/// <exception cref="System.IO.IOException"/>
public override void Close()
{
lock (this)
{
lineRecordReader.Close();
}
}
/// <exception cref="System.IO.IOException"/>
private void TestSplitRecordsForFile(Configuration conf, long firstSplitLength, long
testFileSize, Path testFilePath)
{
conf.SetInt(LineRecordReader.MaxLineLength, int.MaxValue);
NUnit.Framework.Assert.IsTrue("unexpected test data at " + testFilePath, testFileSize
> firstSplitLength);
string delimiter = conf.Get("textinputformat.record.delimiter");
byte[] recordDelimiterBytes = null;
if (null != delimiter)
{
recordDelimiterBytes = Sharpen.Runtime.GetBytesForString(delimiter, Charsets.Utf8
);
}
TaskAttemptContext context = new TaskAttemptContextImpl(conf, new TaskAttemptID()
);
// read the data without splitting to count the records
FileSplit split = new FileSplit(testFilePath, 0, testFileSize, (string[])null);
LineRecordReader reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
int numRecordsNoSplits = 0;
while (reader.NextKeyValue())
{
++numRecordsNoSplits;
}
reader.Close();
// count the records in the first split
split = new FileSplit(testFilePath, 0, firstSplitLength, (string[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
int numRecordsFirstSplit = 0;
while (reader.NextKeyValue())
{
++numRecordsFirstSplit;
}
reader.Close();
// count the records in the second split
split = new FileSplit(testFilePath, firstSplitLength, testFileSize - firstSplitLength
, (string[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
int numRecordsRemainingSplits = 0;
while (reader.NextKeyValue())
{
++numRecordsRemainingSplits;
}
reader.Close();
NUnit.Framework.Assert.AreEqual("Unexpected number of records in split ", numRecordsNoSplits
, numRecordsFirstSplit + numRecordsRemainingSplits);
}
public virtual void TestMultipleClose()
{
Uri testFileUrl = GetType().GetClassLoader().GetResource("recordSpanningMultipleSplits.txt.bz2"
);
NUnit.Framework.Assert.IsNotNull("Cannot find recordSpanningMultipleSplits.txt.bz2"
, testFileUrl);
FilePath testFile = new FilePath(testFileUrl.GetFile());
Path testFilePath = new Path(testFile.GetAbsolutePath());
long testFileSize = testFile.Length();
Configuration conf = new Configuration();
conf.SetInt(LineRecordReader.MaxLineLength, int.MaxValue);
TaskAttemptContext context = new TaskAttemptContextImpl(conf, new TaskAttemptID()
);
// read the data and check whether BOM is skipped
FileSplit split = new FileSplit(testFilePath, 0, testFileSize, null);
LineRecordReader reader = new LineRecordReader();
reader.Initialize(split, context);
//noinspection StatementWithEmptyBody
while (reader.NextKeyValue())
{
}
reader.Close();
reader.Close();
BZip2Codec codec = new BZip2Codec();
codec.SetConf(conf);
ICollection <Decompressor> decompressors = new HashSet <Decompressor>();
for (int i = 0; i < 10; ++i)
{
decompressors.AddItem(CodecPool.GetDecompressor(codec));
}
NUnit.Framework.Assert.AreEqual(10, decompressors.Count);
}
public virtual void TestStripBOM()
{
// the test data contains a BOM at the start of the file
// confirm the BOM is skipped by LineRecordReader
string Utf8Bom = "\uFEFF";
Uri testFileUrl = GetType().GetClassLoader().GetResource("testBOM.txt");
NUnit.Framework.Assert.IsNotNull("Cannot find testBOM.txt", testFileUrl);
FilePath testFile = new FilePath(testFileUrl.GetFile());
Path testFilePath = new Path(testFile.GetAbsolutePath());
long testFileSize = testFile.Length();
Configuration conf = new Configuration();
conf.SetInt(LineRecordReader.MaxLineLength, int.MaxValue);
TaskAttemptContext context = new TaskAttemptContextImpl(conf, new TaskAttemptID()
);
// read the data and check whether BOM is skipped
FileSplit split = new FileSplit(testFilePath, 0, testFileSize, (string[])null);
LineRecordReader reader = new LineRecordReader();
reader.Initialize(split, context);
int numRecords = 0;
bool firstLine = true;
bool skipBOM = true;
while (reader.NextKeyValue())
{
if (firstLine)
{
firstLine = false;
if (reader.GetCurrentValue().ToString().StartsWith(Utf8Bom))
{
skipBOM = false;
}
}
++numRecords;
}
reader.Close();
NUnit.Framework.Assert.IsTrue("BOM is not skipped", skipBOM);
}
public virtual void TestUncompressedInputCustomDelimiterPosValue()
{
Configuration conf = new Configuration();
conf.SetInt("io.file.buffer.size", 10);
conf.SetInt(LineRecordReader.MaxLineLength, int.MaxValue);
string inputData = "abcdefghij++kl++mno";
Path inputFile = CreateInputFile(conf, inputData);
string delimiter = "++";
byte[] recordDelimiterBytes = Sharpen.Runtime.GetBytesForString(delimiter, Charsets
.Utf8);
int splitLength = 15;
FileSplit split = new FileSplit(inputFile, 0, splitLength, (string[])null);
TaskAttemptContext context = new TaskAttemptContextImpl(conf, new TaskAttemptID()
);
LineRecordReader reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
// Get first record: "abcdefghij"
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
LongWritable key = reader.GetCurrentKey();
Text value = reader.GetCurrentValue();
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 10, value.GetLength
());
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 0, key.Get());
// Get second record: "kl"
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 2, value.GetLength
());
// Key should be 12 right after "abcdefghij++"
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 12, key.Get()
);
// Get third record: "mno"
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 3, value.GetLength
());
// Key should be 16 right after "abcdefghij++kl++"
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 16, key.Get()
);
NUnit.Framework.Assert.IsFalse(reader.NextKeyValue());
// Key should be 19 right after "abcdefghij++kl++mno"
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 19, key.Get()
);
// after refresh should be empty
key = reader.GetCurrentKey();
NUnit.Framework.Assert.IsNull("Unexpected key returned", key);
reader.Close();
split = new FileSplit(inputFile, splitLength, inputData.Length - splitLength, (string
[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
// No record is in the second split because the second split dropped
// the first record, which was already reported by the first split.
