Shape / columns · All operations

Set a column's role

A variant of Change a column's type

Mark a column as a key, label, measure, dimension or semantic feature type without changing its values.

Status

In arcform

Not built yet

arcform does not do this yet. Its place on the roadmap

What other tools call it

Meridian's classification; an independent blind rating agreed on 95.7% of 324 rows (variant, version 3).

Alteryx Designer (1)
  • Feature Typesread 2026-09-27

    Detect or set the semantic feature type of each column

Pentaho Data Integration (2)
  • Annotate streamread 2026-09-27

    Mark stream fields as measures, attributes or linked dimensions for the Streamlined Data Refinery

  • Shared dimensionread 2026-09-27

    Define a dimension that models can share, for the Streamlined Data Refinery

Power Query M (3)
RapidMiner Studio (1)
  • Set Roleread 2026-09-27

    Assign a role (label, id, weight, prediction, regular) to columns

VisiData (5)
  • key-colread 2026-09-23

    Toggle current column as a key column (group-by, join key and x-axis role)

  • key-col-offread 2026-09-23

    Unset current column as a key column

  • key-col-onread 2026-09-23

    Set current column as a key column

  • key-off-selectedread 2026-09-23

    Unset selected source columns as key columns

  • key-selectedread 2026-09-23

    Set selected source columns as key columns

Yggdrasil Decision Forests (2)
  • Feature / Columnread 2026-09-27

    Declare one input column: its semantic, vocabulary limits, discretisation bins and monotonic constraint

  • Semanticread 2026-09-27

    Enum of 7 column meanings: NUMERICAL, CATEGORICAL, HASH, CATEGORICAL_SET, BOOLEAN, DISCRETIZED_NUMERICAL …

Background

In RapidMiner, the Set Role operator changes the role of an attribute, and that role describes how other operators handle the attribute.[1] In RapidMiner, the default role is regular, and regular attributes are used as input variables for learning tasks.[1] In Yggdrasil Decision Forests, the interpretation of a feature's values, such as a numerical quantity, a category or a set of tags, is called its semantic.[2] In Yggdrasil Decision Forests, a feature represented by 64-bit integers might have a numerical semantic or a categorical semantic.[2]

When to use it

RapidMiner's label role marks the target attribute for learning operators, which is also often called the target variable or class.[1] An attribute with RapidMiner's id role identifies the examples and should be unique for all of them.[1] RapidMiner's blending operators, such as Join, Union, Transpose and Pivot, use the id attribute to perform their tasks.[1] Some RapidMiner operators expect special roles to be unique and might not know how to handle a special role given to more than one attribute.[1]

Using the wrong semantic for a feature harms both the training speed and the quality of a Yggdrasil Decision Forests model.[2] Yggdrasil Decision Forests automatically recognises integer and floating-point values as numerical features and string features as categorical.[2]

The method

Power Query's Table.AddKey adds a key to a table from a list of the column names that define it, and its isPrimary argument says whether the key is primary.[3] DuckDB's COMMENT ON statement adds metadata to catalog entries such as tables and columns, following the PostgreSQL syntax.[4] A comment set with COMMENT ON COLUMN in DuckDB is read back from the comment column of duckdb_columns().[4] A role written into a free-text comment is only as good as the convention that reads it back, so agree its wording before a later step relies on it. (Meridian's view)

Alteryx's Feature Types tool autodetects the feature type of each column, and any of them can be changed by hand.[5] Yggdrasil Decision Forests determines feature semantics automatically when it trains a model, and a semantic can be overridden by naming it for a feature.[2] RapidMiner's roles also include prediction, which applying a model adds, and cluster, which operators such as k-Means add.[1]

Compared with the alternatives

Renaming an attribute in RapidMiner changes only its name and not its role.[1] In Alteryx, feature type and data type are not the same: feature types serve mainly the machine learning tools of the Intelligence Suite, while data types are used throughout Designer.[5] In Yggdrasil Decision Forests, a feature's semantic is related to, but different from, its representation, the technical data type of the feature.[2]

Sources

  1. [1]Set Role - Rapidminer Documentationdocumentation
  2. [2]How to define model features - YDF documentationdocumentation
  3. [3]Table.AddKey - PowerQuery M | Microsoft Learndocumentation
  4. [4]COMMENT ON Statement – DuckDBdocumentation
  5. [5]Feature Typesdocumentation