# Transform/Propagate

While the Unit abstraction may feel insufficient to express complex logic or limiting, we provide two mechanisms for arbitrary Python code execution that enable quick iteration without giving up the benefits of the Schema type enforcement.

# Map/Transform

MapUnit allows you to apply an arbitrary function to the output of a Unit.

from verdict.common.judge import JudgeUnit
from verdict.transform import MapUnit

JudgeUnit() \
    >> MapUnit(lambda judge: Schema.of(score=((judge.score / 5) * 7))) # normalize to 1-7 scale

It has native support for multiple dependencies, which are indexed first-come first-serve at definition-time.

Reduce

Index

from verdict import Layer

Layer(JudgeUnit(), 5) \
>> MapUnit(lambda judges: Schema.of(score=min(judge.score for judge in judges)))

Layer([
    JudgeUnit(DiscreteScale((1, 5), end_is_worst=False)),
    JudgeUnit(DiscreteScale((1, 5), end_is_worst=True))]) \
>> MapUnit(lambda judges: Schema.of(score=(judges[0].score + (6 - judges[1].score)) / 2))

Additionally, we provide the following built-in MapUnits for convenience.

MapUnit	Description	Output Schema
MeanPoolUnit	Compute the mean of a list of values.	`Schema.of(field_name=field_type)`
MaxPoolUnit	Compute the mode of a list of values.	`Schema.of(field_name=field_type`
MeanVariancePoolUnit	Compute the mean and variance of a list of values.	`Schema.of(mean=float, variance=float)`

# Propagate

You can specify special logic to modify the OutputSchema of a Unit post-execution by using the .propagate() directive. This is particularly useful for propagating forward the OutputSchema of a dependency. Refer to the Previous section for more details.

Propagate Dependency

MapUnit Example

from verdict.schema import Schema

ConversationalUnit("Proponent").propagate(
    lambda unit, previous, input, output: Schema.of(
        thought=previous.thought,
        conversation=output.conversation
    ))

from verdict.common.judge import JudgeUnit

JudgeUnit().propagate(
    lambda unit, previous, input, output: Schema.of(
        score=(previous.score / 5) * 7
    )) \
>> ...