/// <exception cref="System.IO.IOException"/>
public virtual void TestAddInputPathWithFormat()
{
Job conf = Job.GetInstance();
MultipleInputs.AddInputPath(conf, new Path("/foo"), typeof(TextInputFormat));
MultipleInputs.AddInputPath(conf, new Path("/bar"), typeof(KeyValueTextInputFormat
));
IDictionary <Path, InputFormat> inputs = MultipleInputs.GetInputFormatMap(conf);
NUnit.Framework.Assert.AreEqual(typeof(TextInputFormat), inputs[new Path("/foo")]
.GetType());
NUnit.Framework.Assert.AreEqual(typeof(KeyValueTextInputFormat), inputs[new Path(
"/bar")].GetType());
}
/// <exception cref="System.IO.IOException"/>
/// <exception cref="System.Exception"/>
public override IList <InputSplit> GetSplits(JobContext job)
{
Configuration conf = job.GetConfiguration();
Job jobCopy = Job.GetInstance(conf);
IList <InputSplit> splits = new AList <InputSplit>();
IDictionary <Path, InputFormat> formatMap = MultipleInputs.GetInputFormatMap(job);
IDictionary <Path, Type> mapperMap = MultipleInputs.GetMapperTypeMap(job);
IDictionary <Type, IList <Path> > formatPaths = new Dictionary <Type, IList <Path> >();
// First, build a map of InputFormats to Paths
foreach (KeyValuePair <Path, InputFormat> entry in formatMap)
{
if (!formatPaths.Contains(entry.Value.GetType()))
{
formatPaths[entry.Value.GetType()] = new List <Path>();
}
formatPaths[entry.Value.GetType()].AddItem(entry.Key);
}
foreach (KeyValuePair <Type, IList <Path> > formatEntry in formatPaths)
{
Type formatClass = formatEntry.Key;
InputFormat format = (InputFormat)ReflectionUtils.NewInstance(formatClass, conf);
IList <Path> paths = formatEntry.Value;
IDictionary <Type, IList <Path> > mapperPaths = new Dictionary <Type, IList <Path> >();
// Now, for each set of paths that have a common InputFormat, build
// a map of Mappers to the paths they're used for
foreach (Path path in paths)
{
Type mapperClass = mapperMap[path];
if (!mapperPaths.Contains(mapperClass))
{
mapperPaths[mapperClass] = new List <Path>();
}
mapperPaths[mapperClass].AddItem(path);
}
// Now each set of paths that has a common InputFormat and Mapper can
// be added to the same job, and split together.
foreach (KeyValuePair <Type, IList <Path> > mapEntry in mapperPaths)
{
paths = mapEntry.Value;
Type mapperClass = mapEntry.Key;
if (mapperClass == null)
{
try
{
mapperClass = job.GetMapperClass();
}
catch (TypeLoadException e)
{
throw new IOException("Mapper class is not found", e);
}
}
FileInputFormat.SetInputPaths(jobCopy, Sharpen.Collections.ToArray(paths, new Path
[paths.Count]));
// Get splits for each input path and tag with InputFormat
// and Mapper types by wrapping in a TaggedInputSplit.
