is the domainspecific language used to build transformation recipes in .
A recipe is a sequence of transformation steps applied to your dataset in order to produce your results.
Transform and transformation:
A transformation is a userfacing action that you can apply to your dataset through the Transformer page. A transformation is typically a usespecific or more sophisticated manifestation of a transform.
Tip: Except for the reference documentation for individual transforms, the language documentation references transformations that you can apply through the Transformer page. 
For most of these actions, you can pass one or more parameters to define the context (columns, rows, or conditions).
Function:
When you select suggestions in the Transformer Page, your selection is converted into a transformation that you can add to your recipe.
Tip: Where possible, you should make selections in the data grid to build transformation steps. These selections prompt a series of cards to be displayed. You can select different cards to specify a basic transformation for your selected data, choose a variant of that transformation, and then modify the underlying recipe as necessary. For more information, see Overview of Predictive Transformation. For more information on the suggestion cards, see Selection Details Panel. Some complex transformations, such as joins and unions, must be created through dedicated screens. See Transformer Page. 
NOTE: is not SQL. It is a proprietary language of data transformation, purposebuilt for . 
While there are some overlaps between and SQL, here are the key distinctions:
transforms follow this general syntax:
(transform) param1:(expression) param2:(expression) 
Transform Element  Description  

transform  In , a transform (or verb) is a single keyword that identifies the type of change you are applying to your dataset. A transform is always the first keyword in a recipe step. Details are below. The other elements in each step are contextual parameters for the transform. Some transforms do not require parameters.  
parameter1: , parameter2:  Additional parameters may be optional or required for any transform.

Depending on the transform, one or more of value
, col
, and row
parameters may be used. For example, the set
transform can use all three or just value
and col
.
Transform Element  Description 

value:  When present, the An expression can contain combinations of the following:

col:  When present, the Some transforms may support multiple columns as a list, as a range of columns (e.g., 
row:  When present, the row parameter defines the expression to evaluate to determine the rows on which to perform the transform. If the row expression evaluates to true for a row, the transform is performed on the row. 
group:  For aggregating transforms, such as 
order:  For aggregating transforms, such as window , pivot , and derive , the order parameter can be used to specify the column by which the transform results are sorted. In the previous example, you might choose to sort your sum of sales calculation by state: order:State . 
The following types of parameter inputs may be referenced in a transform's parameters.
Other can be referenced as column references. For literal values of these data types, you can insert them into your expressions as strings. Transforms cause the resulting values to be reinferred for their data type.
A reference to the values stored in a column in your dataset. Columns can be referenced by the plaintext value for the column name.
Example: value
parameter references the myCol
column.
Column names with spaces or special characters in a transformation must be wrapped by curly braces.
Example: Below, srcColumn
is renamed to src Column
, which requires no braces because the new name is captured as a string literal:
NOTE: Current column names that have a space in them must be bracketed in curly braces. The above column name reference is the following: 
Some parameters accept functions as inputs. Where values or formulas are calculated, you can reference one of the dozens of functions available in .
Example:
supports the use of variable references to aspects of the source data or dataset. In the following example, the ABS function is applied to each column in a set of them using the $col
reference.
$col
returns the value of the current row. For more information on these variables, see Source Metadata References.
Individual parameters within a function can be computed expressions themselves. These nested expressions can be calculated using constants, other functions, and column references.
Example: Computes a column whose only value is ten divided by three, rounded to the nearest integer (3):
A valid integer value within the accepted range of values for the Integer datatype. For more information, see Supported Data Types.
Example: Generates a column called, my13
which is the sum of the Integer values 5
and 8
:
A valid floating point value within the accepted range of values for the Decimal datatype. For more information, see Supported Data Types.
Example: Generates a column of values that computes the approximate circumference of the values in the diameter
column:
A true
or false
value.
Example: If the value in the order
column is more than 1,000,000, then the value in the bigOrder
column is true
.
A string literal value is the baseline datatype. String literals must be enclosed in single quotes.
Example: Creates a column called, StringCol
containing the value myString
.
supports a special syntax, which simplifies the generation of matching patterns for string values.
Patterns must be enclosed in accent marks ( `MyPattern`
). For more information, see Text Matching.
Example: Extracts up to 10 values from the MyData
column that match the basic pattern for social security numbers (XXXXXXXXX
):
Regular expressions are a common standard for defining matching patterns. Regex is a very powerful tool but can be easily misconfigured.
Regular expressions must be enclosed in slashes ( /MyPattern/
).
Example: Deletes all twodigit numbers from the qty
column:
A valid date or time value that matches the requirements of the Datetime datatype. See Supported Data Types.
Datetime values can be formatted with specific formatting strings. See DATEFORMAT Function.
Example: Generates a new column containing the values from the myDate
column reformatted in yyyymmdd
format:
A valid array of values matching the Array data type.
Example:
[0,1,2,3,4,5,6,7,8] 
See Supported Data Types.
Example: Generates a column with the number of elements in the listed array (7
):
A valid set of values matching the Object data type.
Example:
{"brand":"Subaru","model":"Impreza","color","green"} 
See Supported Data Types.
Example: Generates separate columns for each of the specified keys in the object ( brand
, model
, color
), containing the corresponding value for each row:
This preview is generated by applying the transformation to the sample in the data grid.
NOTE: The generated output applies only to the values displayed in the data grid. The function is applied across the entire dataset only during job execution. 
A transformation is an action for which you can browse or search through the Transform Builder in the Transformer page. When specified and added to the recipe, these sometimes complex actions are rendered in the recipe as steps using the underlying transforms of the language.
Tip: Through transformations, guides you through creation of more sophisticated steps that would be difficult to create in raw . 
For more information on the list of available transformations, see Transformation Reference.
For more information on creating transformation steps in the Transformer page, see Transform Builder.
A function is an action that is applied to a set of values as part of a transform step. Functions can apply to the values in a transform for specific data types, such as strings, or to types of transforms, such as aggregate and window function categories. A function cannot be applied to data without a transform.
Function Category  Description 

These functions are used to perform aggregation calculations on your data, such as sum, mean, and standard deviation.  
Comparison Functions  Comparison functions enable evaluation between two data elements, which are typically nested (Object or Array) elements. 
Math Functions  Perform computations on your data using a variety of math functions and numeric operators. 
Trigonometry Functions  Calculate standard trigonometry functions as well as arc versions of them. 
Date Functions  Use these functions to extract data from or perform operations on objects of Datetime data type. 
String Functions  Manipulate strings, including finding substrings within a string. 
Nested Functions  These functions are designed specifically to assist in wrangling nested data, such as Objects, Arrays, or JSON elements. 
Type Functions  Use the Type functions to identify valid, missing, mismatched, and null values. 
Window Functions  The Window functions enable you to perform calculations on relative windows of data within your dataset. 
Other Functions  Miscellaneous functions that do not fit into the other categories 
An operator is a single character that represents an arithmetic function. For example, the Plus sign (+
) represents the add function.
Operator Category  Description 

Logical Operators  and, or, and not operators 
Numeric Operators  Add, subtract, multiply, and divide 
Comparison Operators  Compare two values with greater than, equals, not equals, and less than operators 
Ternary Operators  Use ternary operators to create if/then/else logic in your transforms. 
A transform, or verb, is an action applied to rows or columns of your data. Transforms are the essential set of changes that you can apply to your dataset.
Transforms are described in the Language Appendices. For more information, see Transforms.
Documentation for is also available through . Select Help menu > Documentation.
Tip: When searching for examples of functions, try using the following form for your search terms within the Product Documentation site:
