Overview

The Data Transformation node is a transformation node in a Weave workflow that applies column-level actions to the upstream data. Each upstream column expands into one or more action rows; each action row picks an action (Replace, Prepend, Pattern, Deidentification, Rename, Tag, Create New, Change Format, Datatype Conversion, Expression, Append) and supplies action-specific parameters. Multiple actions can be stacked on the same column.

Data Transformation is structurally distinct from the Data Preparation nodes (Cleansing, Imputation, Filter, Sorting, etc.) — those nodes apply a single operation type each across selected columns; Data Transformation orchestrates a heterogeneous mix of actions per column, with one node supporting all eleven action types simultaneously.

Note: “Weave” is the current name for what earlier interface versions called “Workflow.” This reference uses “Weave.”

When to use it

• Modifying column values across rows — Replace specific values, Prepend or Append tokens, apply Pattern-based substitution, evaluate Expression formulas.
• Renaming column headers — useful when downstream consumers expect specific naming conventions or when source columns have technical names that need user-friendly labels.
• Converting column data types — decimal to integer, string to date, etc. — to match downstream node or destination requirements.
• Creating computed columns via Expression with arithmetic on upstream values.
• Copying columns (Create New) to preserve originals alongside transformed versions, often paired with Deidentification.
• Tagging related columns for grouped operations, particularly for address-field deidentification.

Action catalog

Eleven actions are available in the action dropdown. Each row in this table summarizes the action, its parameters, and its primary use case.

Deidentification — generator sub-types

Twenty generator sub-types are available in the Deidentification action’s second dropdown. Each generator targets a specific kind of column.

Deidentification — methods

Three methods are available in the Deidentification action’s third dropdown. The method controls the determinism of the generator.

Multi-action stacking on a column

A single column can have multiple action rows. Click the + Action icon on an existing row to add another row under the same column.

Key behaviors

Per-column action orchestration. Data Transformation is unique among Weave’s transformation nodes: it doesn’t apply a single operation type uniformly. Each upstream column has its own independent configuration; each column can have its own action (or stacked actions). This makes Data Transformation the most flexible value-modification node — and the most complex to configure, since each column needs its own decision.

Multi-action stacking on a single column. The + Action icon allows multiple action rows per column. Stacking is unprecedented across other Weave nodes — every other transformation node applies a single operation per node. Stacking enables compositions like “deidentify then copy” or “convert type then apply expression” within a single Data Transformation node, avoiding the chain of multiple specialized nodes.

Deidentification is the article’s deepest action. With twenty generator sub-types covering most common PII shapes (names, addresses, dates, phone, email, numeric, decimal, alphanumeric) and three methods (Random for unpredictable masking, Consistent for joinable masking, Pass Through for staging), Deidentification on its own is comparable in complexity to several other entire Data Preparation nodes. The Method choice is particularly consequential — Consistent preserves source-value relationships through the masking, enabling downstream joins and aggregations to remain meaningful.

Rename changes column names; downstream references must update. Renaming a column to a new name means downstream nodes (Filter, Sorting, Output, Result, other Data Transformations) that reference the column by its original name will break. Either: rename consistently across the pipeline; place the Rename late in the pipeline so upstream references aren’t affected; or update downstream node configurations after renaming.

Apply and Save are distinct commits. Apply commits the Data Transformation popup state and closes the popup. Save commits the node configuration. Both are required for the configuration to persist — clicking only Apply leaves the changes in an uncommitted state at the node level; closing the configuration without Save triggers the warning popup.