org.apache.hadoop.hive.ql
Class Driver

java.lang.Object
  org.apache.hadoop.hive.ql.Driver

All Implemented Interfaces:

CommandProcessor

public class Driver
extends Object
implements CommandProcessor

Field Summary

protected int

tryCount

Constructor Summary

Driver()

Driver(HiveConf conf)
for backwards compatibility with current tests

Method Summary

int	acquireReadWriteLocks() Acquire read and write locks needed by the statement.
int	close()
int	compile(String command) Compile a new query.
int	compile(String command, boolean resetTaskIds) Compile a new query, but potentially reset taskID counter.
void	destroy()
int	execute()
org.apache.hadoop.mapred.ClusterStatus	getClusterStatus() Return the status information about the Map-Reduce cluster
int	getMaxRows() Return the maximum number of rows returned by getResults
QueryPlan	getPlan()
Query	getQueryPlan()
boolean	getResults(ArrayList<String> res)
Schema	getSchema()
static Schema	getSchema(BaseSemanticAnalyzer sem, HiveConf conf) Get a Schema with fields represented with native Hive types
Schema	getThriftSchema() Get a Schema with fields represented with Thrift DDL types
int	getTryCount()
boolean	hasReduceTasks(List<Task<? extends Serializable>> tasks)
void	init()
void	launchTask(Task<? extends Serializable> tsk, String queryId, boolean noName, Map<TaskResult,TaskRunner> running, String jobname, int jobs, DriverContext cxt) Launches a new task
TaskResult	pollTasks(Set<TaskResult> results) Polls running tasks to see if a task has ended.
void	restoreSession(org.apache.hadoop.hive.ql.Driver.QueryState qs)
CommandProcessorResponse	run(String command)
void	saveSession(org.apache.hadoop.hive.ql.Driver.QueryState qs)
void	setMaxRows(int maxRows) Set the maximum number of rows returned by getResults
void	setTryCount(int tryCount)
void	taskCleanup() Cleans up remaining tasks in case of failure

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

tryCount

protected int tryCount

Constructor Detail

Driver

public Driver(HiveConf conf)

for backwards compatibility with current tests

Driver

public Driver()

Method Detail

init

public void init()

Specified by:

init in interface CommandProcessor

getClusterStatus

public org.apache.hadoop.mapred.ClusterStatus getClusterStatus()
                                                        throws Exception

Return the status information about the Map-Reduce cluster

Throws:

Exception

getSchema

public Schema getSchema()

getSchema

public static Schema getSchema(BaseSemanticAnalyzer sem,
                               HiveConf conf)

Get a Schema with fields represented with native Hive types

getThriftSchema

public Schema getThriftSchema()
                       throws Exception

Get a Schema with fields represented with Thrift DDL types

Throws:

Exception

getMaxRows

public int getMaxRows()

Return the maximum number of rows returned by getResults

setMaxRows

public void setMaxRows(int maxRows)

Set the maximum number of rows returned by getResults

hasReduceTasks

public boolean hasReduceTasks(List<Task<? extends Serializable>> tasks)

compile

public int compile(String command)

Compile a new query. Any currently-planned query associated with this Driver is discarded.

Parameters:

command - The SQL query to compile.

saveSession

public void saveSession(org.apache.hadoop.hive.ql.Driver.QueryState qs)

restoreSession

public void restoreSession(org.apache.hadoop.hive.ql.Driver.QueryState qs)

compile

public int compile(String command,
                   boolean resetTaskIds)

Compile a new query, but potentially reset taskID counter. Not resetting task counter is useful for generating re-entrant QL queries.

Parameters:

command - The HiveQL query to compile

resetTaskIds - Resets taskID counter if true.

Returns:

0 for ok

getPlan

public QueryPlan getPlan()

Returns:

The current query plan associated with this Driver, if any.

acquireReadWriteLocks

public int acquireReadWriteLocks()

Acquire read and write locks needed by the statement. The list of objects to be locked are obtained from he inputs and outputs populated by the compiler. The lock acuisition scheme is pretty simple. If all the locks cannot be obtained, error out. Deadlock is avoided by making sure that the locks are lexicographically sorted.

run

public CommandProcessorResponse run(String command)
                             throws CommandNeedRetryException

Specified by:

run in interface CommandProcessor

Throws:

CommandNeedRetryException

execute

public int execute()
            throws CommandNeedRetryException

Throws:

CommandNeedRetryException

launchTask

public void launchTask(Task<? extends Serializable> tsk,
                       String queryId,
                       boolean noName,
                       Map<TaskResult,TaskRunner> running,
                       String jobname,
                       int jobs,
                       DriverContext cxt)

Launches a new task

Parameters:

tsk - task being launched

queryId - Id of the query containing the task

noName - whether the task has a name set

running - map from taskresults to taskrunners

jobname - name of the task, if it is a map-reduce job

jobs - number of map-reduce jobs

cxt - the driver context

taskCleanup

public void taskCleanup()

Cleans up remaining tasks in case of failure

pollTasks

public TaskResult pollTasks(Set<TaskResult> results)

Polls running tasks to see if a task has ended.

Parameters:

results - Set of result objects for running tasks

Returns:

The result object for any completed/failed task

getResults

public boolean getResults(ArrayList<String> res)
                   throws IOException,
                          CommandNeedRetryException

Throws:

IOException

CommandNeedRetryException

getTryCount

public int getTryCount()

setTryCount

public void setTryCount(int tryCount)

close

public int close()

destroy

public void destroy()

getQueryPlan

public Query getQueryPlan()
                   throws IOException

Throws:

IOException