public void testCallback()
{
Configuration conf = new Configuration();
MemoryManager mgr = new MemoryManager(configuredPoolSize);
long pool = mgr.getTotalMemoryPool();
LoggingCallback[] calls = new LoggingCallback[20];
for (int i = 0; i < calls.Length; ++i)
{
calls[i] = new LoggingCallback();
mgr.addWriter(i.ToString(), pool / 4, calls[i]);
}
// add enough rows to get the memory manager to check the limits
for (int i = 0; i < 10000; ++i)
{
mgr.addedRow(1);
}
for (int call = 0; call < calls.Length; ++call)
{
Assert.Equal(2, calls[call].LogLength);
foreach (double argument in calls[call].Log)
{
Assert.Equal(0.2, argument, ERROR);
}
}
}
public OrcTestBase()
{
conf = new Configuration();
workDir = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString());
Directory.CreateDirectory(workDir);
testClassName = GetType().Name;
}
/**
* Create a set of write options based on a set of table properties and
* configuration.
* @param tableProperties the properties of the table
* @param conf the configuration of the query
* @return a WriterOptions object that can be modified
*/
public static WriterOptions writerOptions(Properties tableProperties, Configuration conf)
{
return new WriterOptions(tableProperties, conf);
}
private static MemoryManager getMemoryManager(Configuration conf)
{
return memoryManager.Value;
}
public static ReaderOptions readerOptions(Configuration conf)
{
return new ReaderOptions(conf);
}
/**
* Create a set of writer options based on a configuration.
* @param conf the configuration to use for values
* @return A WriterOptions object that can be modified
*/
public static WriterOptions writerOptions(Configuration conf)
{
return new WriterOptions(null, conf);
}
public double getDouble(Properties tbl, Configuration conf)
{
string value = lookupValue(tbl, conf);
if (value != null)
{
return Double.Parse(value);
}
return ((IConvertible)defaultValue).ToDouble(CultureInfo.InvariantCulture);
}
public static string getString(this OrcConf orcConf, Configuration conf)
{
return OrcConfDetails.details[(int)orcConf].getString(null, conf);
}
public static double getDouble(this OrcConf orcConf, Configuration conf)
{
return OrcConfDetails.details[(int)orcConf].getDouble(null, conf);
}
public void testConfig()
{
Configuration conf = new Configuration();
conf.set("hive.exec.orc.memory.pool", "0.9");
MemoryManager mgr = new MemoryManager(conf);
long mem = ManagementFactory.getMemoryMXBean().getHeapMemoryUsage().getMax();
System.Console.WriteLine("Memory = " + mem);
long pool = mgr.getTotalMemoryPool();
Assert.True("Pool too small: " + pool, mem * 0.899 < pool);
Assert.True("Pool too big: " + pool, pool < mem * 0.901);
}
public static bool getBoolean(this OrcConf orcConf, Configuration conf)
{
return OrcConfDetails.details[(int)orcConf].getBoolean(null, conf);
}
public static double getDouble(this OrcConf orcConf, Properties tbl, Configuration conf)
{
return OrcConfDetails.details[(int)orcConf].getDouble(tbl, conf);
}
private string lookupValue(Properties tbl, Configuration conf)
{
string result = null;
if (tbl != null)
{
result = tbl.getProperty(attribute);
}
if (result == null && conf != null)
{
result = conf.get(attribute);
if (result == null)
{
result = conf.get(hiveConfName);
}
}
return result;
}
public string getString(Properties tbl, Configuration conf)
{
string value = lookupValue(tbl, conf);
return value == null ? (String)defaultValue : value;
}
public long getLong(Properties tbl, Configuration conf)
{
string value = lookupValue(tbl, conf);
