private static void dumpText(List<RubixFile<Tuple, Object>> rfiles, String output, int numRows)
throws IOException, InterruptedException, ClassNotFoundException, InstantiationException,
IllegalAccessException {
Configuration conf = new JobConf();
int totalBlocks = 0;
for (RubixFile<Tuple, Object> rfile : rfiles) {
Path path = rfile.path;
List<KeyData<Tuple>> keyDataList = rfile.getKeyData();
print.f("--- %s", path.toString());
print.f("Schema: %s", rfile.getSchema().toString());
print.f("PartitionKeys: %s", Arrays.toString(rfile.getPartitionKeys()));
print.f("SortKeys %s", Arrays.toString(rfile.getSortKeys()));
print.f("Block Serialization Type: %s", rfile.getBlockSerializationType());
print.f("Number of blocks: %d", keyDataList.size());
totalBlocks += keyDataList.size();
int cumrows = 0;
for (KeyData<Tuple> keyData : keyDataList) {
print.f(
"Block %s. BlockId: %d (Reducer: %d Index:%d)",
keyData,
keyData.blockId,
(keyData.getBlockId() >> 32),
(keyData.getBlockId() & (((long) 1 << 32) - 1)));
if (numRows > 0) {
RubixInputSplit<Tuple, Object> split =
new RubixInputSplit<Tuple, Object>(
conf,
path,
keyData.getKey(),
keyData.getOffset(),
keyData.getLength(),
keyData.getBlockId(),
keyData.getNumRecords(),
rfile.getKeyClass(),
rfile.getValueClass(),
rfile.getSchema(),
rfile.getBlockSerializationType());
RubixRecordReader<Tuple, Object> recordReader = new RubixRecordReader<Tuple, Object>();
recordReader.initialize(split, conf);
int rows = 0;
while (recordReader.nextKeyValue()) {
rows++;
if (rows < numRows) {
System.out.println("\t" + recordReader.getCurrentValue());
} else {
break;
}
}
cumrows += keyData.getNumRecords();
System.out.println(
String.format("\tRows=%d Cummulative=%d", keyData.getNumRecords(), cumrows));
}
}
}
print.f("Total Blocks: %d", totalBlocks);
}
public static void main(String[] args)
throws IOException, ClassNotFoundException, InterruptedException, ParseException,
InstantiationException, IllegalAccessException {
final int VERBOSE_NUM_ROWS = 4;
Options options = new Options();
options.addOption("h", "help", false, "shows this message");
options.addOption("v", "verbose", false, "print summary and first few rows of each block");
options.addOption("m", "metadata", false, "show the metadata");
options.addOption(
"d",
"dump",
false,
"dump the contents of the rubix file. Use -f for specifying format, and -o for specifying "
+ "output location");
options.addOption(
"f", "format", true, "the data format for dumping data (AVRO or TEXT). Default: TEXT");
options.addOption(
"e",
"extract",
true,
"Extract one or more rubix blocks starting from the given blockId. Use -e blockId,numBlocks "
+ "for specifying the blocks to be extracted. Use -o for specifying output location");
options.addOption("o", true, "Store the output at the specified location");
CommandLineParser parser = new BasicParser();
// parse the command line arguments
CommandLine line = parser.parse(options, args);
// show the help message
if (line.hasOption("h")) {
HelpFormatter formatter = new HelpFormatter();
formatter.printHelp(
"RubixFile <rubix file or dir> [options]\nIf no options are provided, print a summary of the blocks.",
options);
return;
}
// validate provided options
if (line.hasOption("d") && line.hasOption("e")) {
System.err.println("Cannot dump (-d) and extract (-e) at the same time!");
return;
}
// obtain the list of rubix files
String[] files = line.getArgs();
if (files == null || files.length == 0) {
System.err.println("Rubix file not specified");
return;
}
Configuration conf = new JobConf();
FileSystem fs = FileSystem.get(conf);
Path path = new Path(files[0]);
FileStatus[] allFiles;
FileStatus status = fs.getFileStatus(path);
if (status.isDir()) {
allFiles =
fs.listStatus(
path,
new PathFilter() {
@Override
public boolean accept(Path path) {
return path.toString().contains(RubixConstants.RUBIX_EXTENSION);
}
});
} else {
allFiles = new FileStatus[] {status};
}
// walk over all files and extract the trailer section
List<RubixFile<Tuple, Object>> rfiles = new ArrayList<RubixFile<Tuple, Object>>();
for (FileStatus s : allFiles) {
Path p = s.getPath();
RubixFile<Tuple, Object> rfile = new RubixFile<Tuple, Object>(conf, p);
// if printing meta data information.. exit after first file (since all files
// have the same meta data)
if (line.hasOption("m")) {
