Overview

Weave is a visual data-processing canvas that combines source, transformation, and terminal nodes into a runnable pipeline. The canvas presents a node panel categorized by node type (Data category for source / terminal nodes, Data Preparation category for transformation nodes, plus additional categories for specialized transformations). Nodes are added by drag-and-drop, connected by drawing lines between them, and configured via per-node settings panels on the right side of the screen.

Weave pipelines are authored from Weave → Design → Weave. Each Weave is a named artifact that can be saved. Saved Weaves are listed under Libraries, where they can be run on a cluster, scheduled, shared, or deleted. A Weave’s data flow is from source nodes (Data Import) through transformation nodes (Cleansing, Imputation, Duplicates, and others) to terminal nodes (Result for design-time inspection, Output for persistent destination writes).

Note: “Weave” is the current name for this artifact and canvas. Earlier interface versions and documentation refer to the same feature as “Workflow.” The terms refer to the same thing; product documentation uses “Weave” exclusively going forward.

When to use it

• Building ETL pipelines that ingest data from one or more sources, transform it through a configured chain, and write the result to a destination.
• Combining data from multiple Datasources through join, merge, or union transformations into a single output stream.
• Preparing data — cleansing nulls, deduplicating, normalizing values — before downstream consumption by Reports (via Custom Scripts) or other Weaves.
• Scheduling recurring data refreshes that run automatically against production destinations.

Anatomy of a Weave

Canvas	The main work area where nodes are placed and connected. Each Weave has its own canvas.
Node panel	Left or top panel listing available node types organized by category — Data, Data Preparation, Transformation, Statistical Plots & ML Nodes.
Configuration panel	Right side of the screen. Shows the selected node’s configuration. Per-node Save / Discard discipline applies.
Toolbar	Top of the canvas. Includes Result, Save, canvas-level Run, Settings, Refresh and RPT Export actions.
Result	Shows the output of the last-run node or canvas-level run, with per-column filter and resize controls.
Per-node icon controls	Three icons appearing above a selected node: Run (per-node execution), Pencil (rename), Cross (delete). Available on every node type.

Data Import — four Data Source Types

The Data Import node is the canonical source node for a Weave. It supports four Data Source Type options, each pulling from a different kind of source.

Datasource	A configured Datasource (database table, API, or similar configured connection)	Standard ETL pipelines pulling from databases or API-backed sources.
Custom Script	A saved Custom Script from Lens	Reusing combined / aggregated / filtered output authored elsewhere — instead of redoing joins and aggregates inside the Weave, consume a pre-built Custom Script.
Connector	A configured Connector (file-based source — CSV, Excel, Other connectors)	File-based ingestion. Connector imports support additional configuration: dataset / file selection, file information, field selection, dataset filtering, row limit.
Data Weave	Another Weave’s Result node output	Chaining Weaves — composing one pipeline’s output as another’s input without persisting to a destination in between.

Node connection and configuration

Add a node	Drag from node panel onto the canvas.
Connect two nodes	Click the tip of the upstream node and drag the line to the downstream node.
Open a node’s configuration	Click the node. The configuration panel appears on the right.
Rename a node	Click the node, click the Pencil icon, enter the new name.
Delete a node	Click the node, click the Cross icon. Downstream connections need to be reconfigured.
Run a single node	Click the node, click the Run icon. Useful for testing during authoring.
Save or discard configuration	After making changes in a node’s right-panel configuration, click Save to commit or Discard to revert. Closing without choosing triggers a warning popup.

Run modes

Weaves support three distinct run modes — pick by what you need from the run.

Design Mode	Canvas-level Run button on the Weave canvas during authoring	Executing the connected nodes in dependency order during design-time. Result displays in the Result Tab. Authoring-time only — does not run on cluster compute.
Single Node	Per-node Run icon on individual nodes (specific nodes only — e.g., SQL Executor Node, Bulk-Data Transformation Node)	Testing a single node’s behavior in isolation. Only certain nodes support single-node Run independent of the rest of the Weave.
Run on Cluster	Run button under Actions on Libraries → Weave for a saved Weave	Production execution on Databricks cluster compute.

Result Tab

The Result Tab at the bottom of the canvas shows the output of the most recent run. Per-column controls allow filter and resize.

Filter (column header icon)	Opens a multi-select dropdown for that column. Use checkboxes or the search bar to pick values; only matching rows display. Filters are per-column and combine across columns. Use Clear to remove the column’s filter.
Resize (column-header divider drag)	Drag the vertical line between two columns to resize. Use to accommodate longer header text or wider data values.
Scope	Available on every node’s Result Tab — not restricted to Data Import.

Lifecycle actions on saved Weaves

Saved Weaves are managed from Libraries → Weave. Each Weave in the list exposes lifecycle actions.

Run	Triggers Run on Cluster for the Weave. Processing happens on Databricks.
Delete	Removes the Weave from the Libraries list. Confirmation popup required. A success popup confirms removal.
Schedule	Configures recurring automatic runs. Schedule popup accepts Title, schedule configuration, Category of Display, and optional description.
Share	Shares the Weave with selected users at a chosen role, group.

Key behaviors

Weave is the artifact and the canvas. “Weave” refers both to the named pipeline artifact (what gets saved and appears in Libraries) and to the visual canvas where it’s authored. The terminology change from “Workflow” to “Weave” applies to both senses uniformly.

Multiple Data Imports combine downstream. A Weave can have multiple Data Import nodes pulling from different sources, with their outputs combined through downstream join / merge / union transformations. The Data Source Type can be different per Data Import — one importing from a Datasource, another from a Connector, another from a Custom Script — all combined into a single transformation chain.

Multiple terminal nodes for fan-out. A Weave can have multiple terminal nodes — multiple Result nodes for inspection at different branch points, multiple Output nodes for fan-out to multiple destinations. They consume from the same upstream branch points and produce parallel outputs.

Save / Discard discipline applies node-wide. Every node’s configuration follows the same Save / Discard pattern with the warning popup on close-without-choice. This is uniform across all node types — Data Import, transformation nodes, Output, Result.