you to separately change the MergePolicy and the MergeScheduler. The MergePolicy is invoked whenever there are changes to the segments in the index. Its role is to select which merges to do, if any, and return a MergePolicy.MergeSpecification describing the merges. The default is LogByteSizeMergePolicy. Then, the MergeScheduler is invoked with the requested merges and it decides when and how to run the merges. The default is ConcurrentMergeScheduler.

    NOTE: if you hit a VirtualMachineError, or disaster strikes during a checkpoint then IndexWriter will close itself. This is a defensive measure in case any internal state (buffered documents, deletions, reference counts) were corrupted. Any subsequent calls will throw an AlreadyClosedException.

    NOTE: IndexWriter instances are completely thread safe, meaning multiple threads can call any of its methods, concurrently. If your application requires external synchronization, you should not synchronize on the IndexWriter instance as this may cause deadlock; use your own (non-Lucene) objects instead.

    NOTE: If you call Thread.interrupt() on a thread that's within IndexWriter, IndexWriter will try to catch this (eg, if it's in a wait() or Thread.sleep()), and will then throw the unchecked exception ThreadInterruptedException and clear the interrupt status on the thread.

    • Constructor Detail

      • IndexWriter

        public IndexWriter(Directory d,
                           IndexWriterConfig conf)
                    throws IOException
        Constructs a new IndexWriter per the settings given in conf. If you want to make "live" changes to this writer instance, use getConfig().

        NOTE: after ths writer is created, the given configuration instance cannot be passed to another writer.

        Parameters:
        d - the index directory. The index is either created or appended according conf.getOpenMode().
        conf - the configuration settings according to which IndexWriter should be initialized.
        Throws:
        IOException - if the directory cannot be read/written to, or if it does not exist and conf.getOpenMode() is OpenMode.APPEND or if there is any other low-level IO error
    • Method Detail

      • numDeletedDocs

        public int numDeletedDocs(SegmentCommitInfo info)
        Obtain the number of deleted docs for a pooled reader. If the reader isn't being pooled, the segmentInfo's delCount is returned.
      • ensureOpen

        protected final void ensureOpen(boolean failIfClosing)
                                 throws AlreadyClosedException
        Used internally to throw an AlreadyClosedException if this IndexWriter has been closed or is in the process of closing.
        Parameters:
        failIfClosing - if true, also fail when IndexWriter is in the process of closing (closing=true) but not yet done closing ( closed=false)
        Throws:
        AlreadyClosedException - if this IndexWriter is closed or in the process of closing
      • close

        public void close()
                   throws IOException
        Closes all open resources and releases the write lock. If LiveIndexWriterConfig.commitOnClose is true, this will attempt to gracefully shut down by writing any changes, waiting for any running merges, committing, and closing. In this case, note that:
        • If you called prepareCommit but failed to call commit, this method will throw IllegalStateException and the IndexWriter will not be closed.
        • If this method throws any other exception, the IndexWriter will be closed, but changes may have been lost.

        Note that this may be a costly operation, so, try to re-use a single writer instead of closing and opening a new one. See commit() for caveats about write caching done by some IO devices.

        NOTE: You must ensure no other threads are still making changes at the same time that this method is invoked.

        Specified by:
        close in interface Closeable
        Specified by:
        close in interface AutoCloseable
        Throws:
        IOException
      • maxDoc

        public int maxDoc()
        Returns total number of docs in this index, including docs not yet flushed (still in the RAM buffer), not counting deletions.
        See Also:
        numDocs()
      • numDocs

        public int numDocs()
        Returns total number of docs in this index, including docs not yet flushed (still in the RAM buffer), and including deletions. NOTE: buffered deletions are not counted. If you really need these to be counted you should call commit() first.
        See Also:
        numDocs()
      • hasDeletions

        public boolean hasDeletions()
        Returns true if this index has deletions (including buffered deletions). Note that this will return true if there are buffered Term/Query deletions, even if it turns out those buffered deletions don't match any documents.
      • addDocument

        public long addDocument(Iterable<? extends IndexableField> doc)
                         throws IOException
        Adds a document to this index.

