Allow nested manipulates

[djsegal]

I want to be able to nest manipulates where inner ones are affected by outer ones.

This currently causes the jupyter notebook to stall out

---

Simple example:

some_dict = Dict(
  "arc" => 3.3, 
  "act1" => 6.8, 
  "act2" => 9.6
)

@manipulate for design=["arc", "act1", "act2"]
    @manipulate for r=linspace(0, some_dict[design], 25)
         cur_design = load_design(design)
         foo(cur_design, r)
    end
end;

[JobJob]

Yep, @manipulate has limitations. If you want to build more complicated UIs you're better off handling the widgets and signals manually. E.g. for the above:

some_dict = Dict(
  "arc" => 3.3, 
  "act1" => 6.8, 
  "act2" => 9.6
)

design_chooser = togglebuttons(["arc", "act1", "act2"]; label="design")
range_sig = map(design_val->linspace(0, some_dict[design_val], 25), signal(design_chooser))
r_slider = slider(value(range_sig); label="r")
set!(r_slider, :range, range_sig) # update the slider's range when the range_sig updates
design_updater = map(signal(design_chooser), signal(r_slider)) do design, r
    cur_design = load_design(design)
    foo(cur_design, r)
end
display.([design_chooser, r_slider, design_updater]);

[JobJob]

The stalling was caused by a couple of bugs arising from recent changes in Reactive to get it work on 0.6. The PRs to fix those, and another issue your code brought up for me, (JuliaGizmos/Reactive.jl#138 and JuliaGizmos/Reactive.jl#139) should be merged soon. In the meantime I'd recommend trying Pkg.checkout("Reactive", "job/flatten-bind-fix") it should be much more reliable. Unfortunately it won't fix the OP, but you'll see that it almost works, just a new slider for r gets created every time you choose a new design.

This is strongly related to #12 and it would be nice to fix this. I'm hopeful I'll get a chance to as part of my GSoC work on Interact

Anyway, thanks for the report, it helped me track down a couple of subtle bugs in Reactive.

[JobJob]

Those PRs are merged. Party time all round 🎉

[djsegal]

@JobJob just confirming this is ok to close?

[djsegal]

If you click an outer scope option, the inner scope options get re-appended

i.e.

[djsegal]

@JobJob, are you able to reproduce this?

[JobJob]

Yep. Leave this issue open so it stays on the radar

[djsegal]

I'm interested in helping fix this issue. Do you know where abouts to look?

---

edit:

the simplest way would probably be to:

• add html ids to the jupyter-widgets divs
• use javascript to enforce unique ids within a cell

i.e. a toggle over "temperature" would have the id #jupyter-toggle-temperature or something

[JobJob]

Yeah, it's gonna be pretty fiddly, and IJulia specific. Probably better to try implement it in InteractNext once it's released, which should be next week, though expect it to be a bit brittle at first.

A basic approach that might work is to get the widgets to become Signal{Widget}s, and then Interact can handle them updating.

As I wrote above, if you want to build more complicated UIs, you're probably better off handling the widgets and signals manually for now, rather than using manipulate.

[jasongrout]

FYI, supporting interacts inside interacts works well (at least in python) in ipywidgets 7.

[piever]

I just tried:

using InteractBulma
some_dict = Dict(
  "arc" => 3.3, 
  "act1" => 6.8, 
  "act2" => 9.6
)

@manipulate for design=["arc", "act1", "act2"]
    @manipulate for r=linspace(0, some_dict[design], 25)
         r
    end
end

and it seems to just work. Let me know if there are more things to be done on this front or whether the issue can simply be closed.

[djsegal]

You can close it

…

On Jul 11, 2018, at 11:36 AM, Pietro Vertechi ***@***.***> wrote: I just tried: using InteractBulma some_dict = Dict( "arc" => 3.3, "act1" => 6.8, "act2" => 9.6 ) @manipulate for design=["arc", "act1", "act2"] @manipulate for r=linspace(0, some_dict[design], 25) r end end and it seems to just work. Let me know if there are more things to be done on this front or whether the issue can simply be closed. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub, or mute the thread.