Testing 1 2 3

[mtennekes]

Hi all (in particular @helgasoft, @mpetroff, @nowosad, @Robinlovelace, @zeileis, and @fmegahed,

The package is imho ready for CRAN, but still in development phase (so version < 1.0). I am still learning about color theory, so indicators may change, as well as the palettes that I recently added (seq/div and biv). Moreover, we are still unsure what the best way is to calculate distance matrices for color vision deficiency. Distances are calculated upstream (colorspace and colorblindcheck), but are used to calculate the color-blind-friendliness scores.

Please let me know if you find any issues.

Best,

Martijn

PS. I still have to document the used quality indicators. Hopefully we can open a discussion what good indicators are and which parameters to use.

[Nowosad]

Hi @mtennekes -- I just run the app and tried several options. A few tiny comments:

1. Why bivu for "Bivariate (sequential x desaturated)" and not bivd?
2. The app looks great when it covers the whole screen; however, some elements (e.g., buttons) do not look great when I resize the window.

[mtennekes]

Thx @Nowosad

1. Good point. Maybe that is better indeed. The reason I initially went for bivu was that it was called "bivariate-uncertainty", and that bivd might be confusing since it could also stand for "bivariate-diverging". (See below)
2. Yes, indeed. I am not a Shiny tiger, so PRs are welcome. Will also take a look myself. Recently I saw colorspace::hcl_wizard, which looks more professional (at least the source code), so maybe I can learn from that. It also has a 'dark mode', which works better (i.e. also the buttons selectizeinputs are updated).

Regarding 1: what do you think of the name bivc? Brewer has two variants for columns: categorical and diverging. However, I noticed that categorical with more than 2 hues does not work for people with CVD. Therefore I only have one variant: an n=3 diverging palette for columns and a sequential scale for rows. I named it bivc, but this may misleading (?) Alternatively, we could use bivd, but this would be a good candidate for the current bivu (your first remark).

What is your take on this?

[zeileis]

Regarding hcl_wizard(): That is some Reto-magic. He would say that he also isn't overly familiar with shiny but he clearly has done solid work here. If you have any concrete questions, you can contact him directly or I can pass the question on to him.

[Nowosad]

@mtennekes I have thought about the abbreviations for some time, but I still do not have a perfect solution.

One idea I have is to rename current bivu to bivl (bivariate (sequential x less saturated)), and use bivd for the diverging one.

[mtennekes]

Thx @Nowosad !

Also related: currently I have put seq x cat and seq x div into one type (currently called bivc). It would probably be better to have them as separate types (bivc and bivd), just to be complete, although I would not recommend any seq x cat palette, simply because they are not color blind friendly.

We can indeed use bivl for the seq x desaturated palettes. A downside is that "l" may also be associated with "light". Alternatively, we can use bivg where the g stands for gra/eyscale, but that is also not perfect...

Meanwhile, I have updated the GUI: smaller fonts and no more columns inside the input panel. Suggestions and PRs are welcome!

@zeileis thanks for the implementation of colorspace convention in linear RGB (gamma=FALSE) into colorspace. When is the next CRAN release of colorspace expected?

[zeileis]

I haven't scheduled a next release, yet. I'm also not fully sure that the current implementation is the most natural one. And the default for gamma is least clear to me.

All of these are indicators that it might help to wait and discuss a bit more before carving stuff into stone (which is what a CRAN release would do). But it would be good not to wait too long.

[Robinlovelace]

Sorry I've not had a chance to look at this yet, will try to take a peak asap but don't wait for me.

[mtennekes]

No worries and thanks for your help @Robinlovelace I'm still finetuning bits and pieces, so it can wait.

From now on I'll use the devel branch for the latest version and the master for a more stable version. Probably it is best to test the devel branch.

[mtennekes]

I've updated the GUI recently. The input widgets should be better, especially for smaller screens. I've also improved the bivariate palettes.

Testers go ahead! (@Robinlovelace @Nowosad @zeileis @helgasoft )

[Nowosad]

Looks good!

[fmegahed]

Hi all,

@mtennekes , I was finally able to host the app on our university server. It is available at http://cols4all.fsb.miamioh.edu/cols4all/.

To create the app, I have created the following two files:

• app.R, which loads the package from GitHub and install/load the dependencies.
• run_app.R, which utilizes on Shiny's runApp() to run the app by running the app.R file

@Robinlovelace @Nowosad @zeileis @helgasoft, if you can test the hosted app as well that would be awesome!!

[zeileis]

I took it for a very quick spin, looks good! 👍

[helgasoft]

Hosted app is great, love it 🏆

[Nowosad]

I cannot see the app neither on my laptop nor smartphone). I assume this my local issue (?) -- maybe some of you know how to fix it...

[fmegahed]

@Nowosad , it is working for me right now. Would you please retry it?

Please use http::// and not https://

[zeileis]

I double checked and it works on my phone and laptop. But it's important to include the trailing slash at the end after "cols4all". Without that the page is not found...

[Nowosad]

Thank you @fmegahed and @zeileis -- now it works with http:// and the trailing slash at the end.

[mtennekes]

I've started a issue-list, see https://github.com/mtennekes/cols4all/issues/13

@fmegahed The server is up again, but I cannot log-in to RStudio yet(?).