        Note that if an Exception is hit (for example disk full) then the index will be consistent, but this document may not have been added. Furthermore, it's possible the index will have one segment in non-compound format even when using compound files (when a merge has partially succeeded).

        This method periodically flushes pending documents to the Directory (see above), and also periodically triggers segment merges in the index according to the MergePolicy in use.

        Merges temporarily consume space in the directory. The amount of space required is up to 1X the size of all segments being merged, when no readers/searchers are open against the index, and up to 2X the size of all segments being merged when readers/searchers are open against the index (see forceMerge(int) for details). The sequence of primitive merge operations performed is governed by the merge policy.

        Note that each term in the document can be no longer than MAX_TERM_LENGTH in bytes, otherwise an IllegalArgumentException will be thrown.

        Note that it's possible to create an invalid Unicode string in java if a UTF16 surrogate pair is malformed. In this case, the invalid characters are silently replaced with the Unicode replacement character U+FFFD.

        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • addDocuments

        public long addDocuments(Iterable<? extends Iterable<? extends IndexableField>> docs)
                          throws IOException
        Atomically adds a block of documents with sequentially assigned document IDs, such that an external reader will see all or none of the documents.

        WARNING: the index does not currently record which documents were added as a block. Today this is fine, because merging will preserve a block. The order of documents within a segment will be preserved, even when child documents within a block are deleted. Most search features (like result grouping and block joining) require you to mark documents; when these documents are deleted these search features will not work as expected. Obviously adding documents to an existing block will require you the reindex the entire block.

        However it's possible that in the future Lucene may merge more aggressively re-order documents (for example, perhaps to obtain better index compression), in which case you may need to fully re-index your documents at that time.

        See addDocument(Iterable) for details on index and IndexWriter state after an Exception, and flushing/merging temporary free space requirements.

        NOTE: tools that do offline splitting of an index (for example, IndexSplitter in contrib) or re-sorting of documents (for example, IndexSorter in contrib) are not aware of these atomically added documents and will likely break them up. Use such tools at your own risk!

        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • tryDeleteDocument

        public long tryDeleteDocument(IndexReader readerIn,
                                      int docID)
                               throws IOException
        Expert: attempts to delete by document ID, as long as the provided reader is a near-real-time reader (from DirectoryReader.open(IndexWriter)). If the provided reader is an NRT reader obtained from this writer, and its segment has not been merged away, then the delete succeeds and this method returns a valid (> 0) sequence number; else, it returns -1 and the caller must then separately delete by Term or Query. NOTE: this method can only delete documents visible to the currently open NRT reader. If you need to delete documents indexed after opening the NRT reader you must use deleteDocuments(Term...)).
        Throws:
        IOException
      • deleteDocuments

        public long deleteDocuments(Term... terms)
                             throws IOException
        Deletes the document(s) containing any of the terms. All given deletes are applied and flushed atomically at the same time.
        Parameters:
        terms - array of terms to identify the documents to be deleted
        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • deleteDocuments

        public long deleteDocuments(Query... queries)
                             throws IOException
        Deletes the document(s) matching any of the provided queries. All given deletes are applied and flushed atomically at the same time.
        Parameters:
        queries - array of queries to identify the documents to be deleted
        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • updateDocument

        public long updateDocument(Term term,
                                   Iterable<? extends IndexableField> doc)
                            throws IOException
        Updates a document by first deleting the document(s) containing term and then adding the new document. The delete and then add are atomic as seen by a reader on the same index (flush may happen only after the add).
        Parameters:
        term - the term to identify the document(s) to be deleted
        doc - the document to be added
        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • updateBinaryDocValue

        public long updateBinaryDocValue(Term term,
                                         String field,
                                         BytesRef value)
                                  throws IOException
        Updates a document's BinaryDocValues for field to the given value. You can only update fields that already exist in the index, not add new fields through this method.

