[Feature request] Support Multi-Level Column Headers

[bdietermann]

Description

Currently, the CSV table view supports only a single row of column headers. It would be useful to support multi-level (hierarchical) headers for datasets that have grouped or nested column structures.

Use Case
Some datasets use structured or hierarchical column names, for example:

• Grouped metrics (e.g. “Revenue” → “Q1”, “Q2”, “Q3”, “Q4”)
• Nested data fields (e.g. “User” → “Name”, “Email”, “ID”)

In such cases, a single header row is insufficient to represent the structure clearly.

Proposed Behavior

• Allow defining multiple header rows in the CSV
• Display headers as a hierarchical structure in the table view.
• Optional, but nice to have: Visually group child columns under parent headers (e.g. spanning header cells).

[Robin-Reiche]

Thanks for this, it is a genuinely thoughtful request 🤔

I looked into it properly and I am going to hold off for now. Here is the honest reasoning.

CSV as a format has no concept of multi-level headers. The "multiple header rows" idea only works as a convention (for example pandas stacks header rows and the reader is told header=[0,1]), so there is no reliable way to detect it from a plain .csv on its own. Supporting it well would need a setting for the number of header rows and a flattening rule for export. It would also touch the core assumption of a single header row that editing, deleting, freezing and find all rely on, which is a lot of complexity and risk for something that mostly comes up in files exported from real spreadsheets.

Where I think it belongs: I do want to grow the extension to other file types like Excel later on, where hierarchical headers are native. That feels like the right moment to do multi-level headers properly, so I am keeping this open as a roadmap item rather than closing it.

If you have a concrete CSV example you are working with, feel free to attach it. It will help when I get to this 🙏

[bdietermann]

That makes sense. My use case for the requested feature is easily navigating timeseries CSVs with multilevel headers which I use for my simulations. For this purpose, freezing the additional header rows would be sufficient. As files may also contain headers consisting of more than 2 rows, adding an option to freeze more than one row would be great. -> I created another feature request for this.
I think this workaround requires way less effort to implement compared to a real multi-level header and also eases comparing rows against each other.

[Robin-Reiche]

Quick update tied to this: v1.10.0 ships freeze-multiple-rows (see #9), which covers the practical side of this for now. You can pin all of the header rows so they stay visible while you scroll a timeseries, without the CSV needing real multi-level headers.

I am keeping this issue open as the longer-term roadmap item for true hierarchical headers (spanning header cells and clean export), which fits better when the extension grows to formats like Excel where that is native. Thanks again for the thoughtful back-and-forth 🙏