Page tree


Support | BlogContact Us | 844.332.2821



This documentation applies to Trifacta Wrangler. Download this free product.
Registered users of this product or Trifacta Wrangler Enterprise should login to Product Docs through the application.


The following changes have been applied to Wrangle in this release.

Release 4.1

Standardization page and transform have been removed

For a number of releases, the Standardization page and its related transform have been available via feature flag, due to a number of issues.

NOTE: This feature was not typically enabled in Trifacta Wrangler Enterprise deployments.

This feature has been removed from the product altogether and will be replaced in the future by a more robust standardization and normalization capability. 

Syntax Changes

New Transforms

Transform NameDocumentationNotes
commentComment Transform

In previous releases, you could insert comments of the following format:

// This is a comment.

Beginning in Release 4.1, comments are supported by formal transform.

New Functions

Function NameDocumentation

Release 4.0.1

Map data type is now Object data type

The Map data type has been renamed to the Object data type. There are no changes to the behavior.

NOTE: This change is not reflected in the Release 4.0.1 PDF documentation.


Splitrows transform now permits specifying of quote escaping character

For text-based formats that format fields between quotes, you can specify the character that is used to signify the escaping of the quote character in the data. The quoteEscapeChar parameter identifies the character in the data the precedes quotes that are supposed to be part of the data, instead of the marker for a field. See Splitrows Transform.

Release 4.0

Script steps displayed in natural language

In Release 3.2.1 and earlier, the steps of your recipe were displayed in raw form of Wrangle, as in the following example:

split col: column1 on: ',' limit: 5 quote: '\"'

Beginning in Release 4.0, by default, recipe steps are displayed in a more natural form of language, so that you can read the intention of the step without having to understand the details of the underlying language syntax. In natural language format, the above step is rendered as the following:

Split column1 on ',' 5 times


  • If you edit a natural language version of your step, you perform your edits in Wrangle
  • If preferred, you can switch back to displaying in source Wrangle. In the data grid, click the Data Grid options button. Select Show Wrangle Script. See Data Grid Panel.
  • Recipe steps are listed in the product documentation in source Wrangle, so that you can copy and paste them into the Transform Builder as needed.

setderive, and window transforms can now perform any type of computation

To support the above capabilities, the following changes appear in the language:

Multi-column input support

  • set and settype transforms now support multiple input columns.
  • When working with multiple columns, set transform now accepts a placeholder variable in the formula. 
  • See Set Transform.
  • See Settype Transform.

Syntax Changes

Terminology Changes

rollingaverage function accepts two windowing parameters.

In Release 3.2.1 and earlier, you could specify your window for computing the rolling average using a single parameter.

window value: ROLLINGAVG(POS_Sales, 3) order: Whse_Nbr

This single parameter determined the row offset after the current row. The above transform captures a window of values from the current row forward two rows for the rolling average value. There was no way to capture a window that included values that were both before and after the current row.

Beginning in Release 4.0, the function accepts an additional parameter, which enables computation across a before/after window. The following example computes the rolling average from two rows before and two rows after the current row:

window value: ROLLINGAVERAGE(POS_Sales, 3, 2) order: Whse_Nbr


  • The function name has changed to ROLLINGAVERAGE.
  • The behavior of the first parameter has changed. It captures rows before the current one, instead of rows after the current one.
    • The default values are -1 and 0, which capture all values from the current row back to the first row of the dataset.
  • During the upgrade process, transform steps using this function are automatically migrated to the new method of specification. 
  • For more information, see ROLLINGAVERAGE Function.
Function Changes

The following name changes have been applied to existing functions to use more familiar names.

Old Function NameNew Function NameNotes

New Datetime functions

These functions generate date and timestamps at execution time:

Function NameDescription
NOWSee NOW Function.
TODAYSee TODAY Function.

New conditional functions

These functions are conditionals based on data validation against a column's data type:

Function NameDescription


See IFMISSING Function.


