List of usage examples for org.apache.lucene.util BytesRef deepCopyOf
public static BytesRef deepCopyOf(BytesRef other)
other The returned BytesRef will have a length of other.length and an offset of zero.
From source file:org.apache.solr.request.UnInvertedField.java
License:Apache License
@Override protected void visitTerm(TermsEnum te, int termNum) throws IOException { if (termNum >= maxTermCounts.length) { // resize by doubling - for very large number of unique terms, expanding // by 4K and resultant GC will dominate uninvert times. Resize at end if material int[] newMaxTermCounts = new int[maxTermCounts.length * 2]; System.arraycopy(maxTermCounts, 0, newMaxTermCounts, 0, termNum); maxTermCounts = newMaxTermCounts; }//from ww w.j a va 2 s .c o m final BytesRef term = te.term(); if (te.docFreq() > maxTermDocFreq) { TopTerm topTerm = new TopTerm(); topTerm.term = BytesRef.deepCopyOf(term); topTerm.termNum = termNum; bigTerms.put(topTerm.termNum, topTerm); if (deState == null) { deState = new SolrIndexSearcher.DocsEnumState(); deState.fieldName = field; deState.liveDocs = searcher.getAtomicReader().getLiveDocs(); deState.termsEnum = te; // TODO: check for MultiTermsEnum in SolrIndexSearcher could now fail? deState.docsEnum = docsEnum; deState.minSetSizeCached = maxTermDocFreq; } docsEnum = deState.docsEnum; DocSet set = searcher.getDocSet(deState); maxTermCounts[termNum] = set.size(); } }
From source file:org.apache.solr.schema.CollationField.java
License:Apache License
/** * analyze the range with the analyzer, instead of the collator. * because jdk collators might not be thread safe (when they are * its just that all methods are synced), this keeps things * simple (we already have a threadlocal clone in the reused TS) *///from w w w . j a va 2 s . co m private BytesRef analyzeRangePart(String field, String part) { TokenStream source = null; try { source = analyzer.tokenStream(field, part); source.reset(); TermToBytesRefAttribute termAtt = source.getAttribute(TermToBytesRefAttribute.class); BytesRef bytes = termAtt.getBytesRef(); // we control the analyzer here: most errors are impossible if (!source.incrementToken()) throw new IllegalArgumentException("analyzer returned no terms for range part: " + part); termAtt.fillBytesRef(); assert !source.incrementToken(); source.end(); return BytesRef.deepCopyOf(bytes); } catch (IOException e) { throw new RuntimeException("Unable to analyze range part: " + part, e); } finally { IOUtils.closeQuietly(source); } }
From source file:org.apache.solr.schema.EntityTextField.java
License:Apache License
public static BytesRef analyzeMultiTerm(String field, String part, Analyzer analyzerIn) { if (part == null || analyzerIn == null) return null; TokenStream source = null;//from w w w. j a v a 2s. co m try { source = analyzerIn.tokenStream(field, part); source.reset(); TermToBytesRefAttribute termAtt = source.getAttribute(TermToBytesRefAttribute.class); BytesRef bytes = termAtt.getBytesRef(); if (!source.incrementToken()) throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, "analyzer returned no terms for multiTerm term: " + part); termAtt.fillBytesRef(); if (source.incrementToken()) throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, "analyzer returned too many terms for multiTerm term: " + part); source.end(); return BytesRef.deepCopyOf(bytes); } catch (IOException e) { throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, "error analyzing range part: " + part, e); } finally { IOUtils.closeWhileHandlingException(source); } }
From source file:org.apache.solr.schema.ICUCollationField.java
License:Apache License
/** * analyze the range with the analyzer, instead of the collator. * because icu collators are not thread safe, this keeps things * simple (we already have a threadlocal clone in the reused TS) */// w ww.java2s . c om private BytesRef analyzeRangePart(String field, String part) { TokenStream source = null; try { source = analyzer.tokenStream(field, part); source.reset(); TermToBytesRefAttribute termAtt = source.getAttribute(TermToBytesRefAttribute.class); BytesRef bytes = termAtt.getBytesRef(); // we control the analyzer here: most errors are impossible if (!source.incrementToken()) throw new IllegalArgumentException("analyzer returned no terms for range part: " + part); termAtt.fillBytesRef(); assert !source.incrementToken(); source.end(); return BytesRef.deepCopyOf(bytes); } catch (IOException e) { throw new RuntimeException("Unable analyze range part: " + part, e); } finally { IOUtils.closeQuietly(source); } }
From source file:org.apache.solr.search.SolrDocumentFetcher.java
License:Apache License
/** * This will fetch and add the docValues fields to a given SolrDocument/SolrInputDocument * * @param doc/*from w ww. j a va2s . c om*/ * A SolrDocument or SolrInputDocument instance where docValues will be added * @param docid * The lucene docid of the document to be populated * @param fields * The list of docValues fields to be decorated */ public void decorateDocValueFields(@SuppressWarnings("rawtypes") SolrDocumentBase doc, int docid, Set<String> fields) throws IOException { final List<LeafReaderContext> leafContexts = searcher.getLeafContexts(); final int subIndex = ReaderUtil.subIndex(docid, leafContexts); final int localId = docid - leafContexts.get(subIndex).docBase; final LeafReader leafReader = leafContexts.get(subIndex).reader(); for (String fieldName : fields) { final SchemaField schemaField = searcher.getSchema().getFieldOrNull(fieldName); if (schemaField == null || !schemaField.hasDocValues() || doc.containsKey(fieldName)) { log.warn("Couldn't decorate docValues for field: [{}], schemaField: [{}]", fieldName, schemaField); continue; } FieldInfo fi = searcher.getFieldInfos().fieldInfo(fieldName); if (fi == null) { continue; // Searcher doesn't have info about this field, hence ignore it. } final DocValuesType dvType = fi.getDocValuesType(); switch (dvType) { case NUMERIC: final NumericDocValues ndv = leafReader.getNumericDocValues(fieldName); if (ndv == null) { continue; } Long val; if (ndv.advanceExact(localId)) { val = ndv.longValue(); } else { continue; } Object newVal = val; if (schemaField.getType().isPointField()) { // TODO: Maybe merge PointField with TrieFields here NumberType type = schemaField.getType().getNumberType(); switch (type) { case INTEGER: newVal = val.intValue(); break; case LONG: newVal = val.longValue(); break; case FLOAT: newVal = Float.intBitsToFloat(val.intValue()); break; case DOUBLE: newVal = Double.longBitsToDouble(val); break; case DATE: newVal = new Date(val); break; default: throw new AssertionError("Unexpected PointType: " + type); } } else { if (schemaField.getType() instanceof TrieIntField) { newVal = val.intValue(); } else if (schemaField.getType() instanceof TrieFloatField) { newVal = Float.intBitsToFloat(val.intValue()); } else if (schemaField.getType() instanceof TrieDoubleField) { newVal = Double.longBitsToDouble(val); } else if (schemaField.getType() instanceof TrieDateField) { newVal = new Date(val); } else if (schemaField.getType() instanceof EnumField) { newVal = ((EnumField) schemaField.getType()).intValueToStringValue(val.intValue()); } } doc.addField(fieldName, newVal); break; case BINARY: BinaryDocValues bdv = leafReader.getBinaryDocValues(fieldName); if (bdv == null) { continue; } BytesRef value; if (bdv.advanceExact(localId)) { value = BytesRef.deepCopyOf(bdv.binaryValue()); } else { continue; } doc.addField(fieldName, value); break; case SORTED: SortedDocValues sdv = leafReader.getSortedDocValues(fieldName); if (sdv == null) { continue; } if (sdv.advanceExact(localId)) { final BytesRef bRef = sdv.binaryValue(); // Special handling for Boolean fields since they're stored as 'T' and 'F'. if (schemaField.getType() instanceof BoolField) { doc.addField(fieldName, schemaField.getType().toObject(schemaField, bRef)); } else { doc.addField(fieldName, bRef.utf8ToString()); } } break; case SORTED_NUMERIC: final SortedNumericDocValues numericDv = leafReader.getSortedNumericDocValues(fieldName); NumberType type = schemaField.getType().getNumberType(); if (numericDv != null) { if (numericDv.advance(localId) == localId) { final List<Object> outValues = new ArrayList<Object>(numericDv.docValueCount()); for (int i = 0; i < numericDv.docValueCount(); i++) { long number = numericDv.nextValue(); switch (type) { case INTEGER: outValues.add((int) number); break; case LONG: outValues.add(number); break; case FLOAT: outValues.add(NumericUtils.sortableIntToFloat((int) number)); break; case DOUBLE: outValues.add(NumericUtils.sortableLongToDouble(number)); break; case DATE: outValues.add(new Date(number)); break; default: throw new AssertionError("Unexpected PointType: " + type); } } assert outValues.size() > 0; doc.addField(fieldName, outValues); } } case SORTED_SET: final SortedSetDocValues values = leafReader.getSortedSetDocValues(fieldName); if (values != null && values.getValueCount() > 0) { if (values.advance(localId) == localId) { final List<Object> outValues = new LinkedList<>(); for (long ord = values.nextOrd(); ord != SortedSetDocValues.NO_MORE_ORDS; ord = values .nextOrd()) { value = values.lookupOrd(ord); outValues.add(schemaField.getType().toObject(schemaField, value)); } assert outValues.size() > 0; doc.addField(fieldName, outValues); } } case NONE: break; } } }
From source file:org.apache.solr.search.SolrIndexSearcher.java
License:Apache License
/** lucene.internal */ public DocSet getDocSet(DocsEnumState deState) throws IOException { int largestPossible = deState.termsEnum.docFreq(); boolean useCache = filterCache != null && largestPossible >= deState.minSetSizeCached; TermQuery key = null;//from w ww .j a v a 2 s.c om if (useCache) { key = new TermQuery(new Term(deState.fieldName, BytesRef.deepCopyOf(deState.termsEnum.term()))); DocSet result = filterCache.get(key); if (result != null) return result; } int smallSetSize = maxDoc() >> 6; int scratchSize = Math.min(smallSetSize, largestPossible); if (deState.scratch == null || deState.scratch.length < scratchSize) deState.scratch = new int[scratchSize]; final int[] docs = deState.scratch; int upto = 0; int bitsSet = 0; OpenBitSet obs = null; DocsEnum docsEnum = deState.termsEnum.docs(deState.liveDocs, deState.docsEnum, DocsEnum.FLAG_NONE); if (deState.docsEnum == null) { deState.docsEnum = docsEnum; } if (docsEnum instanceof MultiDocsEnum) { MultiDocsEnum.EnumWithSlice[] subs = ((MultiDocsEnum) docsEnum).getSubs(); int numSubs = ((MultiDocsEnum) docsEnum).getNumSubs(); for (int subindex = 0; subindex < numSubs; subindex++) { MultiDocsEnum.EnumWithSlice sub = subs[subindex]; if (sub.docsEnum == null) continue; int base = sub.slice.start; int docid; if (largestPossible > docs.length) { if (obs == null) obs = new OpenBitSet(maxDoc()); while ((docid = sub.docsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) { obs.fastSet(docid + base); bitsSet++; } } else { while ((docid = sub.docsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) { docs[upto++] = docid + base; } } } } else { int docid; if (largestPossible > docs.length) { if (obs == null) obs = new OpenBitSet(maxDoc()); while ((docid = docsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) { obs.fastSet(docid); bitsSet++; } } else { while ((docid = docsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) { docs[upto++] = docid; } } } DocSet result; if (obs != null) { for (int i = 0; i < upto; i++) { obs.fastSet(docs[i]); } bitsSet += upto; result = new BitDocSet(obs, bitsSet); } else { result = upto == 0 ? DocSet.EMPTY : new SortedIntDocSet(Arrays.copyOf(docs, upto)); } if (useCache) { filterCache.put(key, result); } return result; }
From source file:org.apache.solr.search.TermsQParserPlugin.java
License:Apache License
@Override public QParser createParser(String qstr, SolrParams localParams, SolrParams params, SolrQueryRequest req) { return new QParser(qstr, localParams, params, req) { @Override/* ww w . j a v a 2 s. com*/ public Query parse() throws SyntaxError { String fname = localParams.get(QueryParsing.F); FieldType ft = req.getSchema().getFieldTypeNoEx(fname); String separator = localParams.get(SEPARATOR, ","); String qstr = localParams.get(QueryParsing.V);//never null Method method = Method.valueOf(localParams.get(METHOD, Method.termsFilter.name())); //TODO pick the default method based on various heuristics from benchmarks //if space then split on all whitespace & trim, otherwise strictly interpret final boolean sepIsSpace = separator.equals(" "); if (sepIsSpace) qstr = qstr.trim(); if (qstr.length() == 0) return new BooleanQuery();//Matches nothing. final String[] splitVals = sepIsSpace ? qstr.split("\\s+") : qstr.split(Pattern.quote(separator), -1); assert splitVals.length > 0; BytesRef[] bytesRefs = new BytesRef[splitVals.length]; BytesRef term = new BytesRef(); for (int i = 0; i < splitVals.length; i++) { String stringVal = splitVals[i]; //logic same as TermQParserPlugin if (ft != null) { ft.readableToIndexed(stringVal, term); } else { term.copyChars(stringVal); } bytesRefs[i] = BytesRef.deepCopyOf(term); } return new SolrConstantScoreQuery(method.makeFilter(fname, bytesRefs)); } }; }
From source file:org.apache.solr.spelling.Token.java
License:Apache License
@Override public Token clone() { final Token t = (Token) super.clone(); if (payload != null) { t.payload = BytesRef.deepCopyOf(payload); }//from www . jav a 2 s . c om return t; }
From source file:org.apache.solr.spelling.Token.java
License:Apache License
@Override public void copyTo(AttributeImpl target) { super.copyTo(target); ((FlagsAttribute) target).setFlags(flags); ((PayloadAttribute) target).setPayload((payload == null) ? null : BytesRef.deepCopyOf(payload)); }
From source file:org.apache.solr.uninverting.DocTermOrds.java
License:Apache License
/** Call this only once (if you subclass!) */ protected void uninvert(final LeafReader reader, Bits liveDocs, final BytesRef termPrefix) throws IOException { final FieldInfo info = reader.getFieldInfos().fieldInfo(field); if (checkForDocValues && info != null && info.getDocValuesType() != DocValuesType.NONE) { throw new IllegalStateException( "Type mismatch: " + field + " was indexed as " + info.getDocValuesType()); }// w w w.j a v a 2 s . co m //System.out.println("DTO uninvert field=" + field + " prefix=" + termPrefix); final long startTime = System.nanoTime(); prefix = termPrefix == null ? null : BytesRef.deepCopyOf(termPrefix); final int maxDoc = reader.maxDoc(); final int[] index = new int[maxDoc]; // immediate term numbers, or the index into the byte[] representing the last number final int[] lastTerm = new int[maxDoc]; // last term we saw for this document final byte[][] bytes = new byte[maxDoc][]; // list of term numbers for the doc (delta encoded vInts) final Terms terms = reader.terms(field); if (terms == null) { // No terms return; } final TermsEnum te = terms.iterator(); final BytesRef seekStart = termPrefix != null ? termPrefix : new BytesRef(); //System.out.println("seekStart=" + seekStart.utf8ToString()); if (te.seekCeil(seekStart) == TermsEnum.SeekStatus.END) { // No terms match return; } // For our "term index wrapper" final List<BytesRef> indexedTerms = new ArrayList<>(); final PagedBytes indexedTermsBytes = new PagedBytes(15); // we need a minimum of 9 bytes, but round up to 12 since the space would // be wasted with most allocators anyway. byte[] tempArr = new byte[12]; // // enumerate all terms, and build an intermediate form of the un-inverted field. // // During this intermediate form, every document has a (potential) byte[] // and the int[maxDoc()] array either contains the termNumber list directly // or the *end* offset of the termNumber list in its byte array (for faster // appending and faster creation of the final form). // // idea... if things are too large while building, we could do a range of docs // at a time (but it would be a fair amount slower to build) // could also do ranges in parallel to take advantage of multiple CPUs // OPTIONAL: remap the largest df terms to the lowest 128 (single byte) // values. This requires going over the field first to find the most // frequent terms ahead of time. int termNum = 0; postingsEnum = null; // Loop begins with te positioned to first term (we call // seek above): for (;;) { final BytesRef t = te.term(); if (t == null || (termPrefix != null && !StringHelper.startsWith(t, termPrefix))) { break; } //System.out.println("visit term=" + t.utf8ToString() + " " + t + " termNum=" + termNum); visitTerm(te, termNum); if ((termNum & indexIntervalMask) == 0) { // Index this term sizeOfIndexedStrings += t.length; BytesRef indexedTerm = new BytesRef(); indexedTermsBytes.copy(t, indexedTerm); // TODO: really should 1) strip off useless suffix, // and 2) use FST not array/PagedBytes indexedTerms.add(indexedTerm); } final int df = te.docFreq(); if (df <= maxTermDocFreq) { postingsEnum = te.postings(postingsEnum, PostingsEnum.NONE); // dF, but takes deletions into account int actualDF = 0; for (;;) { int doc = postingsEnum.nextDoc(); if (doc == DocIdSetIterator.NO_MORE_DOCS) { break; } //System.out.println(" chunk=" + chunk + " docs"); actualDF++; termInstances++; //System.out.println(" docID=" + doc); // add TNUM_OFFSET to the term number to make room for special reserved values: // 0 (end term) and 1 (index into byte array follows) int delta = termNum - lastTerm[doc] + TNUM_OFFSET; lastTerm[doc] = termNum; int val = index[doc]; if ((val & 0xff) == 1) { // index into byte array (actually the end of // the doc-specific byte[] when building) int pos = val >>> 8; int ilen = vIntSize(delta); byte[] arr = bytes[doc]; int newend = pos + ilen; if (newend > arr.length) { // We avoid a doubling strategy to lower memory usage. // this faceting method isn't for docs with many terms. // In hotspot, objects have 2 words of overhead, then fields, rounded up to a 64-bit boundary. // TODO: figure out what array lengths we can round up to w/o actually using more memory // (how much space does a byte[] take up? Is data preceded by a 32 bit length only? // It should be safe to round up to the nearest 32 bits in any case. int newLen = (newend + 3) & 0xfffffffc; // 4 byte alignment byte[] newarr = new byte[newLen]; System.arraycopy(arr, 0, newarr, 0, pos); arr = newarr; bytes[doc] = newarr; } pos = writeInt(delta, arr, pos); index[doc] = (pos << 8) | 1; // update pointer to end index in byte[] } else { // OK, this int has data in it... find the end (a zero starting