Added cards inside PlotModule

[ESCRI11]

Description

Provide the option of using cards inside a PlotModule (Figure 1).

Features

If we place two plots inside the same PlotModule I expect them to be simiar in nature, therefore, the information button will display the same no matter which plot is active. The download button will download the active plot and the options button will display the options for the selected plot.

When displaying the modal output, cards are still inside (Figure 2). This could be changed so that the modal shows the selected card only.

Usage

On the PlotModuleUI we have to use the arguments cards and card_names:

• card: Bool value. If TRUE, cards will be used.
• card_names: Name of the cards (on Figure 1 would be c("card1", "card2")).

UI example code Figure1:

PlotModuleUI(
    ns("pltmod"),
    plotlib = "plotly",
    info.text = info_text,
    download.fmt = c("png", "pdf", "csv"),
    title = "Dataset explorer",
    cards = TRUE,
    card_names = c("card1", "card2")
  )

On the PlotModuleServer we have to create as many servers as cards. In this example, we have two plots, so two servers:

• ID is shared between the UI declaration and all the servers
• We just specify which servers correspond to which card numerically

PlotModuleServer(
      "pltmod",
      plotlib = "plotly",
      func = plot.RENDER,
      func2 = modal_plot.RENDER,
      card = 1
  )
PlotModuleServer(
      "pltmod",
      plotlib = "plotly",
      func = plot.RENDER2,
      func2 = modal_plot.RENDER2,
      card = 2
  )

Additional information

I am working on a developers guide for the PlotModule explaining all the different options it has and how to use them. The provided information in this PR will be included on the wiki.

---

Figure 1

Figure 2

[ivokwee]

should we merge now or after Mauro's grid stuff?

[ivokwee]

I like the solution with multiple server, one ui. Once I thought to make a seperate MultiServer that takes multple plotmoduleservers but it seems not necessary. Your solution is enough. I would like to try this for the heatmap with the new plotly based and old heatmap.

[ESCRI11]

Yes, the heatmap would be a good candidate to try this out.

I was planning on waiting for Mauro grid stuff, but we can push this now too.

[mauromiguelm]

I like this a lot, going to test it today! There are many interesting applications that comes to mind.

I think we can merge first Dev grid cards " #256 into master, and then merge this PR after that.

Or we can merge this PR into Dev grid cards " #256, and later to master.

Not sure what is more efficient, but we should do this asap before creating more conflicts.

[mauromiguelm]

The PR is for the backend tab option, right? I couldnt see any tab in loading module tsne plot.. or maybe its implemented and its just me that cannot see?

An interesting development would be to place the tab UI directly in the card header, see this example: https://rstudio.github.io/bslib/articles/cards.html#multiple-tabs

[ncullen93]

Is this PR still relevant? If it is, let's take care of it. Otherwise, let's close :) @ESCRI11

[ivokwee]

Best is to be resubmit I guess. File names and locations have changed since the playbase merging.

[ESCRI11]

To be tested carefully (also by me)

[ESCRI11]

This PR will close #296

[mauromiguelm]

I really like the structure behind this, I have some minor comments to try to improve the current approach:

• it would be nice to pass the PlotModuleServer "func" as an argument vector of multiple elements, where each element map to card_names in UI. That way we dont have to use PlotModuleServer twice.
• Vertical distance in cards is reduced for showing card names, it would be nice to have a more minimal UI.
• Sometimes option changes in Dynamic are not reflected in Static, for example the row clustering shows genes only in dynamic, but it varies in the static, with static having multiple labels. I think plots should show consistant information, and this might be a legacy issue unrelated to your PR. I guess an easy solution we could hide static row dendogram in case the clustering is the same.
• Plots do not get to the bottom of card, this might have something to do with the fill behavior of the parent div, we can discuss a bit more if you wish, I think a solution is easy.

[ivokwee]

1. I am fine with calling two plotmodules. I see no real advantage putting the extra func3 func4 inside.
2. Vertical space. Would be nice to have the two subtabs either in the header, or type='hidden' and have the navigation buttons in the header or options (as in v2)
3. Both plot should react on the same inputs of the plot options. I think that should be possible.
4. Fill behaviour should be restores. I think it needs height to be set to 100%.

[ESCRI11]

1. OK will leave it the structure as it is.
2. I work on something like that, not 100% sure if it will work out.
3. As long as the functions that generate the plots make proper use of the options, this is no issue, it has nothing to do with the PlotModule, although for this particular plot I can include the fix on the PR.
4. Got it.

[ESCRI11]

Ok point 2 can be achieved easily, I have a working prototype (ugly but functional)

[ivokwee]

Ok point 2 can be achieved easily, I have a working prototype (ugly but functional)

Can you show? My idea was to use bslib::navs_hidden (see https://rstudio.github.io/bslib/reference/navs.html) and add somehow an actionButton in the drowdownmenu (or header?) that does something like updateTabPanel("card_selector", selected = "card1")

[ESCRI11]

@ivokwee Check the latest commit 8d1b84f. Basically I grab the "primitives": navList and content and place them on the header and body sepparately. I just have to make them inline and not affect the buttons.

[ivokwee]

@ivokwee Check the latest commit 8d1b84f. Basically I grab the "primitives": navList and content and place them on the header and body sepparately. I just have to make them inline and not affect the buttons.

Seems to work but the header fill is all over the place now. I guess the fillrow is not set properly if you have a varying number of elements or so. If not, you can rewrite fillrow with non-flexbox layout or add empty elements.

[ESCRI11]

Yep exactly, I am trying to figure out the best solution

[ESCRI11]

@ivokwee @mauromiguelm Please feel free to try the latest developments. Only thing missing is points 3 and 4, which should be easy.

[ESCRI11]

Only missing fixing the heatmap options so that in both plots are relevant. Also found a small bug affecting this new development and the plot editor, now fixed.

[mauromiguelm]

I really like this PR, it looks beautiful Xavier! I tested some of the option and it works quite nice, is there anything broken still?

I did two things (not yet pushed):

• I reduced the font size of the tabs, do you like it?

• create a small loop in the server so we dont call plotmoduleserver twice.. I think its a big more easy to understand, especially if we decide to have multiple plots in the future!

Let me know if I should push these two updates!

[ESCRI11]

Great! Good eye with the smaller font size. Please push the changes. I think we are ready for merge

[mauromiguelm]

Done, I checked again and all looks good!

Problems related to the heatmaps:

• absolute versus relative coloring between the two heatmaps are quite off, we should match color and outcomes as much as possible between the two heatmaps..
• show legend is not working when we select split by phenotypes

This is kind of unrelated to this PR, we could merge and keep working these issues on #296, as we have other priorities

[ESCRI11]

Agreed. Let's finish this PR and solve heatmap problems as different issues