IList <InputSplit> pathSplits = format.GetSplits(jobCopy);
foreach (InputSplit pathSplit in pathSplits)
{
splits.AddItem(new TaggedInputSplit(pathSplit, conf, format.GetType(), mapperClass
));
}
}
}
return(splits);
}
/// <exception cref="System.Exception"/>
public virtual void TestSplitting()
{
Job job = Job.GetInstance();
MiniDFSCluster dfs = null;
try
{
dfs = new MiniDFSCluster.Builder(job.GetConfiguration()).NumDataNodes(4).Racks(new
string[] { "/rack0", "/rack0", "/rack1", "/rack1" }).Hosts(new string[] { "host0"
, "host1", "host2", "host3" }).Build();
FileSystem fs = dfs.GetFileSystem();
Path path = GetPath("/foo/bar", fs);
Path path2 = GetPath("/foo/baz", fs);
Path path3 = GetPath("/bar/bar", fs);
Path path4 = GetPath("/bar/baz", fs);
int numSplits = 100;
FileInputFormat.SetMaxInputSplitSize(job, fs.GetFileStatus(path).GetLen() / numSplits
);
MultipleInputs.AddInputPath(job, path, typeof(TextInputFormat), typeof(TestDelegatingInputFormat.MapClass
));
MultipleInputs.AddInputPath(job, path2, typeof(TextInputFormat), typeof(TestDelegatingInputFormat.MapClass2
));
MultipleInputs.AddInputPath(job, path3, typeof(KeyValueTextInputFormat), typeof(TestDelegatingInputFormat.MapClass
));
MultipleInputs.AddInputPath(job, path4, typeof(TextInputFormat), typeof(TestDelegatingInputFormat.MapClass2
));
DelegatingInputFormat inFormat = new DelegatingInputFormat();
int[] bins = new int[3];
foreach (InputSplit split in (IList <InputSplit>)inFormat.GetSplits(job))
{
NUnit.Framework.Assert.IsTrue(split is TaggedInputSplit);
TaggedInputSplit tis = (TaggedInputSplit)split;
int index = -1;
if (tis.GetInputFormatClass().Equals(typeof(KeyValueTextInputFormat)))
{
// path3
index = 0;
}
else
{
if (tis.GetMapperClass().Equals(typeof(TestDelegatingInputFormat.MapClass)))
{
// path
index = 1;
}
else
{
// path2 and path4
index = 2;
}
}
bins[index]++;
}
NUnit.Framework.Assert.AreEqual("count is not equal to num splits", numSplits, bins
[0]);
NUnit.Framework.Assert.AreEqual("count is not equal to num splits", numSplits, bins
[1]);
NUnit.Framework.Assert.AreEqual("count is not equal to 2 * num splits", numSplits
* 2, bins[2]);
}
finally
{
if (dfs != null)
{
dfs.Shutdown();
}
}
}
public virtual void TestDoMultipleInputs()
{
Path in1Dir = GetDir(In1Dir);
Path in2Dir = GetDir(In2Dir);
Path outDir = GetDir(OutDir);
Configuration conf = CreateJobConf();
FileSystem fs = FileSystem.Get(conf);
fs.Delete(outDir, true);
DataOutputStream file1 = fs.Create(new Path(in1Dir, "part-0"));
file1.WriteBytes("a\nb\nc\nd\ne");
file1.Close();
// write tab delimited to second file because we're doing
// KeyValueInputFormat
DataOutputStream file2 = fs.Create(new Path(in2Dir, "part-0"));
file2.WriteBytes("a\tblah\nb\tblah\nc\tblah\nd\tblah\ne\tblah");
file2.Close();
Job job = Job.GetInstance(conf);
job.SetJobName("mi");
MultipleInputs.AddInputPath(job, in1Dir, typeof(TextInputFormat), typeof(TestMultipleInputs.MapClass
));
MultipleInputs.AddInputPath(job, in2Dir, typeof(KeyValueTextInputFormat), typeof(
TestMultipleInputs.KeyValueMapClass));
job.SetMapOutputKeyClass(typeof(Text));
job.SetMapOutputValueClass(typeof(Text));
job.SetOutputKeyClass(typeof(NullWritable));
job.SetOutputValueClass(typeof(Text));
job.SetReducerClass(typeof(TestMultipleInputs.ReducerClass));
FileOutputFormat.SetOutputPath(job, outDir);
bool success = false;
try
{
success = job.WaitForCompletion(true);
}
catch (Exception ie)
{
throw new RuntimeException(ie);
}
catch (TypeLoadException instante)
{
throw new RuntimeException(instante);
}
if (!success)
{
throw new RuntimeException("Job failed!");
}
// copy bytes a bunch of times for the ease of readLine() - whatever
BufferedReader output = new BufferedReader(new InputStreamReader(fs.Open(new Path
(outDir, "part-r-00000"))));
// reducer should have counted one key from each file
NUnit.Framework.Assert.IsTrue(output.ReadLine().Equals("a 2"));
NUnit.Framework.Assert.IsTrue(output.ReadLine().Equals("b 2"));
NUnit.Framework.Assert.IsTrue(output.ReadLine().Equals("c 2"));
NUnit.Framework.Assert.IsTrue(output.ReadLine().Equals("d 2"));
NUnit.Framework.Assert.IsTrue(output.ReadLine().Equals("e 2"));
}