Skip to content

Provenance ​

Curio keeps two histories while you work. The dataflow's version history records the whole graph each time a node or a data connection is added or removed, and for each person's typing in a node, each agent's change, each File > Save dataflow and each version put back. Each node keeps a record of its runs and edits. Both are saved with the project, so they are still there the next time you open it.

The version history ​

Choose View > Provenance to open the dataflow's history in a drawer on the right. Each row is one version, numbered from the first at the top, with what made it, such as Node added: Data Loading or Connection added, and the time it was recorded. The drawer opens on the version that is on the canvas, marked with a filled dot.

Click a version to open it: a thumbnail of the dataflow at that step, the date and time it was recorded, and how many nodes and connections it holds. Opening a version only shows it; the canvas stays as it is. The pin at the left of the drawer's title keeps the drawer open while you work on the canvas and in the top bar, and Escape leaves it open.

A dataflow opened without a saved history, such as a loaded file or an example opened for the first time, starts one with a version for each of its nodes and connections.

Going back to an earlier version ​

Open a version and click Put this version on the canvas. The rest of the history stays, so you can go forward again the same way. Going back is an edit like any other: it is a version of its own, and it changes the canvas of everyone who has the dataflow open.

If you change the dataflow from there, the new versions branch off from the one you went back to. The line down the left leads to the newest version, and the versions you left behind fold under the one they branched from, as 4 versions on another branch. Click the fold to show them, and again to hide them.

The Dataflow provenance drawer with four versions on another branch
Version 8 was put back after version 12 and changed again, so versions 9 to 12 sit on another branch, shown open under version 8, and the line goes on to version 13, the one on the canvas.

A node's run history ​

The Provenance button among a node's tabs, in its header, lists the node's runs in the same drawer. Each time the node runs, it adds a row with the run's number, what went in and what came out, such as In: DATAFRAME | Out: GEODATAFRAME (or ERROR for a run that failed), and the time it started. In the notebook view, a cell's Provenance button opens the same drawer.

Click a run to open it: when it started, how long it ran, and the code or specification that ran. Put this code back in the node puts that code in the node's editor without running it; if someone changed the node's code meanwhile, nothing changes, and Curio says who did. The next run then branches from the run you picked, so you can try one idea, go back, and try another without losing either. Runs left behind fold under the run they branched from, as versions do.

The Node provenance drawer with three runs of a Python Computation node
The runs of a Python Computation node: run 1's code was put back after run 2, so run 3 branches from run 1 and run 2 sits on another branch, shown open under run 1. Run 3 is open, with when it started, its code and the button that puts that code back in the node.

Your typing in a node is an entry too, named after you, and so is a change an agent makes to the node's code, named after the agent, such as Node Builder; a node an agent added with its code starts with one. Open an entry to see when it was made and the code it left, and put that code back as you would a run's. The Node Builder and the Dataflow Builder read this history, so you can ask either what changed in a node since a run.

What is kept ​

Curio keeps a saved dataflow's histories as each change is made, the same for its owner and the editors the owner invited; someone who only watches the dataflow sees none. Only the owner sees Clear history before this version, in an open version, which removes the versions before it. A file from File > Save dataflow as holds only the dataflow itself, not its history, so a dataflow loaded from a file starts a new one (see Projects and files).