water

Class ExternalFrameReaderClient

java.lang.Object

water.ExternalFrameReaderClient

public final class ExternalFrameReaderClient
extends java.lang.Object

This class is used to read data from H2O Frames from non-H2O environments, such as Spark Executors. It is expected that the frame we want to read is already in the DKV. The check for the presence is up on the user of this class.

Example usage of this class:
First we need to open the connection to H2O and initialize the reader:

 // specify indexes of columns we want to read data from
 int[] selectedColumnIndices = {0, 1};
 // specify expected types for the selected columns
 byte[] expectedTypes = {ExternalFrameHandler.EXPECTED_BOOL, ExternalFrameHandler.EXPECTED_INT};
 ByteChannel channel = ExternalFrameUtils.getConnection("ip:port");
 ExternalFrameReader reader = new ExternalFrameReader(channel, "frameName", 0, selectedColumnIndices);
 
 

In the next step we can read the data we expect, in our case boolean and integer:

 int rowsRead = 0;
 while(rowsRead < reader.getNumRows){
     boolean b = reader.readBool();
     if(reader.isLastNA{
         // it is NA
     }else{
         // it is value
     }

     int i = reader.readInt()
     if(reader.isLastNA{
     // it is NA
     }else{
         // it is value
     }
 }
 
 

And at the end we need to make sure to force to code wait for all data to be read:

 reader.waitUntilAllReceived();
 
 

Constructor Summary

Constructors

Constructor and Description
ExternalFrameReaderClient(java.nio.channels.ByteChannel channel, java.lang.String frameKey, int chunkIdx, int[] selectedColumnIndices, byte[] expectedTypes)

Method Summary

Methods

Modifier and Type	Method and Description
int	getNumRows()
boolean	isLastNA() This method is used to check if the last received value was marked as NA by H2O backend
boolean	readBoolean()
byte	readByte()
char	readChar()
double	readDouble()
float	readFloat()
int	readInt()
long	readLong()
short	readShort()
java.lang.String	readString()
java.sql.Timestamp	readTimestamp()
void	waitUntilAllReceived(int timeout) This method ensures the application waits for all bytes to be received before continuing in the application's control flow.

Methods inherited from class java.lang.Object

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

Constructor Detail

ExternalFrameReaderClient

public ExternalFrameReaderClient(java.nio.channels.ByteChannel channel,
                         java.lang.String frameKey,
                         int chunkIdx,
                         int[] selectedColumnIndices,
                         byte[] expectedTypes)
                          throws java.io.IOException

Parameters:

channel - channel to h2o node

frameKey - name of frame we want to read from

chunkIdx - chunk index from we want to read

selectedColumnIndices - indices of columns we want to read from

expectedTypes - expected types

Throws:

java.io.IOException

Method Detail

getNumRows

public int getNumRows()

readBoolean

public boolean readBoolean()

readByte

public byte readByte()

readChar

public char readChar()

readShort

public short readShort()

readInt

public int readInt()

readLong

public long readLong()

readFloat

public float readFloat()

readDouble

public double readDouble()

readString

public java.lang.String readString()

readTimestamp

public java.sql.Timestamp readTimestamp()

isLastNA

public boolean isLastNA()

This method is used to check if the last received value was marked as NA by H2O backend

waitUntilAllReceived

public void waitUntilAllReceived(int timeout)
                          throws ExternalFrameConfirmationException

This method ensures the application waits for all bytes to be received before continuing in the application's control flow. It has to be called at the end of reading.

Parameters:

timeout - timeout in seconds

Throws:

ExternalFrameConfirmationException