if (value != null)
{
return Int64.Parse(value);
}
return ((IConvertible)defaultValue).ToInt64(CultureInfo.InvariantCulture);
}
public ReaderOptions(Configuration conf)
{
this.conf = conf;
}
public void testZeroCopySeek()
{
ObjectInspector inspector = ObjectInspectorFactory.getReflectionObjectInspector(typeof(BigRow));
const int COUNT = 32768;
long[] intValues = new long[COUNT];
double[] doubleValues = new double[COUNT];
string[] stringValues = new string[COUNT];
byte[][] byteValues = new byte[COUNT][];
string[] words = new string[128];
using (Stream file = FileOpenWrite(TestFilePath))
using (Writer writer = OrcFile.createWriter(TestFilePath, file, OrcFile.writerOptions(conf)
.inspector(inspector)
.stripeSize(200000)
.bufferSize(65536)
.rowIndexStride(1000)))
{
Random rand = new Random(42);
for (int i = 0; i < words.Length; ++i)
{
words[i] = Integer.toHexString(rand.Next());
}
for (int i = 0; i < COUNT / 2; ++i)
{
intValues[2 * i] = rand.NextLong();
intValues[2 * i + 1] = intValues[2 * i];
stringValues[2 * i] = words[rand.Next(words.Length)];
stringValues[2 * i + 1] = stringValues[2 * i];
}
for (int i = 0; i < COUNT; ++i)
{
doubleValues[i] = rand.NextDouble();
byte[] buf = new byte[20];
rand.NextBytes(buf);
byteValues[i] = buf;
}
for (int i = 0; i < COUNT; ++i)
{
writer.addRow(createRandomRow(intValues, doubleValues, stringValues,
byteValues, words, i));
}
}
Reader reader = OrcFile.createReader(TestFilePath, OrcFile.readerOptions(conf));
Assert.Equal(COUNT, reader.getNumberOfRows());
/* enable zero copy record reader */
#if false
Configuration conf = new Configuration();
HiveConf.setBoolVar(conf, HiveConf.ConfVars.HIVE_ORC_ZEROCOPY, true);
#endif
using (RecordReader rows = reader.rows())
{
/* all tests are identical to the other seek() tests */
for (int i = COUNT - 1; i >= 0; --i)
{
rows.seekToRow(i);
OrcStruct row = (OrcStruct)rows.next();
BigRow expected = createRandomRow(intValues, doubleValues,
stringValues, byteValues, words, i);
Assert.Equal(expected.boolean1, row.getFieldValue(0));
Assert.Equal(expected.byte1, row.getFieldValue(1));
Assert.Equal(expected.short1, row.getFieldValue(2));
Assert.Equal(expected.int1, row.getFieldValue(3));
Assert.Equal(expected.long1, row.getFieldValue(4));
Assert.Equal(expected.float1, (float)row.getFieldValue(5), 4);
Assert.Equal(expected.double1, (double)row.getFieldValue(6), 4);
Assert.Equal(expected.bytes1, row.getFieldValue(7));
Assert.Equal(expected.string1, row.getFieldValue(8));
List<InnerStruct> expectedList = expected.middle.list;
List<object> actualList =
(List<object>)((OrcStruct)row.getFieldValue(9)).getFieldValue(0);
compareList(expectedList, actualList);
compareList(expected.list, (List<object>)row.getFieldValue(10));
}
}
IList<StripeInformation> stripes = reader.getStripes();
long offsetOfStripe2 = 0;
long offsetOfStripe4 = 0;
long lastRowOfStripe2 = 0;
for (int i = 0; i < 5; ++i)
{
StripeInformation stripe = stripes[i];
if (i < 2)
{
lastRowOfStripe2 += stripe.getNumberOfRows();
}
else if (i == 2)
{
offsetOfStripe2 = stripe.getOffset();
lastRowOfStripe2 += stripe.getNumberOfRows() - 1;
}
else if (i == 4)
{
offsetOfStripe4 = stripe.getOffset();
}
}
bool[] columns = new bool[reader.getStatistics().Length];
columns[5] = true; // long colulmn
columns[9] = true; // text column
/* use zero copy record reader */
using (RecordReader rows = reader.rowsOptions(new RecordReaderOptions()
.range(offsetOfStripe2, offsetOfStripe4 - offsetOfStripe2)
.include(columns)))
{
rows.seekToRow(lastRowOfStripe2);
