Calculation

Private Preview

This feature is currently in Private Preview.

This feature has the following limitations:

• Only available for Realtime Feature Views.
• Set of supported SQL functions is limited and is actively being expanded. See SQL Expression Reference below for details.
• Please reach out to Tecton Support for questions and feedback!

If you would like to participate in the preview, please file a support ticket.

Summary

The Calculation class describes a Calculation feature that is applied to a Realtime Feature View via the features param.

Example

from tecton import Calculation, RealtimeFeatureView

transaction_counts_fv = RealtimeFeatureView(
    name="transaction_counts_features",
    sources=[request_ds, user_transaction_metrics],
    # mode=None implied
    ...
    features = [
        Calculation(
            name="transaction_count_fill_na",
            expr="COALESCE(user_transaction_metrics.transaction_count, request_ds.default_count)",
        ),
        Calculation(
            name="transaction_count_add_1",
            expr="user_transaction_metrics.transaction_count + 1",
            description="a test calculation utilizing addition",
            tags={"tag": "value"}
        ),
    ],
)

Attributes

Name	Data Type	Description
description	Optional[str]	A human-readable description of the feature.
tags	Optional[Dict[str, str]]	Tags associated with the feature (key-value pairs of user-defined metadata).
name	str	The name of this feature. Must be explicitly defined.
expr	str	The calculation string expressing the operations and input operands.

__init__(...)

Parameters

• description: Optional[str] = None A human-readable description of the feature.
• tags: Optional[Dict[str, str]] = None Tags associated with the feature (key-value pairs of user-defined metadata).
• name: str = None The name of this feature. Must be explicitly defined.
• expr: str = None The calculation string expressing the operations and input operands.

SQL Expression Reference

The following SQL functions are currently supported in Calculation Features:

Arithmetic

Standard SQL arithmetic operations are supported.

amount + 1
amount - value
value * 2
3 / 2

info

Please note the following:

• For addition, subtraction, and multiplication: the result of arithmetic operation will be the type required to preserve the precision of the inputs. The precision hierarchy is Float64 -> Float32 -> Int64 -> Int32. For example, a Float32 value + Int64 value will return a value of type Float32 to preserve the float input's precision.
• For division: the result will always be a Float64. If division by zero occurs, the result will be type Float64:
  • +Infinity if the numerator is positive.
  • -Infinity if the numerator is negative.
  • NaN if the numerator is 0.
• Currently, Realtime Feature Views do not support Int32 features. As a workaround for your calculations, you can cast your Int32 value to Int64 using Int32_value + 0.

CASE Statements

CASE statements are supported.

CASE
    WHEN my_source.amount > 100 THEN 'high'
    WHEN my_source.amount > 50 THEN 'medium'
    ELSE 'low'
END

COALESCE

COALESCE(value, default_value)

Returns the first non-null value in a list of expressions.

COALESCE(amount, 0)

Comparison Operators

Standard SQL comparisons are supported.

• Comparisons between Float32, Int32, Float64, Int64, and Bool are supported. For example, you can compare True and 1 (True = 1 returns True).
• If one of the elements is a String element, it can only be compared against another String element. Strings are compared using lexicographical order.

amount > threshold
value >= 0
3 != 2

DATEDIFF

DATEDIFF(datepart, start_date, end_date)

Returns the count of the specified datepart boundaries crossed between two timestamps. All datetimes provided are converted to UTC and all datediffs are performed on the UTC equivalent.

DATEDIFF('minutes', start_time, end_time)

info

DATEDIFF logic follows DuckDB convention where the difference is equal to the number of partition boundaries between the two timestamps.

For example, consider DATEDIFF('days', my_source.start_date, my_source.end_date). If start_date is 2022-01-01 23:59:00 and end_date is 2022-01-02 00:00:00, the result will be 1.

Logical Operators

Standard SQL logical operators are supported.

amount > 100 AND amount < 200
value OR another_value
NOT value