List of usage examples for org.apache.lucene.index IndexFileNames getExtension
public static String getExtension(String filename)
From source file:perf.DiskUsage.java
License:Apache License
static Set<FieldStats> analyzeFields(SegmentReader reader) throws Exception { Map<String, FieldStats> stats = new HashMap<>(); Map<String, String> dvSuffixes = new HashMap<>(); Map<String, String> postingsSuffixes = new HashMap<>(); for (FieldInfo field : reader.getFieldInfos()) { FieldStats fieldStats = new FieldStats(field.name); stats.put(field.name, fieldStats); Map<String, String> attributes = field.attributes(); if (attributes != null) { String postingsSuffix = attributes.get(PerFieldPostingsFormat.PER_FIELD_SUFFIX_KEY); if (postingsSuffix != null) { postingsSuffixes.put(postingsSuffix, field.name); }/*from w ww.ja v a 2s. c o m*/ String dvSuffix = attributes.get(PerFieldDocValuesFormat.PER_FIELD_SUFFIX_KEY); if (dvSuffix != null) { dvSuffixes.put(dvSuffix, field.name); } } Bits docsWithField = reader.getDocsWithField(field.name); if (docsWithField != null) { int count = 0; for (int docID = 0; docID < reader.maxDoc(); docID++) { if (docsWithField.get(docID)) { count++; } } fieldStats.docCountWithField = count; } } Directory directory = reader.directory(); for (String file : directory.listAll()) { String suffix = parseSuffix(file); long bytes = directory.fileLength(file); if (suffix != null) { switch (IndexFileNames.getExtension(file)) { case "dvd": case "dvm": stats.get(dvSuffixes.get(suffix)).dvBytes += bytes; break; case "tim": case "tip": stats.get(postingsSuffixes.get(suffix)).termsBytes += bytes; break; case "doc": stats.get(postingsSuffixes.get(suffix)).postingsBytes += bytes; break; case "pos": case "pay": stats.get(postingsSuffixes.get(suffix)).proxBytes += bytes; break; default: throw new AssertionError("unexpected suffixed file: " + file); } } else { // not a per-field file, but we can hackishly do this for the points case. if ("dii".equals(IndexFileNames.getExtension(file))) { System.err.println( "retrieving per-field point usage, if you see a scary corruption error, its probably just this tool!!!!"); try (ChecksumIndexInput in = directory.openChecksumInput(file, IOContext.READONCE)) { // fail hard if its not exactly the version we do this hack for. CodecUtil.checkIndexHeader(in, "Lucene60PointsFormatMeta", 0, 0, reader.getSegmentInfo().info.getId(), ""); int fieldCount = in.readVInt(); // strangely, bkd offsets are not in any guaranteed order TreeMap<Long, String> offsetToField = new TreeMap<>(); for (int i = 0; i < fieldCount; i++) { int field = in.readVInt(); long offset = in.readVLong(); offsetToField.put(offset, reader.getFieldInfos().fieldInfo(field).name); } // now we can traverse in order long previousOffset = 0; for (Map.Entry<Long, String> entry : offsetToField.entrySet()) { long offset = entry.getKey(); String field = entry.getValue(); stats.get(field).pointsBytes += (offset - previousOffset); previousOffset = offset; } CodecUtil.checkFooter(in); } } } } return new TreeSet<FieldStats>(stats.values()); }
From source file:perf.DiskUsage.java
License:Apache License
static void report(SegmentReader reader, Set<FieldStats> stats) throws Exception { long totalSize = 0; long storeSize = 0; long vectorSize = 0; long normsSize = 0; long dvsSize = 0; long postingsSize = 0; long pointsSize = 0; long termsSize = 0; long proxSize = 0; for (String file : reader.directory().listAll()) { long size = reader.directory().fileLength(file); totalSize += size;/* w ww .ja v a 2s. c om*/ String extension = IndexFileNames.getExtension(file); if (extension != null) { switch (extension) { case "fdt": case "fdx": storeSize += size; break; case "tvx": case "tvd": vectorSize += size; break; case "nvd": case "nvm": normsSize += size; break; case "dvd": case "dvm": dvsSize += size; break; case "tim": case "tip": termsSize += size; break; case "pos": case "pay": proxSize += size; break; case "doc": postingsSize += size; break; case "dii": case "dim": pointsSize += size; break; } } } DecimalFormat df = new DecimalFormat("#,##0"); System.out.printf("total disk: %15s\n", df.format(totalSize)); System.out.printf("num docs: %15s\n", df.format(reader.numDocs())); System.out.printf("stored fields: %15s\n", df.format(storeSize)); System.out.printf("term vectors: %15s\n", df.format(vectorSize)); System.out.printf("norms: %15s\n", df.format(normsSize)); System.out.printf("docvalues: %15s\n", df.format(dvsSize)); System.out.printf("postings: %15s\n", df.format(postingsSize)); System.out.printf("prox: %15s\n", df.format(proxSize)); System.out.printf("points: %15s\n", df.format(pointsSize)); System.out.printf("terms: %15s\n", df.format(termsSize)); System.out.println(); int maxFieldNameLength = 0; for (FieldStats field : stats) { maxFieldNameLength = Math.max(maxFieldNameLength, field.name.length()); } // Make sure we format to enough room for the max field length: String fieldNameFormat = "%" + maxFieldNameLength + "s"; System.out.printf(fieldNameFormat + " %15s %15s %15s %15s %15s %15s %15s %20s\n", "field", "total", "terms dict", "postings", "proximity", "points", "docvalues", "% with dv", "features"); System.out.printf(fieldNameFormat + " %15s %15s %15s %15s %15s %15s %15s %20s\n", "=====", "=====", "==========", "========", "=========", "=========", "=========", "========", "========"); for (FieldStats field : stats) { System.out.printf(fieldNameFormat + " %15s %15s %15s %15s %15s %15s %14.1f%% %20s\n", field.name, df.format(field.totalBytes()), df.format(field.termsBytes), df.format(field.postingsBytes), df.format(field.proxBytes), df.format(field.pointsBytes), df.format(field.dvBytes), (100.0 * field.docCountWithField) / reader.maxDoc(), features(reader.getFieldInfos().fieldInfo(field.name))); } }
From source file:perf.DiskUsage.java
License:Apache License
static Set<FieldStats> analyzeFields(SegmentReader reader) throws Exception { Map<String, FieldStats> stats = new HashMap<>(); Map<String, String> dvSuffixes = new HashMap<>(); Map<String, String> postingsSuffixes = new HashMap<>(); for (FieldInfo field : reader.getFieldInfos()) { FieldStats fieldStats = new FieldStats(field.name); stats.put(field.name, fieldStats); Map<String, String> attributes = field.attributes(); if (attributes != null) { String postingsSuffix = attributes.get(PerFieldPostingsFormat.PER_FIELD_SUFFIX_KEY); if (postingsSuffix != null) { postingsSuffixes.put(postingsSuffix, field.name); }/*from w ww .j ava 2 s . c om*/ String dvSuffix = attributes.get(PerFieldDocValuesFormat.PER_FIELD_SUFFIX_KEY); if (dvSuffix != null) { dvSuffixes.put(dvSuffix, field.name); } } DocIdSetIterator docsWithField; switch (field.getDocValuesType()) { case NUMERIC: docsWithField = reader.getNumericDocValues(field.name); break; case BINARY: docsWithField = reader.getBinaryDocValues(field.name); break; case SORTED: docsWithField = reader.getSortedDocValues(field.name); break; case SORTED_NUMERIC: docsWithField = reader.getSortedNumericDocValues(field.name); break; case SORTED_SET: docsWithField = reader.getSortedSetDocValues(field.name); break; case NONE: docsWithField = null; break; default: docsWithField = null; break; } if (docsWithField != null) { int count = 0; while (docsWithField.nextDoc() != DocIdSetIterator.NO_MORE_DOCS) { count++; } fieldStats.docCountWithField = count; } } Directory directory = reader.directory(); for (String file : directory.listAll()) { String suffix = parseSuffix(file); long bytes = directory.fileLength(file); if (suffix != null) { switch (IndexFileNames.getExtension(file)) { case "dvd": case "dvm": stats.get(dvSuffixes.get(suffix)).dvBytes += bytes; break; case "tim": case "tip": stats.get(postingsSuffixes.get(suffix)).termsBytes += bytes; break; case "doc": stats.get(postingsSuffixes.get(suffix)).postingsBytes += bytes; break; case "pos": case "pay": stats.get(postingsSuffixes.get(suffix)).proxBytes += bytes; break; default: throw new AssertionError("unexpected suffixed file: " + file); } } else { // not a per-field file, but we can hackishly do this for the points case. if ("dii".equals(IndexFileNames.getExtension(file))) { System.err.println( "retrieving per-field point usage, if you see a scary corruption error, its probably just this tool!!!!"); try (ChecksumIndexInput in = directory.openChecksumInput(file, IOContext.READONCE)) { // fail hard if its not exactly the version we do this hack for. CodecUtil.checkIndexHeader(in, "Lucene60PointsFormatMeta", 0, 0, reader.getSegmentInfo().info.getId(), ""); int fieldCount = in.readVInt(); // strangely, bkd offsets are not in any guaranteed order TreeMap<Long, String> offsetToField = new TreeMap<>(); for (int i = 0; i < fieldCount; i++) { int field = in.readVInt(); long offset = in.readVLong(); offsetToField.put(offset, reader.getFieldInfos().fieldInfo(field).name); } // now we can traverse in order long previousOffset = 0; for (Map.Entry<Long, String> entry : offsetToField.entrySet()) { long offset = entry.getKey(); String field = entry.getValue(); stats.get(field).pointsBytes += (offset - previousOffset); previousOffset = offset; } CodecUtil.checkFooter(in); } } } } return new TreeSet<FieldStats>(stats.values()); }