for (int i = 0; i < 2; ++i)
{
OrcStruct row = (OrcStruct)rows.next();
BigRow expected = createRandomRow(intValues, doubleValues,
stringValues, byteValues, words,
(int)(lastRowOfStripe2 + i));
Assert.Equal(expected.long1, row.getFieldValue(4));
Assert.Equal(expected.string1, row.getFieldValue(8));
}
}
}
public WriterOptions(Properties tableProperties, Configuration conf)
{
configuration = conf;
memoryManagerValue = getMemoryManager(conf);
stripeSizeValue = OrcConf.STRIPE_SIZE.getLong(tableProperties, conf);
blockSizeValue = OrcConf.BLOCK_SIZE.getLong(tableProperties, conf);
rowIndexStrideValue =
(int)OrcConf.ROW_INDEX_STRIDE.getLong(tableProperties, conf);
bufferSizeValue = (int)OrcConf.BUFFER_SIZE.getLong(tableProperties,
conf);
blockPaddingValue =
OrcConf.BLOCK_PADDING.getBoolean(tableProperties, conf);
compressValue = (CompressionKind)Enum.Parse(
typeof(CompressionKind),
OrcConf.COMPRESS.getString(tableProperties, conf),
true);
string versionName = OrcConf.WRITE_FORMAT.getString(tableProperties,
conf);
versionValue = VersionHelper.byName(versionName);
string enString = OrcConf.ENCODING_STRATEGY.getString(tableProperties,
conf);
_encodingStrategy = (EncodingStrategy)Enum.Parse(typeof(EncodingStrategy), enString, true);
string compString =
OrcConf.COMPRESSION_STRATEGY.getString(tableProperties, conf);
compressionStrategy = (CompressionStrategy)Enum.Parse(typeof(CompressionStrategy), compString, true);
_paddingTolerance =
OrcConf.BLOCK_PADDING_TOLERANCE.getDouble(tableProperties, conf);
_bloomFilterColumns = OrcConf.BLOOM_FILTER_COLUMNS.getString(tableProperties,
conf);
_bloomFilterFpp = OrcConf.BLOOM_FILTER_FPP.getDouble(tableProperties,
conf);
timeZone = TimeZoneInfo.Local.Id;
}
public void testMaxLengthToReader()
{
Configuration conf = new Configuration();
OrcProto.Type rowType = OrcProto.Type.CreateBuilder()
.SetKind(OrcProto.Type.Types.Kind.STRUCT).Build();
OrcProto.Footer footer = OrcProto.Footer.CreateBuilder()
.SetHeaderLength(0).SetContentLength(0).SetNumberOfRows(0)
.SetRowIndexStride(0).AddTypes(rowType).Build();
OrcProto.PostScript ps = OrcProto.PostScript.CreateBuilder()
.SetCompression(OrcProto.CompressionKind.NONE)
.SetFooterLength((ulong)footer.SerializedSize)
.SetMagic("ORC").AddVersion(0).AddVersion(11).Build();
DataOutputBuffer buffer = new DataOutputBuffer();
footer.WriteTo(buffer);
ps.WriteTo(buffer);
buffer.write(ps.SerializedSize);
FileSystem fs = Mockito.mock(typeof(FileSystem), settings);
FSDataInputStream file = new FSDataInputStream(new BufferInStream(buffer.getData(),
buffer.getLength()));
string p = "/dir/file.orc";
Mockito.when(fs.open(p)).thenReturn(file);
OrcFile.ReaderOptions options = OrcFile.readerOptions(conf);
options.filesystem(fs);
options.maxLength(buffer.getLength());
Mockito.when(fs.getFileStatus(p))
.thenReturn(new FileStatus(10, false, 3, 3000, 0, p));
Reader reader = OrcFile.createReader(p, options);
}
public static long getLong(this OrcConf orcConf, Properties tbl, Configuration conf)
{
return OrcConfDetails.details[(int)orcConf].getLong(tbl, conf);
}
public bool getBoolean(Properties tbl, Configuration conf)
{
string value = lookupValue(tbl, conf);
if (value != null)
{
return Boolean.Parse(value);
}
return (bool)defaultValue;
}
public MyMemoryManager(Configuration conf, long totalSpace, double rate)
: base(totalSpace)
{
this.totalSpace = totalSpace;
this.rate = rate;
}
public OrcOptions(Configuration conf)
: base(conf)
{
}
/**
* Create a reader that merge sorts the ACID events together.