See IFNULL Function.


See IFVALID Function.



These functions compute specific values based on conditionals:

Function NameDescription
ANYIFSee ANYIF Function.
See AVERAGEIF Function.
See COUNTAIF Function.
See COUNTIF Function.
See .LISTIF Function.
See MAXIF Function.
See MINIF Function.
See SUMIF Function.
See VARIF Function.

Other new functions

Function NameDescription

 transform no longer accepts row parameter

In Release 3.2.1 and earlier, the row parameter could be used to filter the rows in a dataset to which the set transform value is applied, as in the following example:

set col: results value: 'Outstanding!' row: (score == 100)

Beginning in Release 4.0, the row parameter has been removed. Instead, you can specify conditionals in the value parameter. During upgrade, the above transform step is converted to the following:

set col: results value: IF(score == 100, 'Outstanding!', '')

In addition to the standard IF function, you can apply any of the new conditional functions listed below.

For more information, see Set Transform

Format string with # before 0 is no longer supported in the NUMFORMAT function

In Release 3.2.1, the NUMFORMAT function supported a format string of ##.#0 in the Javascript running environment. This string was not supported in the Photon running environment. 

For Release 4.0 and later, this format string is no longer supported and must be changed.

NOTE: After you have upgraded to Release 4.0 or later, you must change references format strings with a # before 0 for the NUMFORMAT function to use a supported formatting string. See NUMFORMAT Function.

ARRAYUNIQUE function can now take a single column as input

In Release 3.2.1 and earlier, the ARRAYUNIQUE function required at least two functions to generate an output. 

Beginning in Release 4.0, this function can accept a single array or column as input, generating an output array containing only the unique values in the source. See ARRAYUNIQUE Function.

Execution Changes

Ternary predicates evaluating to null return false expressions

Suppose you have the following function expression:


In Release 3.2.1 and earlier, predicates that returned null values returned null for the entire expression. In this case, the expression returned a null value. 

In Release 4.0 and later, this expression returns 2.

See IF Function.

Null values no longer automatically filtered in limiting transforms

In Release 3.2.1 and earlier, when filtering the set of rows using a recipe step, such as a keep or delete transform, any null values in the evaluated in the condition would result in the filtering being applied. Example:

delete row:invAge >=90

If invAge contained a null value for a row, the row was deleted.

NOTE: In Release 4.0 and later, null values used as inputs to filtering transforms do not result in the row being filtered. This is a change in behavior for null values.

For each of the transforms below, you can review how to retain the Release 3.2.1 and earlier behavior in Release 4.0 and later.

Delete transform:

Release 3.2.1 example:

delete row:invAge >=90

Release 4.0 example:

delete row:(invAge >=90 && invAge == null())

Keep transform:

Release 3.2.1 example:

keep row:POS_Sales < 100

Release 4.0 example:

keep row:(POS_Sales < 100 && POS_Sales != null())

IF transform:

Release 3.2.1 example:

derive value:IF(rating > 9.0, 'ok','retry') as:'status'

Release 4.0 example:

derive value:(IF(rating > 9.0, 'ok','retry') && rating != null()) as:'status'

Tip: Release 4.0 introduces a series of conditional functions that can streamline computation and action. These functions test conditionals based on type (e.g. IFNULL) or based on computation of an aggregate function (e.g. SUMIF). See New conditional functions above.

 For more information:

Release 3.2.1

Syntax Changes

Terminology Changes

NOTE: Beginning in Release 3.2.1, values that are considered empty are now referred to as missing.

Function Changes

The following name changes have been applied to existing functions to use more familiar names.

Old Function NameNew Function NameNotes
ISEMPTYISMISSINGRemoval of duplicate function name.

Execution Changes

Transforms that nest null values in arrays now write null literals on Photon

In Release 3.2 and earlier, when a transform step was nested a null value within an array, an empty string value was written in the Photon running environment.

