![JAR search and dependency download from the Maven repository](/logo.png)
org.apache.cassandra.hadoop.pig.CassandraStorage Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of cassandra-all Show documentation
Show all versions of cassandra-all Show documentation
Palantir open source project
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.cassandra.hadoop.pig;
import java.io.IOException;
import java.io.UnsupportedEncodingException;
import java.net.URLDecoder;
import java.nio.ByteBuffer;
import java.nio.charset.CharacterCodingException;
import java.util.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.cassandra.auth.PasswordAuthenticator;
import org.apache.cassandra.config.CFMetaData;
import org.apache.cassandra.config.ColumnDefinition;
import org.apache.cassandra.db.marshal.*;
import org.apache.cassandra.db.SystemKeyspace;
import org.apache.cassandra.exceptions.ConfigurationException;
import org.apache.cassandra.exceptions.InvalidRequestException;
import org.apache.cassandra.exceptions.SyntaxException;
import org.apache.cassandra.hadoop.ColumnFamilyRecordReader;
import org.apache.cassandra.hadoop.ConfigHelper;
import org.apache.cassandra.hadoop.HadoopCompat;
import org.apache.cassandra.schema.LegacySchemaTables;
import org.apache.cassandra.serializers.CollectionSerializer;
import org.apache.cassandra.thrift.*;
import org.apache.cassandra.utils.ByteBufferUtil;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.Hex;
import org.apache.cassandra.utils.UUIDGen;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.mapreduce.*;
import org.apache.pig.Expression;
import org.apache.pig.LoadFunc;
import org.apache.pig.LoadMetadata;
import org.apache.pig.ResourceSchema;
import org.apache.pig.ResourceStatistics;
import org.apache.pig.StoreFuncInterface;
import org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.PigSplit;
import org.apache.pig.data.*;
import org.apache.pig.impl.util.UDFContext;
import org.apache.pig.ResourceSchema.ResourceFieldSchema;
import org.apache.thrift.TDeserializer;
import org.apache.thrift.TException;
import org.apache.thrift.TSerializer;
import org.apache.thrift.protocol.TBinaryProtocol;
/**
* A LoadStoreFunc for retrieving data from and storing data to Cassandra
*
* A row from a standard CF will be returned as nested tuples: (key, ((name1, val1), (name2, val2))).
*/
@Deprecated
public class CassandraStorage extends LoadFunc implements StoreFuncInterface, LoadMetadata
{
public final static String PIG_ALLOW_DELETES = "PIG_ALLOW_DELETES";
public final static String PIG_WIDEROW_INPUT = "PIG_WIDEROW_INPUT";
public final static String PIG_USE_SECONDARY = "PIG_USE_SECONDARY";
private final static ByteBuffer BOUND = ByteBufferUtil.EMPTY_BYTE_BUFFER;
private static final Logger logger = LoggerFactory.getLogger(CassandraStorage.class);
private ByteBuffer slice_start = BOUND;
private ByteBuffer slice_end = BOUND;
private boolean slice_reverse = false;
private boolean allow_deletes = false;
private RecordReader> reader;
private RecordWriter> writer;
private boolean widerows = false;
private int limit;
protected String DEFAULT_INPUT_FORMAT;
protected String DEFAULT_OUTPUT_FORMAT;
protected enum MarshallerType { COMPARATOR, DEFAULT_VALIDATOR, KEY_VALIDATOR, SUBCOMPARATOR };
protected String username;
protected String password;
protected String keyspace;
protected String column_family;
protected String loadSignature;
protected String storeSignature;
protected Configuration conf;
protected String inputFormatClass;
protected String outputFormatClass;
protected int splitSize = 64 * 1024;
protected String partitionerClass;
protected boolean usePartitionFilter = false;
protected String initHostAddress;
protected String rpcPort;
protected int nativeProtocolVersion = 1;
// wide row hacks
private ByteBuffer lastKey;
private Map lastRow;
private boolean hasNext = true;
public CassandraStorage()
{
this(1024);
}
/**@param limit number of columns to fetch in a slice */
public CassandraStorage(int limit)
{
super();
this.limit = limit;
DEFAULT_INPUT_FORMAT = "org.apache.cassandra.hadoop.ColumnFamilyInputFormat";
DEFAULT_OUTPUT_FORMAT = "org.apache.cassandra.hadoop.ColumnFamilyOutputFormat";
}
public int getLimit()
{
return limit;
}
public void prepareToRead(RecordReader reader, PigSplit split)
{
this.reader = reader;
}
/** read wide row*/
public Tuple getNextWide() throws IOException
{
