Pivot

Short Description

The component reads input records and treats them as groups. A group is defined either by a Group key or a number of records forming the group. Pivot then produces a single record from each group. In other words, the component creates a pivot table.

Pivot has two principal attributes which instruct it to treat some input values as output field names and other inputs as output values.

The component is a simple form of Denormalizer.

Note: When using the Group key attribute, input records should be sorted. See Details below.

Ports

Metadata

Pivot does not propagate metadata.
Pivot has no metadata template.

Pivot Attributes

[1] One of the Group key or Group size attributes has to be always set.
[2] These two values can either be given as an attribute or in your own transformation.
[3] One of these attributes has to be set if you do not control the transformation by means of Field defining output field name and Field defining output field value.

Details

You can define the data transformation in two ways:

1. Set the Group key or Group size attributes. See Group Data by Setting Attributes below.
2. Write the transformation yourself in CTL/Java or provide it in an external file/Java class. See Define Your Own Transformation - Java/CTL below.

Group Data by Setting Attributes

Group Data Using Group Key
If you group data using the Group key attribute, your input should be sorted according to Group key values. To tell the component how your input is sorted, specify Sort order. If the Group key fields appear in the output metadata as well, Group key values are copied automatically.

Group Data Using Group Size
When you are grouping using the Group size attribute, the component ignores the data itself, takes e.g. 3 records (for Group size = 3) and treats them as one group. Naturally, you have to have an adequate number of input records otherwise errors on reading will occur. The number has to be a multiple of Group size, e.g. 3, 6, 9 etc. for Group size = 3.

Mapping
There are the two major attributes which describe the "mapping". They say:

• which input field’s value will designate the output field - Field defining output field name
• which input field’s value will be used as a value for that field Field defining output field value

As for the output metadata, it is arbitrary but fixed to field names. If your input data has extra fields, they are simply ignored (only fields defined as a value/name matter). Likewise, output fields without any corresponding input records will be null.

If a value of Field defining output field name does not correspond to any of names of output metadata fields, the component fails.

Define Your Own Transformation - Java/CTL

In Pivot, you can write the transformation function yourself. That can be done either in CTL or Java, see Advanced attributes in Pivot Attributes above.
Before writing the transformation, you might want to refer to some of the sections touching the subject:

• Defining Transformations
• writing transformations in Denormalizer, the component Pivot is derived from: CTL Interface and Java Interface.

CTL Interface

You can implement methods getOutputFieldIndex and getOutputFieldValue or you can set one of the attributes and implement the other one with a method.
So you can, for example, set valueField and implement getOutputFieldIndex. Or you can set nameField and implement getOutputFieldValue.
For a better understanding, examine the methods' documentation directly in the Transform editor.

Java Interface

Compared to Denormalizer, the Pivot component has new significant attributes: nameField and valueField. These can be defined either as attributes (see above) or by methods. If the transformation is not defined, the component uses com.opensys.cloveretl.component.pivot.DataRecordPivotTransform which copies values from valueField to nameField.

In Java, you can implement your own PivotTransform that overrides DataRecordPivotTransform. However, you can override only one method, e.g. getOutputFieldValue, getOutputFieldIndex or others from PivotTransform (that extends RecordDenormalize).

Examples

Data Transformation with Pivot - Using Key

Let us have the following input values:

Because we are going to group the data according to the groupID field, the input has to be sorted (mind the ascending order of groupIDs). In the Pivot component, we will make the following settings:
Group key = groupID (to group all input records with the same groupID)
Field defining output field name = fieldName (to say we want to take output fields' names from this input field)
Field defining output field value = fieldValue (to say we want to take output fields' values from this input field)

Processing that data with Pivot produces the following output:

Notice the input recordNo field has been ignored. Similarly, the output comment had no corresponding fields on the input, that is why it remains null. groupID makes part in the output metadata and thus was copied automatically.

If the input is not sorted (not like in the example), grouping records according to their count is especially handy. Omit Group key and set Group size instead to read sequences of records that have exactly the number of records you need.

Converting fixed number of records to single record

Input metadata have fields filedName and fieldValue. The records contain a timestamp, IP address and username.

timestamp|2014-10-30 13:51:12
address  |192.168.10.15
username |Alice
timestamp|2014-10-30 13:52:14
address  |192.168.3.151
username |Bob
timestamp|2014-10-30 13:52:40
address  |192.168.102.105
username |Eve

Convert the data to a one line structure having a timestamp, IP address and username.

Solution
Use attributes Group size, Field defining output field name and Field defining output field value.

Output metadata has to have fields timestamp, address and user.

Converting fixed number of records to single record using CTL

This example is similar to the previous one: input records contain a timestamp, IP address and username, but there is no field indicating which one is a timestamp, IP address or username. The order of the input records within the group is always the same: timestamp is before IP address and IP address is before username.

2014-10-30 13:51:12
192.168.10.15
Alice
2014-10-30 13:52:14
192.168.3.151
Bob
2014-10-30 13:52:40
192.168.102.105
Eve

Solution
One output record correspond to three input records, so we use the Group size attribute. Mapping to the output record is defined in Pivot transformation.

//#CTL2

function integer getOutputFieldIndex(integer idx) {
    return idx % 3;
}

function string getOutputFieldValue(integer idx) {
    return $in.0.value;
}

The order of input records corresponds to the order of output metadata fields. If you need a different order, rearrange the output metadata or change the content of the getOutputFieldIndex() function.

Passing trough Fields to Output

The input records have customerId, batchId, fieldName and value metadata fields:

C0001|B001|firstName|John
C0001|B001|lastName |Doe
C0001|B001|accountNo|A0001

Convert data to following the format:

C0001|B001|John|Doe|A0001

Solution

Note that Group key fields have been passed to the corresponding output fields.

Best Practices

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

See also

Denormalizer
MetaPivot
Normalizer
Common Properties of Components
Specific attribute types
Common Properties of Transformers