public static void validateKeyRange(KeyRange range) throws InvalidRequestException {
if ((range.start_key == null) != (range.end_key == null)) {
throw new InvalidRequestException(
"start key and end key must either both be non-null, or both be null");
}
if ((range.start_token == null) != (range.end_token == null)) {
throw new InvalidRequestException(
"start token and end token must either both be non-null, or both be null");
}
if ((range.start_key == null) == (range.start_token == null)) {
throw new InvalidRequestException(
"exactly one of {start key, end key} or {start token, end token} must be specified");
}
if (range.start_key != null) {
IPartitioner p = StorageService.getPartitioner();
Token startToken = p.getToken(range.start_key);
Token endToken = p.getToken(range.end_key);
if (startToken.compareTo(endToken) > 0 && !endToken.equals(p.getMinimumToken())) {
if (p instanceof RandomPartitioner)
throw new InvalidRequestException(
"start key's md5 sorts after end key's md5. this is not allowed; you probably should not specify end key at all, under RandomPartitioner");
else
throw new InvalidRequestException(
"start key must sort before (or equal to) finish key in your partitioner!");
}
}
if (range.count <= 0) {
throw new InvalidRequestException("maxRows must be positive");
}
}
public TokenRange getRange(ByteBuffer key) {
Token t = partitioner.getToken(key);
com.datastax.driver.core.Token driverToken =
metadata.newToken(partitioner.getTokenFactory().toString(t));
for (TokenRange range : rangeMap.keySet()) {
if (range.contains(driverToken)) {
return range;
}
}
throw new RuntimeException(
"Invalid token information returned by describe_ring: " + rangeMap);
}
public ByteBuffer execute(List<ByteBuffer> parameters) throws InvalidRequestException {
ColumnNameBuilder builder = cfDef.getKeyNameBuilder();
for (ByteBuffer bb : parameters) builder.add(bb);
return partitioner.getTokenFactory().toByteArray(partitioner.getToken(builder.build()));
}
public List<InputSplit> getSplits(JobContext context) throws IOException {
Configuration conf = context.getConfiguration();
validateConfiguration(conf);
// cannonical ranges and nodes holding replicas
List<TokenRange> masterRangeNodes = getRangeMap(conf);
keyspace = ConfigHelper.getInputKeyspace(context.getConfiguration());
cfName = ConfigHelper.getInputColumnFamily(context.getConfiguration());
partitioner = ConfigHelper.getInputPartitioner(context.getConfiguration());
logger.debug("partitioner is " + partitioner);
// cannonical ranges, split into pieces, fetching the splits in parallel
ExecutorService executor = Executors.newCachedThreadPool();
List<InputSplit> splits = new ArrayList<InputSplit>();
try {
List<Future<List<InputSplit>>> splitfutures = new ArrayList<Future<List<InputSplit>>>();
KeyRange jobKeyRange = ConfigHelper.getInputKeyRange(conf);
Range<Token> jobRange = null;
if (jobKeyRange != null) {
if (jobKeyRange.start_key == null) {
logger.warn("ignoring jobKeyRange specified without start_key");
} else {
if (!partitioner.preservesOrder())
throw new UnsupportedOperationException(
"KeyRange based on keys can only be used with a order preserving paritioner");
if (jobKeyRange.start_token != null)
throw new IllegalArgumentException("only start_key supported");
if (jobKeyRange.end_token != null)
throw new IllegalArgumentException("only start_key supported");
jobRange =
new Range<Token>(
partitioner.getToken(jobKeyRange.start_key),
partitioner.getToken(jobKeyRange.end_key),
partitioner);
}
}
for (TokenRange range : masterRangeNodes) {
if (jobRange == null) {
// for each range, pick a live owner and ask it to compute bite-sized splits
splitfutures.add(executor.submit(new SplitCallable(range, conf)));
} else {
Range<Token> dhtRange =
new Range<Token>(
partitioner.getTokenFactory().fromString(range.start_token),
partitioner.getTokenFactory().fromString(range.end_token),
partitioner);
if (dhtRange.intersects(jobRange)) {
for (Range<Token> intersection : dhtRange.intersectionWith(jobRange)) {
range.start_token = partitioner.getTokenFactory().toString(intersection.left);
range.end_token = partitioner.getTokenFactory().toString(intersection.right);
// for each range, pick a live owner and ask it to compute bite-sized splits
splitfutures.add(executor.submit(new SplitCallable(range, conf)));
}
}
}
}
// wait until we have all the results back
for (Future<List<InputSplit>> futureInputSplits : splitfutures) {
try {
splits.addAll(futureInputSplits.get());
} catch (Exception e) {
throw new IOException("Could not get input splits", e);
}
}
} finally {
executor.shutdownNow();
}
assert splits.size() > 0;
Collections.shuffle(splits, new Random(System.nanoTime()));
return splits;
}