CfDef cfDef = getCfDef(loadSignature);
ByteBuffer key = null;
Tuple tuple = null;
DefaultDataBag bag = new DefaultDataBag();
try
{
while(true)
{
hasNext = reader.nextKeyValue();
if (!hasNext)
{
if (tuple == null)
tuple = TupleFactory.getInstance().newTuple();
if (lastRow != null)
{
if (tuple.size() == 0) // lastRow is a new one
{
key = reader.getCurrentKey();
tuple = keyToTuple(key, cfDef, parseType(cfDef.getKey_validation_class()));
}
for (Map.Entry entry : lastRow.entrySet())
{
bag.add(columnToTuple(entry.getValue(), cfDef, parseType(cfDef.getComparator_type())));
}
lastKey = null;
lastRow = null;
tuple.append(bag);
return tuple;
}
else
{
if (tuple.size() == 1) // rare case of just one wide row, key already set
{
tuple.append(bag);
return tuple;
}
else
return null;
}
}
if (key != null && !(reader.getCurrentKey()).equals(key)) // key changed
{
// read too much, hold on to it for next time
lastKey = reader.getCurrentKey();
lastRow = reader.getCurrentValue();
// but return what we have so far
tuple.append(bag);
return tuple;
}
if (key == null) // only set the key on the first iteration
{
key = reader.getCurrentKey();
if (lastKey != null && !(key.equals(lastKey))) // last key only had one value
{
if (tuple == null)
tuple = keyToTuple(lastKey, cfDef, parseType(cfDef.getKey_validation_class()));
else
addKeyToTuple(tuple, lastKey, cfDef, parseType(cfDef.getKey_validation_class()));
for (Map.Entry entry : lastRow.entrySet())
{
bag.add(columnToTuple(entry.getValue(), cfDef, parseType(cfDef.getComparator_type())));
}
tuple.append(bag);
lastKey = key;
lastRow = reader.getCurrentValue();
return tuple;
}
if (tuple == null)
tuple = keyToTuple(key, cfDef, parseType(cfDef.getKey_validation_class()));
else
addKeyToTuple(tuple, lastKey, cfDef, parseType(cfDef.getKey_validation_class()));
}
SortedMap row =
(SortedMap)reader.getCurrentValue();
if (lastRow != null) // prepend what was read last time
{
for (Map.Entry entry : lastRow.entrySet())
{
bag.add(columnToTuple(entry.getValue(), cfDef, parseType(cfDef.getComparator_type())));
}
lastKey = null;
lastRow = null;
}
for (Map.Entry entry : row.entrySet())
{
bag.add(columnToTuple(entry.getValue(), cfDef, parseType(cfDef.getComparator_type())));
}
}
}
catch (InterruptedException e)
{
throw new IOException(e.getMessage());
}
}
/** read next row */
public Tuple getNext() throws IOException
{
if (widerows)
return getNextWide();
try
{
// load the next pair
if (!reader.nextKeyValue())
return null;
CfDef cfDef = getCfDef(loadSignature);
ByteBuffer key = reader.getCurrentKey();
Map cf = reader.getCurrentValue();
assert key != null && cf != null;
// output tuple, will hold the key, each indexed column in a tuple, then a bag of the rest
// NOTE: we're setting the tuple size here only for the key so we can use setTupleValue on it
Tuple tuple = keyToTuple(key, cfDef, parseType(cfDef.getKey_validation_class()));
DefaultDataBag bag = new DefaultDataBag();
// we must add all the indexed columns first to match the schema
Map added = new HashMap(cfDef.column_metadata.size());
// take care to iterate these in the same order as the schema does
for (ColumnDef cdef : cfDef.column_metadata)
{
boolean hasColumn = false;
boolean cql3Table = false;
try
{
hasColumn = cf.containsKey(cdef.name);
}
catch (Exception e)
{
cql3Table = true;
}
if (hasColumn)
{
tuple.append(columnToTuple(cf.get(cdef.name), cfDef, parseType(cfDef.getComparator_type())));
}
else if (!cql3Table)
{ // otherwise, we need to add an empty tuple to take its place
tuple.append(TupleFactory.getInstance().newTuple());
}
added.put(cdef.name, true);
}
// now add all the other columns
for (Map.Entry entry : cf.entrySet())
{
if (!added.containsKey(entry.getKey()))
bag.add(columnToTuple(entry.getValue(), cfDef, parseType(cfDef.getComparator_type())));
}
tuple.append(bag);
// finally, special top-level indexes if needed
if (usePartitionFilter)
{
for (ColumnDef cdef : getIndexes())
{
Tuple throwaway = columnToTuple(cf.get(cdef.name), cfDef, parseType(cfDef.getComparator_type()));
tuple.append(throwaway.get(1));
}
}
return tuple;
}
catch (InterruptedException e)
{
throw new IOException(e.getMessage());
}
}
/** write next row */
public void putNext(Tuple t) throws IOException
{
/*
We support two cases for output:
First, the original output:
(key, (name, value), (name,value), {(name,value)}) (tuples or bag is optional)
For supers, we only accept the original output.
