At least for the online version, it'd be nice to have gifs that link to the interactive.

For the print version, it prob make most sense to explicitly write the hyperlink in the figure caption

Setting this in index.Rmd might do the trick

getOption("knitr.graphics.auto_pdf", !knitr::is_html_output())

This could force include_graphics() to use the pdf version of an image in the print version, if it exists.

And ideally already have plotly loaded so folks can just copy/paste examples to get the interactive.

Mention this instance in the initial overview.

As in the first chart at this page:
https://plotly-book.cpsievert.me/key-frame-animations.html
the one resulting from this code:
data(gapminder, package = "gapminder")
gg <- ggplot(gapminder, aes(gdpPercap, lifeExp, color = continent)) +
geom_point(aes(size = pop, frame = year, ids = country)) +
scale_x_log10()
ggplotly(gg)

The legend for size variable is not visible properly, the text is half cut.
Is there anything to do to see it in full?

• zikar
• eechidna
• bcviz

To acknowledge all contributors

Lots of reviewers said these chapters are "not complete"

Hi Carson,

Thanks for the plotly book - it's extremely helpful!

I just wanted to point out a small issue: the Figures in Section 3.1 do not seem to render in the book. The code works if I copy it into an RMarkdown file though!

Best,
Rebecca

Perhaps similar to some code from @aloy:

# Script accessing country-level population data via Census Bureau
library(idbr)
library(dplyr)
library(purrr)
library(tidyr)

female <- idb1(country = "BE", year = 2018:2050, variables = c("AGE", "NAME", "POP"), sex = "female")
male   <- idb1(country = "BE", year = 2018:2050, variables = c("AGE", "NAME", "POP"), sex = "male")

belgium <- bind_rows(female, male)

make_ids <- function(.data, .order) {
  base_ids <- 1:nrow(.data)
  extra_ids <- if(.order > 0) seq(from = max(base_ids) + 1, to = max(base_ids) + .order)
  ids <- c(extra_ids, base_ids)
  ids[1:101]
}

be_nested <- belgium %>%
  group_by(SEX, time) %>%
  tidyr::nest() %>%
  mutate(order = time - 2018) %>%
  mutate(ids = map2(data, order, .f = make_ids)) 

belgium2 <- be_nested %>%
  select(-order) %>%
  unnest()

plot_ly(belgium2, color = ~factor(SEX), size = I(5), alpha  = 0.5, colors = c("red", "black")) %>%
  add_segments(x = ~POP, xend = 0, y = ~AGE, yend = ~AGE, frame = ~time)

1st of all, congratulations to @cpsievert for a great and enlightening book.

I'm using the standard variety of R, from the current package of the stable Debian repo. It's still R 3.3.3, while some of the packages in the book, like trelliscopejs demand R 3.4.0, at least.

So, my question is: what about backporting trelliscope to R 3.3.3 ?

Or maybe using only packages compatible while R 3.3.3 ?

While downloading and compiling R 3.4.0 is probably not that difficult, maybe some of your readers are also using versions older than 3.4.0.

Thank you !

On arriving at section 1.2.2 Modifying Layers, p was previously assigned as follows:

p <- ggplot(fortify(gold), aes(x, y)) + geom_line()

This section says that there are three traces within the data, however there is only a "0" appearing under data. In contrast, if p is assigned as follows, I see a 0, 1 and 2 which I understand to be the three traces referred to in the section

`p <- ggplot(mtcars, aes(x = wt, y = mpg)) +
geom_point() + geom_smooth()

plotly_json(p)
`
This is my first look at plotly, so I could be wrong! but otherwise it is not clear ....

Already have a section on creating figures on plot.ly, but could use a section modifying them?

A much faster alternative to rendering polygons compared to scattermapbox

Could maybe recreate the cartogram examples with style='white-bg'?

Add an example of drilling down in a single view, like https://edgarruiz.shinyapps.io/db-dashboard/

Also, now that we have sunburst, would be nice to mention this 'native' drill-down ability, and how it can be linked to other views

Based on discussion in plotly/plotly.R#1441

Arrow turns to crosshairs when entering barplot

Perhaps extending the freqpoly example in the intro to ggplotly

This section is currently written for an academic audience -- it should have less theory and more practice (i.e., many more examples!)

linked_highlighting is tagged in the following sections (but clicking on 'linked_highlighting' does not link to more information on the topic)

2.1.3.2 Parallel Coordinates
Section 2.3 Bars & Histograms, note 7: This has some interesting applications for linked highlighting as it allows for summary statistics to be computed on-the-fly based on a selection

As in HTML5 <summary>/<details>. This could provide a light-weight solution to the lack of code-folding.

Also unable to create from code

Perhaps make this note in polygons section and make a nod to it in mapping

Also, make sure all the good ideas from here are in the book https://blog.cpsievert.me/2018/03/30/visualizing-geo-spatial-data-with-sf-and-plotly/

https://plotly-r.com/translating-custom-ggplot2-geoms.html to
https://plotly-r.com/custom-geoms.html

Custom geographic and mapbox layout options are not displayed when viewing the plotly book on Chrome.

Additionally, when pasting the code onto my computer (using RStudio) the custom geographic layout is blank. For mapbox, I was informed that a token was required (I did not look into this further).
On inspecting the blank plot region for custom geographic layout and hitting the console tab, I see an error "Failed to load resource: Unable to init SSL Context:"

I have just reviewed Plotly issues 356 and 1111 and wondering if it is a problem with working behind a firewall? Although would that also affect viewing the book on Chrome?

Using albersusa, extending the current example

Folks seem to want this

• Get installation of package working (#72)
• Include images/ folder

So that printed links can be ported to another location, if need be

404

e.g. ch 4.2.3
I'm using plotly 4.5.6.9000

Why interactive data visualization?

• Data science workflow
• Interactivity augments exploration
  • Some historical context and examples (https://talks.cpsievert.me/20180202/#7)
  • "Interactive graphics enable the analyst to pursue follow-up questions" -- Cook & Swayne, GGobi book, p15
  • A case study of US election data
• GUI and command-line - conflict or synergy?
  • https://talks.cpsievert.me/20180305/#12
• Web-based visualization
  • New capabilites (e.g. easy sharing, portability) brings new set of concerns (e.g., client-server, security, etc)
  • Great tools for expository vis (d3.js, vega, plotly.js)
  • Lack of tools for exploratory vis (i.e. tools for iteration)

The plotly for R ecosystem

• plotly is a FOSS tool built with plotly.js and htmlwidgets
• Plotly's cloud services
  • collaboration, point-and-click editing & collaboration
• Getting started (installation, etc)
• Two approaches, one object

The ggplotly() toolbox

• Makes some plots easier to acheive (e.g., geom_hex(), geom_smooth(), etc)
• Caveats and limitations
• Learn plotly.js via ggplotly()

The plot_ly() toolbox

• Scatter traces
• Maps
• Bars & histograms
• 2D frequencies
• Raster images
• 3D graphics

Animating views

• Key frame animations
• Linking animated views

Arranging multiple views

• Arranging htmlwidgets
• Merging plotly objects
• Navigating many views

Multiple linked views

• Linking views with shiny
• Linking views without shiny

Exporting views

• Saving local HTML files
• Embedding local HTML files (i.e., iframe, widgetFrame, etc)
• Using orca() to export to various file formats
• Using config(p, editable = T) to manually edit charts
• Exporting manually edited charts
  • Should this just leverage config(p, toImageOptions = ...)? Or provide a way to get at the underlying JSON?

Performance considerations

• What's in a plotly object?
• What happens at print-time?
• Build-time versus render time
• Build time
  • profvis
• Render time
  • SVG vs Web-GL rendering
  • Data summary/simplification

Custom JavaScript

• Motivate through examples
  • Opening hyperlinks
  • Adding annotations in response to direct manipulation

Other ideas

• Interactive annotations via https://github.com/plotly/react-chart-editor (would require a proper wrapper)
• Manipulating widgets with https://github.com/rte-antares-rpackage/manipulateWidget

Advanced case studies

• storm paths -- https://vimeo.com/257149623
• https://github.com/cpsievert/pedestrians
• https://github.com/cpsievert/zikar
• https://github.com/cpsievert/eechidna
• https://github.com/cpsievert/accidents

plotly extension packages

• Hmisc, heatmaply, TSstudio, autoplotly

Like

s <- schema()
x <- grep("-", s$traces$scatter$attributes$marker$symbol$values, value = T)
plot_ly(x = seq_along(x), y = seq_along(x), symbol = I(x), span = I(1), stroke = I("black"), marker = list(size = 30), text = x, hoverinfo = "text")

Sort of like this example, which currently requires plotly/plotly.R#1392

library(shiny)
library(plotly)

ui <- function(req) {
  fluidPage(
    bookmarkButton(),
    plotlyOutput("plot")
  )
}

server <- function(input, output, session) {
  
  output$plot <- renderPlotly({
    plot_ly(economics, x = ~pop, source = "pop_hist")
  })
  
  onRestored(function(state) {
    hist_relayout <- jsonlite::parse_json(
      state$input[["plotly_relayout-pop_hist-input"]]
    )
    xrng <- as.numeric(hist_relayout[c("xaxis.range[0]", "xaxis.range[1]")])
    
    if (length(xrng)) {
      plotlyProxy("plot") %>% 
        plotlyProxyInvoke("relayout", list(xaxis = list(range = xrng)))
    }
  })
}

shinyApp(ui, server, enableBookmarking = "url")

"the x and y attributes do not have to be provided, but if they are, the length of x should match the number of rows in the matrix and y should match the number of columns."

The statement is not correct. It's exactly the opposite: the length of x should match the number of columns in the matrix and y should match the number of rows. (i agree that plotly is bit weird in this regard.)

x <- seq_len(ncol(volcano)) + 100
y <- seq_len(nrow(volcano)) + 500
plot_ly() %>% add_surface(x = ~x, y = ~y, z = ~volcano)

If screenshots are needed, put them in images

e.g., what is a trace? what is a figure?

(not sure this is the right place for this comment. If this is not, I apologize).
I have a caveat with the example of section
"4.2.3 Linking with other htmlwidgets"
The example seems dependent upon
install_github("rstudio/leaflet#346")
otherwise, if I just install
install_github("rstudio/leaflet")
I get blue color for points on the right map when selecting on the left scatterplot (where points are red or
the selected color), which is inconvenient as selected points cannot be identified on the map.
As this type of linked brushing is very important for me, I would like to know if this code will be
valid for future standard releases of leaflet: sticking to #346 might not be possible in the future or interfere with newer improvements.
Thanks

At least for the static plots on the website

This is very minor, but when I copy/paste in Windows some of the code examples give an error since “ is used instead of ".

Currently gifs autoplay, which can be quite distracting and annoying for longer gifs if you want to revisit a particular section

One option is to add play/pause to gifs with http://rubentd.com/gifplayer/

Assuming we can get Figure numbering working, embedding vimeos might be a better option. Note that you can embed without the cruft with https://player.vimeo.com/video/257149623?title=0&byline=0&portrait=0

e.g.,

• More cohesive explanation of updatemenus than https://plot.ly/r/dropdowns/
  • Perhaps draw connections to plotlyProxy()?
• How to add margin(s) to a plot?
• How to customize axes (e.g., titles, tickvals/ticktext, etc)?
• Why does my line plot appear blank (i.e., malformed group definition)?
• Why is my plot being sluggish
  • Point to SVG vs WebGL
• Why is my plot's file size so big?
  • Point to partial_bundle()
• How can a plotly plot from graying-out in shiny?
  • Point to shiny performance section

What software you are using to create your animated gifs?

It should take more time to build the core concepts, like http://bit.ly/jj-shiny-day does

Also, take advantage of plotly/plotly.R#1655, when fixed

Works fine locally, but apparently not on shinyapps

• drag_markers
• drag_lines
• drag_brush
• crossfilter_scatter
• crossfilter_kde

Namely the examples towards the end here https://blog.cpsievert.me/2018/01/30/learning-improving-ggplotly-geom-sf/

Hey there,
I just skimmed over your map chapters, really interesting! And I know that you focus on plotly but I would suggest to at least mention the popular and powerful packages mapview (https://github.com/r-spatial/mapview) and tmap (https://github.com/mtennekes/tmap) when it comes to interactive mapping in R.

https://plotly-r.com/flexdashboard.html

Expand on the current section to include:

• A brief intro to JavaScript (for R users)
• A brief intro to (htmlwidgets::onRender()) (includes limitations -- does it work in shiny?)
• More compelling examples (some could be borrowed from https://rstudio.cloud/project/14090)

Maybe take an example from ?scale_color_gradient2