Support one bond defining another when setting both simultaneously

[fonsp]

See JuliaPluto/PlutoSliderServer.jl#38

[lungben]

Copied from JuliaPluto/PlutoSliderServer.jl#38 :

This currently works in Pluto, but not in PlutoSliderServer:

@bind x Slider(1:10)

@bind y Slider(1:x)

x, y

Why? When moving the y slider in normal Pluto, we just send the y value to the Pluto server, it sets it in the notebook process and runs the dependent cells (just cell 3). In PlutoSliderServer, moving y will send x and y together, this is our stateless idea.

Pluto does not handle this correctly (this is a bug in Pluto, not PSS): it first sets both x and y in the notebook process, and then runs the cells that depend on them: cells 2 and 3. The problem is: running cell 2 will set y to its initial value: 1, and you will always see x, 1 in the third cell.

What should be done?

If the SliderServer submits updated values for (x, y) then:

1. Update the Slider in @bind y Slider(1:x)
2. Do not set y as the value from the Slider, but keep y as the value in the request from SliderServer.

As far as I understand, a situation where a variable value (here y) is requested both from the SliderServer and result of a dependent cell in the notebook itself can only happen for @bind values (otherwise it would be an invalid multiple definition of a variable).
Thus, a fix could be to reset a variable back to its request value if the variable is set in a @bind.

@fonsp could you please give me a quick pointer which part of the Pluto code receives the update requests of the SliderServer?

[lungben]

As far as I can see from the code, the updates arrive in Pluto here: https://github.com/fonsp/Pluto.jl/blob/main/src/evaluation/RunBonds.jl#L1

The information which variables are in the request is in bound_sym_names.

This information needs to be populated downstream to run_reactive!(), i.e. https://github.com/fonsp/Pluto.jl/blob/main/src/evaluation/Run.jl#L129, where the value in the request from PlutoSliderServer should overwrite the new (default) value obtained from recalculation of the cell (if they are different).

Is this the right approach?

[fonsp]

Fixed by #2014