        NOTE: this method currently replaces the existing value of all affected documents with the new value.

        Parameters:
        term - the term to identify the document(s) to be updated
        field - field name of the BinaryDocValues field
        value - new value for the field
        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • updateDocValues

        public long updateDocValues(Term term,
                                    Field... updates)
                             throws IOException
        Updates documents' DocValues fields to the given values. Each field update is applied to the set of documents that are associated with the Term to the same value. All updates are atomically applied and flushed together.
        Parameters:
        updates - the updates to apply
        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
      • forceMerge

        public void forceMerge(int maxNumSegments)
                        throws IOException
        Forces merge policy to merge segments until there are <= maxNumSegments. The actual merges to be executed are determined by the MergePolicy.

        This is a horribly costly operation, especially when you pass a small maxNumSegments; usually you should only call this if the index is static (will no longer be changed).

        Note that this requires free space that is proportional to the size of the index in your Directory: 2X if you are not using compound file format, and 3X if you are. For example, if your index size is 10 MB then you need an additional 20 MB free for this to complete (30 MB if you're using compound file format). This is also affected by the Codec that is used to execute the merge, and may result in even a bigger index. Also, it's best to call commit() afterwards, to allow IndexWriter to free up disk space.

        If some but not all readers re-open while merging is underway, this will cause > 2X temporary space to be consumed as those new readers will then hold open the temporary segments at that time. It is best not to re-open readers while merging is running.

        The actual temporary usage could be much less than these figures (it depends on many factors).

        In general, once this completes, the total size of the index will be less than the size of the starting index. It could be quite a bit smaller (if there were many pending deletes) or just slightly smaller.

        If an Exception is hit, for example due to disk full, the index will not be corrupted and no documents will be lost. However, it may have been partially merged (some segments were merged but not all), and it's possible that one of the segments in the index will be in non-compound format even when using compound file format. This will occur when the Exception is hit during conversion of the segment into compound format.

        This call will merge those segments present in the index when the call started. If other threads are still adding documents and flushing segments, those newly created segments will not be merged unless you call forceMerge again.

        Parameters:
        maxNumSegments - maximum number of segments left in the index after merging finishes
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
        See Also:
        MergePolicy.findMerges(org.apache.lucene.index.MergeTrigger, org.apache.lucene.index.SegmentInfos, org.apache.lucene.index.IndexWriter)
      • forceMerge

        public void forceMerge(int maxNumSegments,
                               boolean doWait)
                        throws IOException
        Just like forceMerge(int), except you can specify whether the call should block until all merging completes. This is only meaningful with a MergeScheduler that is able to run merges in background threads.
        Throws:
        IOException
      • forceMergeDeletes

        public void forceMergeDeletes()
                               throws IOException
        Forces merging of all segments that have deleted documents. The actual merges to be executed are determined by the MergePolicy. For example, the default TieredMergePolicy will only pick a segment if the percentage of deleted docs is over 10%.

        This is often a horribly costly operation; rarely is it warranted.

        To see how many deletions you have pending in your index, call IndexReader.numDeletedDocs().

        NOTE: this method first flushes a new segment (if there are indexed documents), and applies all buffered deletes.

        Throws:
        IOException
      • maybeMerge

        public final void maybeMerge()
                              throws IOException
        Expert: asks the mergePolicy whether any merges are necessary now and if so, runs the requested merges and then iterate (test again if merges are needed) until no more merges are returned by the mergePolicy. Explicit calls to maybeMerge() are usually not necessary. The most common case is when merge policy parameters have changed. This method will call the MergePolicy with MergeTrigger.EXPLICIT.
        Throws:
        IOException
      • getMergingSegments

        public Collection<SegmentCommitInfogetMergingSegments()
        Expert: to be used by a MergePolicy to avoid selecting merges for segments already being merged. The returned collection is not cloned, and thus is only safe to access if you hold IndexWriter's lock (which you do when IndexWriter invokes the MergePolicy).

