Version

    Partition

    Short Description
    Ports
    Metadata
    Partition Attributes
    Details
    CTL Interface
    Java Interface
    Examples
    Best Practices
    See also

    Short Description

    Partition distributes individual input data records among different output ports.

    ComponentSame input metadataSorted inputsInputsOutputsJavaCTLAuto-propagated metadata
    Partition-11-n[1][1]

    [1]  Partition can use either a transformation or two other attributes (Ranges and/or Partition key).

    Ports

    Port typeNumberRequiredDescriptionMetadata
    Input0For input data recordsAny
    Output0For output data recordsInput 0
    1-NFor output data recordsInput 0

    Metadata

    Partition propagates metadata in both directions. Partition does not change priority of propagated metadata.

    Partition has no metadata template.

    Input and output fields can have any data types.

    Metadata on input and output ports cannot differ. (Input and output records can have different names but the metadata fields of both records must be identical.)

    Partition Attributes

    AttributeReqDescriptionPossible values
    Basic
    Partition [1]Definition of the way how records should be distributed among output ports written in the graph in CTL or Java. 
    Partition URL[1]

    The name of the external file, including the path, containing the definition of the way how records should be distributed among output ports written in CTL or Java.

     
    Partition class[1]

    The name of the external class defining the way how records should be distributed among output ports.

     
    Ranges[1] [2]

    Ranges expressed as a sequence of individual ranges separated from each other by a semicolon. Each individual range is a sequence of intervals for some set of fields that are adjacent to each other without any delimiter. It is expressed also whether the minimum and maximum margin is included in the interval or not by a bracket and parenthesis, respectively. Example of Ranges: <1,9)(,31.12.2008);<1,9)<31.12.2008,);<9,)(,31.12.2008);<9,)<31.12.2008).

     
    Partition key[1]  [2]

    Key according to which input records are distributed among different output ports. Expressed as a sequence of individual input field names separated from each other by a semicolon. Example of Partition key: first_name;last_name.

     
    Advanced
    Partition source charset 

    Encoding of the external file defining the transformation.

    The default encoding depends on DEFAULT_SOURCE_CODE_CHARSET in defaultProperties.

    UTF-8 | other encoding
    Deprecated
    Locale 

    Locale to be used when internationalization is set to true. By default, system value is used unless the value of Locale specified in the defaultProperties file is uncommented and set to the desired Locale. For more information on how Locale may be changed in the defaultProperties, see Chapter 17, Engine Configuration.

    system value or specified default value (default) | other locale
    Use internationalization 

    By default, no internationalization is used. If set to true, sorting according to national properties is performed.

    false (default) | true

    [1]  If one of these transformation attributes is specified, both Ranges and Partition key will be ignored since they have lesser priority.

    [2]  If no transformation attribute is defined, Ranges and Partition key are used in one of the three ways as described in details.

    Details

    To distribute data records, user-defined transformation, ranges of Partition key or RoundRobin algorithm may be used. In this component, no mapping may be defined since it does not change input data records. It only distributes them unchanged among output ports.

    Transformation uses a CTL template for Partition or implements a PartitionFunction interface. Its methods are listed below.

    If no transformation attribute is defined, Ranges and Partition key are used in one of following ways:

    • Both Ranges and Partition key are set.

      The records in which the values of the fields are inside the margins of specified range will be sent to the same output port. The number of the output port corresponds to the order of the range within all values of the fields.

    • Ranges are not defined. Only Partition key is set.

      Records will be distributed among output ports in such a way that all records with the same values of Partition key fields will be sent to the same port.

      The output port number will be determined as the hash value computed from the key fields modulo the number of output ports.

    • Neither Ranges nor Partition key are defined.

      RoundRobin algorithm will be used to distribute records among output ports.

    [Tip]Tip

    Note that you can use the Partition component as a filter similarly to Filter. With the Partition component, you can define much more sophisticated filter expressions and distribute input data records among more than 2 outputs.

    Neither Partition nor Filter allow to modify records.

    [Important]Important

    Partition is a high-performance component, thus you cannot modify input and output records - it would result in an error. If you need to do so, consider using Reformat instead.

    CTL Interface

    CTL Templates for Partition (or ParallelPartition)
    Access to input and output fields

    Transformation in CTL can be specified in the Partition or Partition URL attributes.

    CTL Templates for Partition (or ParallelPartition)

    This transformation template is used in Partition, and ParallelPartition.

    You can convert existing transformation in CTL to Java language code using the button at the upper right corner of the tab.

    You can open the transformation definition as another tab of a graph (in addition to the Graph and Source tabs of Graph Editor) by clicking the corresponding button at the upper right corner of the tab.

    Table 57.5. Functions in Partition (or ParallelPartition)

    CTL Template Functions
    void init(integer partitionCount)
    RequiredNo
    DescriptionInitialize the component, setup the environment, global variables.
    InvocationCalled before processing the first record.
    Input Parametersinteger partitionCount
    Returnsvoid
    integer getOutputPort()
    Requiredyes
    Input Parametersnone
    ReturnsInteger numbers. For detailed information, see Return Values of Transformations.
    InvocationCalled repeatedly for each input record.
    Description

    It does not transform the records, it does not change them nor remove them, it only returns integer numbers. Each of these returned numbers is a number of the output port to which individual record should be sent. In ParallelPartition, these ports are virtual and mean Cluster nodes.

