Chiming in on an old thread – I think the visual dataflow graph is immensely clever, and a big step in a good direction, but I don’t find myself using it all that much.
I suspect this is for a few reasons –
First, the minimap is trying to strike a balance between showing local dependency structure and maintaining a view of all the cells. I agree with j-pb in that I don’t think this is necessary – typically I either want to see either the local dependencies, or the entire graph for debugging.
Second, there is no context provided with the minimap — just the fact that a relationship exists. I think that showing a little summary of how a cell is used — maybe by just grabbing the line (truncated or not) where the cell is referenced — would help remind the notebook author about the relationship between the two cells, without jumping around the notebook.
Third, in agreement with tau, is that the minimap is small and off to the side, rather than local to each cell. If feels a little jarring to be editing a cell, move the cursor over to the minimap to see relationships, click to jump to another cell, make another edit, go back to the map, jump again, etc. I think that a local minimap should be as tightly integrated with each cell as possible.
My basic proposal would be something along these lines – having a minimap button, which expand a local dependency graph under the currently selected cell. This dependency graph would show some contextual information about how the current cell is being used.
This approach could be further refined — for example, maybe clicking a button to display the graph isn’t ideal, and instead you could have some subtle local minimap always visible with each cell. Maybe you don’t have a graph at all, and just a list of Dependencies and References, each with some transcluded context showing how the current cell is used.
Anyways, many thanks to the Observable team for your hard work – these notebooks are quickly becoming my first choice for an environment to think in.
