Skip to main content

Columns Flow

Columns Flow

1. What Columns Flow Is​

Columns Flow is a visual, node-based workspace for building data workflows in Columns. Each node represents a dataset, and each edge represents a transformation from one node to the next. You can sync data into nodes, preview results, and save visuals from any node.

2. What Columns Flow Can Do​

  • Connect input nodes to existing data sources and sync them into the flow.
  • Transform data with generated compute steps such as transform, aggregate, pivot, post-process, and sort.
  • Preview sample inputs and outputs at each step before approving changes.
  • Update downstream nodes by approving a transformation on an edge.
  • Visualize any node once it has schema and data samples.

3. How You Use Columns Flow​

  1. Create or open a flow and start from the input node.
  2. Select a data source for the input node, then sync to load schema and data samples.
  3. Add a new node, connect it from the source node, and open the edge details.
  4. Describe your transformation intent and click Generate to produce a compute plan.
  5. Preview the plan and sample output, then click Approve to update the target node.
  6. Repeat to build multi-step workflows, and visualize any node when needed.

4. How to Reference Other Nodes​

In your intent, you can easily reference values from another node to filter your data, for example, you can say:

Filter the data to include only rows where Location is the top location from "Top Locations" node

This statement implies there is an existing node called "Top Locations", and it will fetch its top value of field "location" and use that to filter the upstream data you are transforming.

This gives you flexibility to split your workflow into multiple nodes, and dynamically compose them together.

tip

Give each node a meaningful name so that you can easily referenc them.