    If getOutputPort() fails and user has not defined any getOutputPortOnError(), the whole graph will fail.

    If any part of the getOutputPort() function for some output record causes fail of the getOutputPort() function and if the user has defined the function getOutputPortOnError(), processing continues in getOutputPortOnError() at the place where getOutputPort() failed.

    The getOutputPortOnError() function gets the information gathered by getOutputPort() that was get from previously successfully processed code. Also an error message and stack trace are passed to getOutputPortOnError().

    Example
    function integer getOutputPort() {
        switch (expression) {
          case const0 : return 0; break;
          case const1 : return 1; break;
          ...
          case constN : return N; break;
          [default : return N+1;]
       }
    }
    integer getOutputPortOnError(string errorMessage, string stackTrace)
    Requiredno
    Input Parametersstring errorMessage
    string stackTrace
    ReturnsInteger numbers. For detailed information, see Return Values of Transformations.
    InvocationCalled if getOutputPort() throws an exception.
    Description

    It does not transform the records, it does not change them nor remove them, it only returns integer numbers. Each of these returned numbers is a number of the output port to which individual record should be sent. In ParallelPartition, these ports are virtual and mean Cluster nodes.

    If any part of the getOutputPort() function for some output record causes a failure of the getOutputPort() function and if the user has defined the function getOutputPortOnError(), processing continues in this getOutputPortOnError() at the place where getOutputPort() failed.

    The getOutputPortOnError() function gets the information gathered by getOutputPort() that was get from previously successfully processed code. Also error message and stack trace are passed to getOutputPortOnError().

    Example
    function integer getOutputPortOnError(
                        string errorMessage, 
                        string stackTrace) {
       printErr(errorMessage);
       printErr(stackTrace);
    }
    string getMessage()
    RequiredNo
    DescriptionPrints an error message specified and invoked by user.
    Invocation

    Called in any time specified by user (called only when either getOutputPort() or getOutputPortOnError() returns a value less than or equal to -2).

    Returnsstring
    void preExecute()
    RequiredNo
    Input parametersNone
    Returnsvoid
    Description

    May be used to allocate and initialize resources. All resources allocated within this function should be released by the postExecute() function.

    InvocationCalled during each graph run before the transform is executed.
    void postExecute()
    RequiredNo
    Input parametersNone
    Returnsvoid
    DescriptionShould be used to free any resources allocated withinthe preExecute() function.
    InvocationCalled during each graph run after the entire transform was executed.

    Access to input and output fields

    Input records or fields

    Input records or fields are accessible within the getOutputPort() and getOutputPortOnError() functions only.

    Output records or fields

    Output records or fields are not accessible at all as records are mapped to the output without any modification and mapping.

    [Warning]Warning

    All of the other CTL template functions allow to access neither inputs nor outputs.

    Remember that if you do not hold these rules, NPE will be thrown!

    Java Interface

    The transformation implements methods of the PartitionFunction interface and inherits other common methods from the Transform interface. See Common Java Interfaces and Public CloverDX API.

    Following are the methods of PartitionFunction interface:

    • void init(int numPartitions,RecordKey partitionKey)

      Called before getOutputPort() is used. The numPartitions argument specifies how many partitions should be created. The RecordKey argument is the set of fields composing a key based on which the partition should be determined.

    • boolean supportsDirectRecord()

      Indicates whether the partition function supports operation on serialized records /aka direct. Returns true if the getOutputPort(ByteBuffer) method can be called.

    • int getOutputPort(DataRecord record)

      Returns the port number which should be used for sending data out. For more information about return values and their meaning, see Return Values of Transformations.

    • int getOutputPortOnError(Exception exception, DataRecord record)

      Returns the port number which should be used for sending data out. Called only if getOutputPort(DataRecord) throws an exception.

    • int getOutputPort(ByteBuffer directRecord)

      Returns the port number which should be used for sending data out. For more information about return values and their meaning, see Return Values of Transformations.

    • int getOutputPortOnError(Exception exception, ByteBuffer directRecord)

      Returns port number which should be used for sending data out. Called only if getOutputPort(ByteBuffer) throws an exception.

    Examples

    Simple example
    Partitioning even and odd numbers

    Simple example

    Split data into 2 parts. Each part has to contain the same number of records. The number of records can differ by one if the number of input records is odd.

    Solution

    Place the Partition component into graph and connect the corresponding edges. No attribute has to be set up.

    Partitioning even and odd numbers

    Partition records according to the value of field id. Send record with even id to output port 0 and odd numbers to output port 1. If id is not known, send record to port 2.

    Solution

    Use Partition attribute.

    AttributeValue
    PartitionSee the code below
    //#CTL2
    
    function integer getOutputPort() {
        return $in.0.id % 2;
    }
    
    function integer getOutputPortOnError(string errorMessage, string stackTrace) {
        return 2;
    }

    Best Practices

    If the transformation is specified in an external file (Partition URL), we recommend the user to explicitly specify Partition source charset.