rfile.getKeyData();
System.out.println(new ObjectMapper().writer().writeValueAsString(rfile.metadataJson));
break;
}
rfiles.add(rfile);
}
// dump the data
if (line.hasOption("d")) {
String format = line.getOptionValue("f");
if (format == null) format = "TEXT";
format = format.trim().toUpperCase();
if (format.equals("AVRO")) {
// dumpAvro(rfiles, line.getOptionValue("o"));
throw new UnsupportedOperationException(
"Dumping to avro is not currently supporting. Please write a Cubert (map-only) script to store data in avro format");
} else if (format.equals("TEXT")) {
if (line.hasOption("o")) {
System.err.println("Dumping TEXT format data *into a file* is not currently supported");
return;
}
dumpText(rfiles, line.getOptionValue("o"), Integer.MAX_VALUE);
} else {
System.err.println("Invalid format [" + format + "] for dumping. Please use AVRO or TEXT");
return;
}
}
// extract arguments: -e blockId,numBlocks(contiguous) -o ouputLocation
else if (line.hasOption("e")) {
String extractArguments = line.getOptionValue("e");
String outputLocation;
if (line.hasOption("o")) {
outputLocation = line.getOptionValue("o");
} else {
System.err.println("Need to specify the location to store the output");
return;
}
long blockId;
int numBlocks = 1;
if (extractArguments.contains(",")) {
String[] splitExtractArgs = extractArguments.split(",");
blockId = Long.parseLong(splitExtractArgs[0]);
numBlocks = Integer.parseInt(splitExtractArgs[1]);
} else {
blockId = Long.parseLong(extractArguments);
}
extract(rfiles, blockId, numBlocks, outputLocation);
} else
// print summary
{
dumpText(rfiles, null, line.hasOption("v") ? VERBOSE_NUM_ROWS : 0);
}
}
private static void extract(
List<RubixFile<Tuple, Object>> rfiles, long blockId, int numBlocks, String output)
throws IOException, ClassNotFoundException, InstantiationException, IllegalAccessException {
Configuration conf = new JobConf();
File outFile = new File(output);
if (outFile.exists()) {
outFile.delete();
}
outFile.createNewFile();
BufferedOutputStream bos = new BufferedOutputStream(new FileOutputStream(outFile));
ByteArrayOutputStream keySectionStream = new ByteArrayOutputStream();
DataOutput keySectionOut = new DataOutputStream(keySectionStream);
SerializationFactory serializationFactory = new SerializationFactory(conf);
RubixFile<Tuple, Object> lastrFile = null;
JsonNode json;
long totalLength = 0;
final int BUF_SIZE = 32 * 1024;
long blockIds[] = new long[numBlocks];
int foundBlocks = 0;
for (int i = 0; i < numBlocks; i++) blockIds[i] = blockId + i;
for (int i = 0; i < numBlocks; i++) {
boolean found = false;
for (RubixFile<Tuple, Object> rfile : rfiles) {
print.f("Checking %s", rfile.path.toString());
List<KeyData<Tuple>> keyDataList = rfile.getKeyData();
for (KeyData<Tuple> keyData : keyDataList) {
if (keyData.getBlockId() == blockIds[i]) {
long offset = keyData.getOffset();
long length = keyData.getLength();
Tuple key = keyData.getKey();
print.f(
"Extracting block %d (off=%d len=%d) from %s",
keyData.getBlockId(), offset, length, rfile.path.toString());
// copy the data
if (length > 0) {
FileSystem fs = FileSystem.get(conf);
FSDataInputStream in = fs.open(rfile.path);
in.seek(offset);
byte[] data = new byte[BUF_SIZE];
long toRead = length;
while (toRead > 0) {
int thisRead = toRead > BUF_SIZE ? BUF_SIZE : (int) toRead;
in.readFully(data, 0, thisRead);
bos.write(data, 0, thisRead);
toRead -= thisRead;
System.out.print(".");
}
System.out.println();
}
// copy the key section
Serializer<Tuple> keySerializer =
serializationFactory.getSerializer(rfile.getKeyClass());
keySerializer.open(keySectionStream);
keySerializer.serialize(key);
keySectionOut.writeLong(totalLength); // position
keySectionOut.writeLong(keyData.getBlockId());
keySectionOut.writeLong(keyData.getNumRecords());
foundBlocks++;
totalLength += length;
lastrFile = rfile;
found = true;
break;
}
}
if (found) {
break;
}
}
if (!found) System.err.println("Cannot locate block with id " + blockIds[i]);
}
byte[] trailerBytes = keySectionStream.toByteArray();
json = JsonUtils.cloneNode(lastrFile.metadataJson);
((ObjectNode) json).put("numberOfBlocks", foundBlocks);
DataOutput out = new DataOutputStream(bos);
out.writeUTF(json.toString());
out.writeInt(trailerBytes.length);
out.write(trailerBytes);
out.writeLong(totalLength); // trailer start offset
bos.close();
}
