The Variables pane

View > Variables lists what the notebook's Python session holds on this PC: each variable's name, type, size and a short value. It's the same explorer Fabric and Databricks notebooks have. Reading it runs no cells and no queries; the only code of yours it can touch is a class's own repr, used for the short value, as any Python explorer does.

Filter and sort#

• Type in Filter by name or type to narrow the list, for example DataFrame or sales.
• Click Name, Type or Size to sort; click again to reverse. Size sorts by the memory a variable holds, biggest first; variables whose memory can't be told cheaply go last.
• The line under the table counts the variables and adds up their memory.

Memory is an estimate: a Polars or Arrow frame's buffers, a NumPy array's bytes, or a built-in list, dict or set with its first 10,000 items, scaled up (an object inside counts its base size). A Spark frame, a Polars LazyFrame and your own classes have no figure, because measuring them would mean running their query or their code. Hover a size to see the estimate.

Open a variable#

The arrow next to a data frame, list, tuple, set or dict opens it in place:

Values whose name or content looks like a secret (password, token, key and the like, or a value read with getSecret) show as hidden, at the top level and inside an opened dict, and a hidden variable can't be opened. Opened variables close when the pane refreshes after a run, and reopen with the new values when you open them again.

Clicking a variable's row still adds a cell that displays it.

When the debugging Preview features are on, the same table is the pane's Now tab, beside watches and history.