*/
if (t.size() < 1)
{
// simply nothing here, we can't even delete without a key
logger.warn("Empty output skipped, filter empty tuples to suppress this warning");
return;
}
ByteBuffer key = objToBB(t.get(0));
if (t.getType(1) == DataType.TUPLE)
writeColumnsFromTuple(key, t, 1);
else if (t.getType(1) == DataType.BAG)
{
if (t.size() > 2)
throw new IOException("No arguments allowed after bag");
writeColumnsFromBag(key, (DataBag) t.get(1));
}
else
throw new IOException("Second argument in output must be a tuple or bag");
}
/** set hadoop cassandra connection settings */
protected void setConnectionInformation() throws IOException
{
StorageHelper.setConnectionInformation(conf);
if (System.getenv(StorageHelper.PIG_INPUT_FORMAT) != null)
inputFormatClass = getFullyQualifiedClassName(System.getenv(StorageHelper.PIG_INPUT_FORMAT));
else
inputFormatClass = DEFAULT_INPUT_FORMAT;
if (System.getenv(StorageHelper.PIG_OUTPUT_FORMAT) != null)
outputFormatClass = getFullyQualifiedClassName(System.getenv(StorageHelper.PIG_OUTPUT_FORMAT));
else
outputFormatClass = DEFAULT_OUTPUT_FORMAT;
if (System.getenv(PIG_ALLOW_DELETES) != null)
allow_deletes = Boolean.parseBoolean(System.getenv(PIG_ALLOW_DELETES));
}
/** get the full class name */
protected String getFullyQualifiedClassName(String classname)
{
return classname.contains(".") ? classname : "org.apache.cassandra.hadoop." + classname;
}
/** set read configuration settings */
public void setLocation(String location, Job job) throws IOException
{
conf = HadoopCompat.getConfiguration(job);
setLocationFromUri(location);
if (ConfigHelper.getInputSlicePredicate(conf) == null)
{
SliceRange range = new SliceRange(slice_start, slice_end, slice_reverse, limit);
SlicePredicate predicate = new SlicePredicate().setSlice_range(range);
ConfigHelper.setInputSlicePredicate(conf, predicate);
}
if (System.getenv(PIG_WIDEROW_INPUT) != null)
widerows = Boolean.parseBoolean(System.getenv(PIG_WIDEROW_INPUT));
if (System.getenv(PIG_USE_SECONDARY) != null)
usePartitionFilter = Boolean.parseBoolean(System.getenv(PIG_USE_SECONDARY));
if (System.getenv(StorageHelper.PIG_INPUT_SPLIT_SIZE) != null)
{
try
{
ConfigHelper.setInputSplitSize(conf, Integer.parseInt(System.getenv(StorageHelper.PIG_INPUT_SPLIT_SIZE)));
}
catch (NumberFormatException e)
{
throw new IOException("PIG_INPUT_SPLIT_SIZE is not a number", e);
}
}
if (usePartitionFilter && getIndexExpressions() != null)
ConfigHelper.setInputRange(conf, getIndexExpressions());
if (username != null && password != null)
ConfigHelper.setInputKeyspaceUserNameAndPassword(conf, username, password);
if (splitSize > 0)
ConfigHelper.setInputSplitSize(conf, splitSize);
if (partitionerClass!= null)
ConfigHelper.setInputPartitioner(conf, partitionerClass);
if (rpcPort != null)
ConfigHelper.setInputRpcPort(conf, rpcPort);
if (initHostAddress != null)
ConfigHelper.setInputInitialAddress(conf, initHostAddress);
ConfigHelper.setInputColumnFamily(conf, keyspace, column_family, widerows);
setConnectionInformation();
if (ConfigHelper.getInputRpcPort(conf) == 0)
throw new IOException("PIG_INPUT_RPC_PORT or PIG_RPC_PORT environment variable not set");
if (ConfigHelper.getInputInitialAddress(conf) == null)
throw new IOException("PIG_INPUT_INITIAL_ADDRESS or PIG_INITIAL_ADDRESS environment variable not set");
if (ConfigHelper.getInputPartitioner(conf) == null)
throw new IOException("PIG_INPUT_PARTITIONER or PIG_PARTITIONER environment variable not set");
if (loadSignature == null)
loadSignature = location;
initSchema(loadSignature);