byte - not // part of another number, hence not following a byte with the high bit set). int ipos; if (val == 0) { ipos = 0; } else if ((val & 0x0000ff80) == 0) { ipos = 1; } else if ((val & 0x00ff8000) == 0) { ipos = 2; } else if ((val & 0xff800000) == 0) { ipos = 3; } else { ipos = 4; } //System.out.println(" ipos=" + ipos); int endPos = writeInt(delta, tempArr, ipos); //System.out.println(" endpos=" + endPos); if (endPos <= 4) { //System.out.println(" fits!"); // value will fit in the integer... move bytes back for (int j = ipos; j < endPos; j++) { val |= (tempArr[j] & 0xff) << (j << 3); } index[doc] = val; } else { // value won't fit... move integer into byte[] for (int j = 0; j < ipos; j++) { tempArr[j] = (byte) val; val >>>= 8; } // point at the end index in the byte[] index[doc] = (endPos << 8) | 1; bytes[doc] = tempArr; tempArr = new byte[12]; } } } setActualDocFreq(termNum, actualDF); } termNum++; if (te.next() == null) { break; } } numTermsInField = termNum; long midPoint = System.nanoTime(); if (termInstances == 0) { // we didn't invert anything // lower memory consumption. tnums = null; } else { this.index = index; // // transform intermediate form into the final form, building a single byte[] // at a time, and releasing the intermediate byte[]s as we go to avoid // increasing the memory footprint. // for (int pass = 0; pass < 256; pass++) { byte[] target = tnums[pass]; int pos = 0; // end in target; if (target != null) { pos = target.length; } else { target = new byte[4096]; } // loop over documents, 0x00ppxxxx, 0x01ppxxxx, 0x02ppxxxx // where pp is the pass (which array we are building), and xx is all values. // each pass shares the same byte[] for termNumber lists. for (int docbase = pass << 16; docbase < maxDoc; docbase += (1 << 24)) { int lim = Math.min(docbase + (1 << 16), maxDoc); for (int doc = docbase; doc < lim; doc++) { //System.out.println(" pass=" + pass + " process docID=" + doc); int val = index[doc]; if ((val & 0xff) == 1) { int len = val >>> 8; //System.out.println(" ptr pos=" + pos); index[doc] = (pos << 8) | 1; // change index to point to start of array if ((pos & 0xff000000) != 0) { // we only have 24 bits for the array index throw new IllegalStateException( "Too many values for UnInvertedField faceting on field " + field); } byte[] arr = bytes[doc]; /* for(byte b : arr) { //System.out.println(" b=" + Integer.toHexString((int) b)); } */ bytes[doc] = null; // IMPORTANT: allow GC to avoid OOM if (target.length <= pos + len) { int newlen = target.length; /*** we don't have to worry about the array getting too large * since the "pos" param will overflow first (only 24 bits available) if ((newlen<<1) <= 0) { // overflow... newlen = Integer.MAX_VALUE; if (newlen <= pos + len) { throw new SolrException(400,"Too many terms to uninvert field!"); } } else { while (newlen <= pos + len) newlen<<=1; // doubling strategy } ****/ while (newlen <= pos + len) newlen <<= 1; // doubling strategy byte[] newtarget = new byte[newlen]; System.arraycopy(target, 0, newtarget, 0, pos); target = newtarget; } System.arraycopy(arr, 0, target, pos, len); pos += len + 1; // skip single byte at end and leave it 0 for terminator } } } // shrink array if (pos < target.length) { byte[] newtarget = new byte[pos]; System.arraycopy(target, 0, newtarget, 0, pos); target = newtarget; } tnums[pass] = target; if ((pass << 16) > maxDoc) break; } } indexedTermsArray = indexedTerms.toArray(new BytesRef[indexedTerms.size()]); long endTime = System.nanoTime(); total_time = (int) TimeUnit.MILLISECONDS.convert(endTime - startTime, TimeUnit.NANOSECONDS); phase1_time = (int) TimeUnit.MILLISECONDS.convert(midPoint - startTime, TimeUnit.NANOSECONDS); }