* @param conf the configuration
* @param collapseEvents should the events on the same row be collapsed
* @param isOriginal is the base file a pre-acid file
* @param bucket the bucket we are reading
* @param options the options to read with
* @param deltaDirectory the list of delta directories to include
* @
*/
OrcRawRecordMerger(Configuration conf,
bool collapseEvents,
Reader reader,
bool isOriginal,
int bucket,
ValidTxnList validTxnList,
Reader.Options options,
Path[] deltaDirectory)
{
this.conf = conf;
this.collapse = collapseEvents;
this.offset = options.getOffset();
this.length = options.getLength();
this.validTxnList = validTxnList;
TypeDescription typeDescr = OrcUtils.getDesiredRowTypeDescr(conf);
if (typeDescr == null) {
throw new IOException(ErrorMsg.SCHEMA_REQUIRED_TO_READ_ACID_TABLES.getErrorCodedMsg());
}
objectInspector = OrcRecordUpdater.createEventSchema
(OrcStruct.createObjectInspector(0, OrcUtils.getOrcTypes(typeDescr)));
// modify the options to reflect the event instead of the base row
Reader.Options eventOptions = createEventOptions(options);
if (reader == null)
{
baseReader = null;
}
else
{
// find the min/max based on the offset and length
if (isOriginal)
{
discoverOriginalKeyBounds(reader, bucket, options);
}
else
{
discoverKeyBounds(reader, options);
}
LOG.info("min key = " + minKey + ", max key = " + maxKey);
// use the min/max instead of the byte range
ReaderPair pair;
ReaderKey key = new ReaderKey();
if (isOriginal)
{
options = options.clone();
options.range(options.getOffset(), Long.MAX_VALUE);
pair = new OriginalReaderPair(key, reader, bucket, minKey, maxKey,
options);
}
else
{
pair = new ReaderPair(key, reader, bucket, minKey, maxKey,
eventOptions, 0);
}
// if there is at least one record, put it in the map
if (pair.nextRecord != null)
{
readers.put(key, pair);
}
baseReader = pair.recordReader;
}
// we always want to read all of the deltas
eventOptions.range(0, Long.MAX_VALUE);
if (deltaDirectory != null)
{
foreach (Path delta in deltaDirectory)
{
ReaderKey key = new ReaderKey();
Path deltaFile = AcidUtils.createBucketFile(delta, bucket);
AcidUtils.ParsedDelta deltaDir = AcidUtils.parsedDelta(delta);
FileSystem fs = deltaFile.getFileSystem(conf);
long length = getLastFlushLength(fs, deltaFile);
if (length != -1 && fs.exists(deltaFile))
{
Reader deltaReader = OrcFile.createReader(deltaFile,
OrcFile.readerOptions(conf).maxLength(length));
Reader.Options deltaEventOptions = null;
if (eventOptions.getSearchArgument() != null)
{
// Turn off the sarg before pushing it to delta. We never want to push a sarg to a delta as
// it can produce wrong results (if the latest valid version of the record is filtered out by
// the sarg) or ArrayOutOfBounds errors (when the sarg is applied to a delete record)
// unless the delta only has insert events
OrcRecordUpdater.AcidStats acidStats = OrcRecordUpdater.parseAcidStats(deltaReader);
if (acidStats.deletes > 0 || acidStats.updates > 0)
{
deltaEventOptions = eventOptions.clone().searchArgument(null, null);
}
}
ReaderPair deltaPair;
deltaPair = new ReaderPair(key, deltaReader, bucket, minKey,
maxKey, deltaEventOptions != null ? deltaEventOptions : eventOptions, deltaDir.getStatementId());
if (deltaPair.nextRecord != null)
{
readers.put(key, deltaPair);
}
}
}
}
// get the first record
Map.Entry<ReaderKey, ReaderPair> entry = readers.pollFirstEntry();
if (entry == null)
{
columns = 0;
primary = null;
}
else
{
primary = entry.getValue();
if (readers.isEmpty())
{
secondaryKey = null;
}
else
{
secondaryKey = readers.firstKey();
}
// get the number of columns in the user's rows
columns = primary.getColumns();
}
}
public static void printJsonMetaData(List<string> files, Configuration conf,
List<int> rowIndexCols, bool prettyPrint, bool printTimeZone)
{
JsonWriter writer = new JsonWriter();
bool multiFile = files.Count > 1;
if (multiFile)
{
writer.array();
}
else
{
writer.newObject();
}
foreach (string filename in files)
{
if (multiFile)
{
writer.newObject();
}
writer.key("fileName").value(Path.GetFileName(filename));
Reader reader = OrcFile.createReader(filename, OrcFile.readerOptions(conf));
writer.key("fileVersion").value(OrcFile.VersionHelper.getName(reader.getFileVersion()));
writer.key("writerVersion").value(reader.getWriterVersion().ToString());
using (RecordReaderImpl rows = (RecordReaderImpl)reader.rows())
{