NUnit.Framework.Assert.IsFalse("Unexpected record returned", reader.NextKeyValue(
));
key = reader.GetCurrentKey();
NUnit.Framework.Assert.IsNull("Unexpected key returned", key);
reader.Close();
// multi char delimiter with starting part of the delimiter in the data
inputData = "abcd+efgh++ijk++mno";
inputFile = CreateInputFile(conf, inputData);
splitLength = 5;
split = new FileSplit(inputFile, 0, splitLength, (string[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
// Get first record: "abcd+efgh"
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
key = reader.GetCurrentKey();
value = reader.GetCurrentValue();
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 0, key.Get());
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 9, value.GetLength
());
// should have jumped over the delimiter, no record
NUnit.Framework.Assert.IsFalse(reader.NextKeyValue());
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 11, key.Get()
);
// after refresh should be empty
key = reader.GetCurrentKey();
NUnit.Framework.Assert.IsNull("Unexpected key returned", key);
reader.Close();
// next split: check for duplicate or dropped records
split = new FileSplit(inputFile, splitLength, inputData.Length - splitLength, (string
[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
key = reader.GetCurrentKey();
value = reader.GetCurrentValue();
// Get second record: "ijk" first in this split
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 11, key.Get()
);
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 3, value.GetLength
());
// Get third record: "mno" second in this split
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 16, key.Get()
);
NUnit.Framework.Assert.AreEqual("Wrong length for record value", 3, value.GetLength
());
// should be at the end of the input
NUnit.Framework.Assert.IsFalse(reader.NextKeyValue());
NUnit.Framework.Assert.AreEqual("Wrong position after record read", 19, key.Get()
);
reader.Close();
inputData = "abcd|efgh|+|ij|kl|+|mno|pqr";
inputFile = CreateInputFile(conf, inputData);
delimiter = "|+|";
recordDelimiterBytes = Sharpen.Runtime.GetBytesForString(delimiter, Charsets.Utf8
);
// walking over the buffer and split sizes checks for proper processing
// of the ambiguous bytes of the delimiter
for (int bufferSize = 1; bufferSize <= inputData.Length; bufferSize++)
{
for (int splitSize = 1; splitSize < inputData.Length; splitSize++)
{
// track where we are in the inputdata
int keyPosition = 0;
conf.SetInt("io.file.buffer.size", bufferSize);
split = new FileSplit(inputFile, 0, bufferSize, (string[])null);
reader = new LineRecordReader(recordDelimiterBytes);
reader.Initialize(split, context);
// Get the first record: "abcd|efgh" always possible
NUnit.Framework.Assert.IsTrue("Expected record got nothing", reader.NextKeyValue(
));
key = reader.GetCurrentKey();
value = reader.GetCurrentValue();
NUnit.Framework.Assert.IsTrue("abcd|efgh".Equals(value.ToString()));
// Position should be 0 right at the start
NUnit.Framework.Assert.AreEqual("Wrong position after record read", keyPosition,
key.Get());
// Position should be 12 right after the first "|+|"
keyPosition = 12;
// get the next record: "ij|kl" if the split/buffer allows it
if (reader.NextKeyValue())
{
// check the record info: "ij|kl"
NUnit.Framework.Assert.IsTrue("ij|kl".Equals(value.ToString()));
NUnit.Framework.Assert.AreEqual("Wrong position after record read", keyPosition,
key.Get());
// Position should be 20 after the second "|+|"
keyPosition = 20;
}
// get the third record: "mno|pqr" if the split/buffer allows it
if (reader.NextKeyValue())
{
// check the record info: "mno|pqr"
NUnit.Framework.Assert.IsTrue("mno|pqr".Equals(value.ToString()));
NUnit.Framework.Assert.AreEqual("Wrong position after record read", keyPosition,
key.Get());
// Position should be the end of the input
keyPosition = inputData.Length;
}
NUnit.Framework.Assert.IsFalse("Unexpected record returned", reader.NextKeyValue(
));
// no more records can be read we should be at the last position
NUnit.Framework.Assert.AreEqual("Wrong position after record read", keyPosition,
key.Get());
// after refresh should be empty
key = reader.GetCurrentKey();
NUnit.Framework.Assert.IsNull("Unexpected key returned", key);
reader.Close();
}
}
}