What pain point is this feature intended to address? Please describe.
Vizier's autoscroll feels... not quite intuitive. This is hard to pin down, but simply put, there are situations where Vizier should scroll something into view, but doesn't. A few specific examples:
When a cell is running, cell outputs don't automatically scroll into view.
Note 1: We don't want cell messages to always scroll into view... if we're looking at a different part of the workflow, it can be annoying if the notebook keeps scrolling cells into view. One thought for a heuristic might be:
Is the bottom of the cell in view prior to the insertion of the new message?
Has the user scrolled in the extremely recent past (e.g., 100-200ms)
Note 2: The real pain point here is when I'm editing the last cell in a workflow. Re-running the cell clears the cell messages, scrolling the window so that the output is exactly hidden.
When scrolling to a cell (e.g., clicking in the ToC or clicking in the artifact list), the scroll jumps to the cell midpoint. This works for small cells (e.g., cells with height <= display height), but once the cell gets bigger (e.g., as a result of lots of outputs) the midpoint is useless. If the cell is bigger than the display, here's some thoughts on heuristics
Scrolling to a specific artifact should probably display the artifact itself
Scrolling to the cell from the ToC should align the top of the cell with the top of the display. If the code block is bigger than the screen, maybe we align the bottom of the code block with the bottom of the display?
What pain point is this feature intended to address? Please describe. Vizier's autoscroll feels... not quite intuitive. This is hard to pin down, but simply put, there are situations where Vizier should scroll something into view, but doesn't. A few specific examples:
height <= display height), but once the cell gets bigger (e.g., as a result of lots of outputs) the midpoint is useless. If the cell is bigger than the display, here's some thoughts on heuristics