writer.key("numberOfRows").value(reader.getNumberOfRows());
writer.key("compression").value(reader.getCompression().ToString());
if (reader.getCompression() != CompressionKind.NONE)
{
writer.key("compressionBufferSize").value(reader.getCompressionSize());
}
writer.key("schemaString").value(reader.getObjectInspector().getTypeName());
writer.key("schema").array();
writeSchema(writer, reader.getTypes());
writer.endArray();
writer.key("stripeStatistics").array();
List<StripeStatistics> stripeStatistics = reader.getStripeStatistics();
for (int n = 0; n < stripeStatistics.Count; n++)
{
writer.newObject();
writer.key("stripeNumber").value(n + 1);
StripeStatistics ss = stripeStatistics[n];
writer.key("columnStatistics").array();
for (int i = 0; i < ss.getColumnStatistics().Length; i++)
{
writer.newObject();
writer.key("columnId").value(i);
writeColumnStatistics(writer, ss.getColumnStatistics()[i]);
writer.endObject();
}
writer.endArray();
writer.endObject();
}
writer.endArray();
ColumnStatistics[] stats = reader.getStatistics();
int colCount = stats.Length;
writer.key("fileStatistics").array();
for (int i = 0; i < stats.Length; ++i)
{
writer.newObject();
writer.key("columnId").value(i);
writeColumnStatistics(writer, stats[i]);
writer.endObject();
}
writer.endArray();
writer.key("stripes").array();
int stripeIx = -1;
foreach (StripeInformation stripe in reader.getStripes())
{
++stripeIx;
long stripeStart = stripe.getOffset();
OrcProto.StripeFooter footer = rows.readStripeFooter(stripe);
writer.newObject(); // start of stripe information
writer.key("stripeNumber").value(stripeIx + 1);
writer.key("stripeInformation");
writeStripeInformation(writer, stripe);
if (printTimeZone)
{
writer.key("writerTimezone").value(
footer.HasWriterTimezone ? footer.WriterTimezone : FileDump.UNKNOWN);
}
long sectionStart = stripeStart;
writer.key("streams").array();
foreach (OrcProto.Stream section in footer.StreamsList)
{
writer.newObject();
string kind = section.HasKind ? section.Kind.ToString() : FileDump.UNKNOWN;
writer.key("columnId").value(section.Column);
writer.key("section").value(kind);
writer.key("startOffset").value(sectionStart);
writer.key("length").value(section.Length);
sectionStart += (long)section.Length;
writer.endObject();
}
writer.endArray();
writer.key("encodings").array();
for (int i = 0; i < footer.ColumnsCount; ++i)
{
writer.newObject();
OrcProto.ColumnEncoding encoding = footer.ColumnsList[i];
writer.key("columnId").value(i);
writer.key("kind").value(encoding.Kind.ToString());
if (encoding.Kind == OrcProto.ColumnEncoding.Types.Kind.DICTIONARY ||
encoding.Kind == OrcProto.ColumnEncoding.Types.Kind.DICTIONARY_V2)
{
writer.key("dictionarySize").value(encoding.DictionarySize);
}
writer.endObject();
}
writer.endArray();
if (rowIndexCols != null && rowIndexCols.Count != 0)
{
// include the columns that are specified, only if the columns are included, bloom filter
// will be read
bool[] sargColumns = new bool[colCount];
foreach (int colIdx in rowIndexCols)
{
sargColumns[colIdx] = true;
}
RecordReaderImpl.Index indices = rows.readRowIndex(stripeIx, null, sargColumns);
writer.key("indexes").array();
foreach (int col in rowIndexCols)
{
writer.newObject();
writer.key("columnId").value(col);
writeRowGroupIndexes(writer, col, indices.getRowGroupIndex());
writeBloomFilterIndexes(writer, col, indices.getBloomFilterIndex());
writer.endObject();
}
writer.endArray();
}
writer.endObject(); // end of stripe information
}
writer.endArray();
long fileLen = new FileInfo(filename).Length;
long paddedBytes = FileDump.getTotalPaddingSize(reader);
// empty ORC file is ~45 bytes. Assumption here is file length always >0
double percentPadding = ((double)paddedBytes / (double)fileLen) * 100;
writer.key("fileLength").value(fileLen);
writer.key("paddingLength").value(paddedBytes);
writer.key("paddingRatio").value(percentPadding);
rows.close();
}
writer.endObject();
}
if (multiFile)
{
writer.endArray();
}
if (prettyPrint)
{
#if false
string prettyJson;
if (multiFile)
{
JSONArray jsonArray = new JSONArray(writer.toString());
prettyJson = jsonArray.toString(2);
}
else
{
JSONObject jsonObject = new JSONObject(writer.toString());
prettyJson = jsonObject.toString(2);
}
#else
string prettyJson = writer.ToString();
#endif
System.Console.WriteLine(prettyJson);
}
else
{
System.Console.WriteLine(writer.ToString());
}
}