}
/** set store configuration settings */
public void setStoreLocation(String location, Job job) throws IOException
{
conf = HadoopCompat.getConfiguration(job);
// don't combine mappers to a single mapper per node
conf.setBoolean("pig.noSplitCombination", true);
setLocationFromUri(location);
if (username != null && password != null)
ConfigHelper.setOutputKeyspaceUserNameAndPassword(conf, username, password);
if (splitSize > 0)
ConfigHelper.setInputSplitSize(conf, splitSize);
if (partitionerClass!= null)
ConfigHelper.setOutputPartitioner(conf, partitionerClass);
if (rpcPort != null)
{
ConfigHelper.setOutputRpcPort(conf, rpcPort);
ConfigHelper.setInputRpcPort(conf, rpcPort);
}
if (initHostAddress != null)
{
ConfigHelper.setOutputInitialAddress(conf, initHostAddress);
ConfigHelper.setInputInitialAddress(conf, initHostAddress);
}
ConfigHelper.setOutputColumnFamily(conf, keyspace, column_family);
setConnectionInformation();
if (ConfigHelper.getOutputRpcPort(conf) == 0)
throw new IOException("PIG_OUTPUT_RPC_PORT or PIG_RPC_PORT environment variable not set");
if (ConfigHelper.getOutputInitialAddress(conf) == null)
throw new IOException("PIG_OUTPUT_INITIAL_ADDRESS or PIG_INITIAL_ADDRESS environment variable not set");
if (ConfigHelper.getOutputPartitioner(conf) == null)
throw new IOException("PIG_OUTPUT_PARTITIONER or PIG_PARTITIONER environment variable not set");
// we have to do this again here for the check in writeColumnsFromTuple
if (System.getenv(PIG_USE_SECONDARY) != null)
usePartitionFilter = Boolean.parseBoolean(System.getenv(PIG_USE_SECONDARY));
initSchema(storeSignature);
}
/** Methods to get the column family schema from Cassandra */
protected void initSchema(String signature) throws IOException
{
Properties properties = UDFContext.getUDFContext().getUDFProperties(CassandraStorage.class);
// Only get the schema if we haven't already gotten it
if (!properties.containsKey(signature))
{
try
{
Cassandra.Client client = ConfigHelper.getClientFromInputAddressList(conf);
client.set_keyspace(keyspace);
if (username != null && password != null)
{
Map credentials = new HashMap(2);
credentials.put(PasswordAuthenticator.USERNAME_KEY, username);
credentials.put(PasswordAuthenticator.PASSWORD_KEY, password);
try
{
client.login(new AuthenticationRequest(credentials));
}
catch (AuthenticationException e)
{
logger.error("Authentication exception: invalid username and/or password");
throw new IOException(e);
}
}
// compose the CfDef for the columfamily
CfDef cfDef = getCfDef(client);
if (cfDef != null)
{
StringBuilder sb = new StringBuilder();
sb.append(cfdefToString(cfDef));
properties.setProperty(signature, sb.toString());
}
else
throw new IOException(String.format("Table '%s' not found in keyspace '%s'",
column_family,
keyspace));
}
catch (Exception e)
{
throw new IOException(e);
}
}
}
public void checkSchema(ResourceSchema schema) throws IOException
{
// we don't care about types, they all get casted to ByteBuffers
}
/** define the schema */
public ResourceSchema getSchema(String location, Job job) throws IOException
{
setLocation(location, job);
CfDef cfDef = getCfDef(loadSignature);
if (cfDef.column_type.equals("Super"))
return null;
/*
Our returned schema should look like this:
(key, index1:(name, value), index2:(name, value), columns:{(name, value)})
Which is to say, columns that have metadata will be returned as named tuples, but unknown columns will go into a bag.
This way, wide rows can still be handled by the bag, but known columns can easily be referenced.