        Do not alter the returned collection!

      • hasPendingMerges

        public boolean hasPendingMerges()
        Expert: returns true if there are merges waiting to be scheduled.
      • rollback

        public void rollback()
                      throws IOException
        Close the IndexWriter without committing any changes that have occurred since the last commit (or since it was opened, if commit hasn't been called). This removes any temporary files that had been created, after which the state of the index will be the same as it was when commit() was last called or when this writer was first opened. This also clears a previous call to prepareCommit().
        Specified by:
        rollback in interface TwoPhaseCommit
        Throws:
        IOException - if there is a low-level IO error
      • addIndexes

        public long addIndexes(Directory... dirs)
                        throws IOException
        Adds all segments from an array of indexes into this index.

        This may be used to parallelize batch indexing. A large document collection can be broken into sub-collections. Each sub-collection can be indexed in parallel, on a different thread, process or machine. The complete index can then be created by merging sub-collection indexes with this method.

        NOTE: this method acquires the write lock in each directory, to ensure that no IndexWriter is currently open or tries to open while this is running.

        This method is transactional in how Exceptions are handled: it does not commit a new segments_N file until all indexes are added. This means if an Exception occurs (for example disk full), then either no indexes will have been added or they all will have been.

        Note that this requires temporary free space in the Directory up to 2X the sum of all input indexes (including the starting index). If readers/searchers are open against the starting index, then temporary free space required will be higher by the size of the starting index (see forceMerge(int) for details).

        This requires this index not be among those to be added.

        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
        IllegalArgumentException - if addIndexes would cause the index to exceed MAX_DOCS, or if the indoming index sort does not match this index's index sort
      • addIndexes

        public long addIndexes(CodecReader... readers)
                        throws IOException
        Merges the provided indexes into this index.

        The provided IndexReaders are not closed.

        See addIndexes(org.apache.lucene.store.Directory...) for details on transactional semantics, temporary free space required in the Directory, and non-CFS segments on an Exception.

        NOTE: empty segments are dropped by this method and not added to this index.

        NOTE: this method merges all given LeafReaders in one merge. If you intend to merge a large number of readers, it may be better to call this method multiple times, each time with a small set of readers. In principle, if you use a merge policy with a mergeFactor or maxMergeAtOnce parameter, you should pass that many readers in one call.

        Returns:
        The sequence number for this operation
        Throws:
        CorruptIndexException - if the index is corrupt
        IOException - if there is a low-level IO error
        IllegalArgumentException - if addIndexes would cause the index to exceed MAX_DOCS
      • doAfterFlush

        protected void doAfterFlush()
                             throws IOException
        A hook for extending classes to execute operations after pending added and deleted documents have been flushed to the Directory but before the change is committed (new segments_N file written).
        Throws:
        IOException
      • doBeforeFlush

        protected void doBeforeFlush()
                              throws IOException
        A hook for extending classes to execute operations before pending added and deleted documents are flushed to the Directory.
        Throws:
        IOException
      • prepareCommit

        public final long prepareCommit()
                                 throws IOException

        Expert: prepare for commit. This does the first phase of 2-phase commit. This method does all steps necessary to commit changes since this writer was opened: flushes pending added and deleted docs, syncs the index files, writes most of next segments_N file. After calling this you must call either commit() to finish the commit, or rollback() to revert the commit and undo all changes done since the writer was opened.

        You can also just call commit() directly without prepareCommit first in which case that method will internally call prepareCommit.

        Specified by:
        prepareCommit in interface TwoPhaseCommit
        Returns:
        The sequence number of the last operation in the commit. All sequence numbers <= this value will be reflected in the commit, and all others will not.
        Throws:
        IOException
      • setLiveCommitData

        public final void setLiveCommitData(Iterable<Map.Entry<String,String>> commitUserData)
        Sets the iterator to provide the commit user data map at commit time. Calling this method is considered a committable change and will be committed even if there are no other changes this writer. Note that you must call this method before prepareCommit(). Otherwise it won't be included in the follow-on commit().