In Release 3.2.1 and later, the value written for a nested null value in the array is the literal: null.

Suppose your data looks like the following:



  • empty_str_col contains an empty string value.
  • null_col contains a null value

If you add the following recipe step:

nest col: text_col, empty_str_col, null_col into: array as: 'result'

In Release 3.1, the result was the following:

myText  ["myText", "", ""]

In Release 3.2.1, the result is the following:

myText  ["myText", "", null]

This change was made to align the behaviors of the Photon running environment with the JavaScript running environment.

Release 3.2

Syntax Changes

Transform Changes

  • multisplit transform has been deprecated. All multisplit capabilities are now supported by the split transform. See Split Transform.
  • pivot transform now supports multiple columns. See Pivot Transform.
  • unnest transform now requires the keys parameter, which was optional in previous releases. See Unnest Transform.
    • To unnest arrays without specifying keys, use the flatten transform. See Flatten Transform.
  • arraylength and arraystomap now accept functions that return arrays as inputs to the function.
  • domain and subdomain functions have been updated to reflect standard interpretations of domain and sub-domain values for URLs:

    ReleaseExample URLDomain FunctionSubdomain Function
    Release 3.1.2 and earlier
    Release 3.2 and later
  • The following parameter values are no longer supported as special capture groups in the with parameter for the replace transform. These references in the with parameter do not work in the Photon and Hadoop running environments and are unlikely to work at scale in any other running environment:


Parameter Changes

Applicable Transform(s)Old TermNew TermNotes
Extract Transformurlparam 

Removed from use in the extract transform. Use of this parameter in that transform prevented the use of other extract parameters. See Extract Transform.

Countpattern Transformquote Removed from use in the countpattern transform. Parameter was not being respected. See Countpattern Transform.
Extractkv Transformquote Remove from use in the extractkv transform. Parameter was not being used. See Extractkv Transform.
Extractlist Transformquote Parameter is now used exclusively for matching against delimiters. Parameter does not match against patterned values. See Extractlist Transform.
Split Transform and Extract Transformlimit 

For these transforms, the limit parameter can no longer be used in conjunction with the following parameters: positions, delimiters, at, and urlparam.

In previous releases, these combinations did not actually work, even though the transform step was consumed. Now, it generates an error.

Function Changes

More consistent results for DATEDIFF functions:

Prior to Release 3.2, the DATEDIFF function generated inconsistent results between the Pig and Javascript running environments for DayOfYear calculations.

Beginning in Release 3.2, the DATEDIFF function has been updated to generate more consistent results. See DATEDIF Function.  

Trifacta Pattern Changes

Changes to alpha-numeric pattern:

The alpha-numeric Trifacta pattern now applies to a single character and does not match on underscores (_). Previously, it was applied to one or more alpha-numeric characters, as well as underscores.

NOTE: Beginning in Release 3.2, the alpha-numeric Trifacta pattern applies to a single character. If you used it in your recipes prior to Release 3.2, these references have been converted to regular expressions to support matching with multiple characters.

New alphanum-underscore pattern:

To support previous functionality, you can use the new alphanum-underscore pattern, which matches on a single alpha-numeric character or underscore. 

For more information, see Text Matching.

RANGE function

The RANGE function now accepts negative start and stop values.

Deprecated Items

Aggregate Tool

In prior releases, you could build aggregation steps using a separate tool, which was available through the Transform Editor.

In Release 3.2, this tool has been replaced by building aggregate transforms in the Transform Builder. See Transform Builder.

Language Cheat Sheet

In prior releases, the Language Cheat Sheet was available through the User Profile menu. This menu option has been removed. Additional contextual documentation is available through the Transform Builder. See Transform Builder.

You can still access the Language Cheat Sheet by adding /docs to the base URL. For example:

NOTE: The above option is likely to be deprecated in a future release.

Your Rating: Results: PatheticBadOKGoodOutstanding! 10 rates

This page has no comments.