*/
// top-level schema, no type
ResourceSchema schema = new ResourceSchema();
// get default marshallers and validators
Map marshallers = getDefaultMarshallers(cfDef);
Map validators = getValidatorMap(cfDef);
// add key
ResourceFieldSchema keyFieldSchema = new ResourceFieldSchema();
keyFieldSchema.setName("key");
keyFieldSchema.setType(StorageHelper.getPigType(marshallers.get(MarshallerType.KEY_VALIDATOR)));
ResourceSchema bagSchema = new ResourceSchema();
ResourceFieldSchema bagField = new ResourceFieldSchema();
bagField.setType(DataType.BAG);
bagField.setName("columns");
// inside the bag, place one tuple with the default comparator/validator schema
ResourceSchema bagTupleSchema = new ResourceSchema();
ResourceFieldSchema bagTupleField = new ResourceFieldSchema();
bagTupleField.setType(DataType.TUPLE);
ResourceFieldSchema bagcolSchema = new ResourceFieldSchema();
ResourceFieldSchema bagvalSchema = new ResourceFieldSchema();
bagcolSchema.setName("name");
bagvalSchema.setName("value");
bagcolSchema.setType(StorageHelper.getPigType(marshallers.get(MarshallerType.COMPARATOR)));
bagvalSchema.setType(StorageHelper.getPigType(marshallers.get(MarshallerType.DEFAULT_VALIDATOR)));
bagTupleSchema.setFields(new ResourceFieldSchema[] { bagcolSchema, bagvalSchema });
bagTupleField.setSchema(bagTupleSchema);
bagSchema.setFields(new ResourceFieldSchema[] { bagTupleField });
bagField.setSchema(bagSchema);
// will contain all fields for this schema
List allSchemaFields = new ArrayList();
// add the key first, then the indexed columns, and finally the bag
allSchemaFields.add(keyFieldSchema);
if (!widerows)
{
// defined validators/indexes
for (ColumnDef cdef : cfDef.column_metadata)
{
// make a new tuple for each col/val pair
ResourceSchema innerTupleSchema = new ResourceSchema();
ResourceFieldSchema innerTupleField = new ResourceFieldSchema();
innerTupleField.setType(DataType.TUPLE);
innerTupleField.setSchema(innerTupleSchema);
innerTupleField.setName(new String(cdef.getName()));
ResourceFieldSchema idxColSchema = new ResourceFieldSchema();
idxColSchema.setName("name");
idxColSchema.setType(StorageHelper.getPigType(marshallers.get(MarshallerType.COMPARATOR)));
ResourceFieldSchema valSchema = new ResourceFieldSchema();
AbstractType validator = validators.get(cdef.name);
if (validator == null)
validator = marshallers.get(MarshallerType.DEFAULT_VALIDATOR);
valSchema.setName("value");
valSchema.setType(StorageHelper.getPigType(validator));
innerTupleSchema.setFields(new ResourceFieldSchema[] { idxColSchema, valSchema });
allSchemaFields.add(innerTupleField);
}
}
// bag at the end for unknown columns
allSchemaFields.add(bagField);
// add top-level index elements if needed
if (usePartitionFilter)
{
for (ColumnDef cdef : getIndexes())
{
ResourceFieldSchema idxSchema = new ResourceFieldSchema();
idxSchema.setName("index_" + new String(cdef.getName()));
AbstractType validator = validators.get(cdef.name);
if (validator == null)
validator = marshallers.get(MarshallerType.DEFAULT_VALIDATOR);
idxSchema.setType(StorageHelper.getPigType(validator));
allSchemaFields.add(idxSchema);
}
}
// top level schema contains everything
schema.setFields(allSchemaFields.toArray(new ResourceFieldSchema[allSchemaFields.size()]));
return schema;
}
/** set partition filter */
public void setPartitionFilter(Expression partitionFilter) throws IOException
{
UDFContext context = UDFContext.getUDFContext();
Properties property = context.getUDFProperties(CassandraStorage.class);
property.setProperty(StorageHelper.PARTITION_FILTER_SIGNATURE, indexExpressionsToString(filterToIndexExpressions(partitionFilter)));
}
/** prepare writer */
public void prepareToWrite(RecordWriter writer)
{
this.writer = writer;
}
/** convert object to ByteBuffer */
protected ByteBuffer objToBB(Object o)
{
if (o == null)
return nullToBB();
if (o instanceof java.lang.String)
return ByteBuffer.wrap(new DataByteArray((String)o).get());
if (o instanceof Integer)
return Int32Type.instance.decompose((Integer)o);
if (o instanceof Long)
return LongType.instance.decompose((Long)o);
if (o instanceof Float)
return FloatType.instance.decompose((Float)o);
if (o instanceof Double)
return DoubleType.instance.decompose((Double)o);
if (o instanceof UUID)
return ByteBuffer.wrap(UUIDGen.decompose((UUID) o));
if(o instanceof Tuple) {
List