        NOTE: the iterator is late-binding: it is only visited once all documents for the commit have been written to their segments, before the next segments_N file is written

      • commit

        public final long commit()
                          throws IOException

        Commits all pending changes (added and deleted documents, segment merges, added indexes, etc.) to the index, and syncs all referenced index files, such that a reader will see the changes and the index updates will survive an OS or machine crash or power loss. Note that this does not wait for any running background merges to finish. This may be a costly operation, so you should test the cost in your application and do it only when really necessary.

        Note that this operation calls Directory.sync on the index files. That call should not return until the file contents and metadata are on stable storage. For FSDirectory, this calls the OS's fsync. But, beware: some hardware devices may in fact cache writes even during fsync, and return before the bits are actually on stable storage, to give the appearance of faster performance. If you have such a device, and it does not have a battery backup (for example) then on power loss it may still lose data. Lucene cannot guarantee consistency on such devices.

        If nothing was committed, because there were no pending changes, this returns -1. Otherwise, it returns the sequence number such that all indexing operations prior to this sequence will be included in the commit point, and all other operations will not.

        Specified by:
        commit in interface TwoPhaseCommit
        Returns:
        The sequence number of the last operation in the commit. All sequence numbers <= this value will be reflected in the commit, and all others will not.
        Throws:
        IOException
        See Also:
        prepareCommit()
      • hasUncommittedChanges

        public final boolean hasUncommittedChanges()
        Returns true if there may be changes that have not been committed. There are cases where this may return true when there are no actual "real" changes to the index, for example if you've deleted by Term or Query but that Term or Query does not match any documents. Also, if a merge kicked off as a result of flushing a new segment during commit(), or a concurrent merged finished, this method may return true right after you had just called commit().
      • numRamDocs

        public final int numRamDocs()
        Expert: Return the number of documents currently buffered in RAM.
      • isLocked

        @Deprecated
        public static boolean isLocked(Directory directory)
                                            throws IOException
        Deprecated. Use of this method can only lead to race conditions. Try to actually obtain a lock instead.
        Returns true iff the index in the named directory is currently locked.
        Parameters:
        directory - the directory to check for a lock
        Throws:
        IOException - if there is a low-level IO error
      • getTragicException

        public Throwable getTragicException()
        If this IndexWriter was closed as a side-effect of a tragic exception, e.g. disk full while flushing a new segment, this returns the root cause exception. Otherwise (no tragic exception has occurred) it returns null.
      • isOpen

        public boolean isOpen()
        Returns true if this IndexWriter is still open.
      • deleteUnusedFiles

        public void deleteUnusedFiles()
                               throws IOException
        Expert: remove any index files that are no longer used.

        IndexWriter normally deletes unused files itself, during indexing. However, on Windows, which disallows deletion of open files, if there is a reader open on the index then those files cannot be deleted. This is fine, because IndexWriter will periodically retry the deletion.

        However, IndexWriter doesn't try that often: only on open, close, flushing a new segment, and finishing a merge. If you don't do any of these actions with your IndexWriter, you'll see the unused files linger. If that's a problem, call this method to delete them (once you've closed the open readers that were preventing their deletion).

        In addition, you can call this method to delete unreferenced index commits. This might be useful if you are using an IndexDeletionPolicy which holds onto index commits until some criteria are met, but those commits are no longer needed. Otherwise, those commits will be deleted the next time commit() is called.

        Throws:
        IOException
      • getMaxCompletedSequenceNumber

        public long getMaxCompletedSequenceNumber()
        Returns the highest sequence number across all completed operations, or 0 if no operations have finished yet. Still in-flight operations (in other threads) are not counted until they finish.