multiple (R) plotly figures generated in a Rmarkdown (knitr chunk) document

Question

I try to create multiple plotly figures in a Rmarkdown document using loop or lapply.

The R script:

require(plotly)
data(iris)
b <- lapply(setdiff(names(iris), c("Sepal.Length","Species")),
            function(x) {
              plot_ly(iris, 
                      x = iris[["Sepal.Length"]],
                      y = iris[[x]], 
                      mode = "markers")
            })
print(b)

works well, but it fails when included in a knitr chunk:

---
output: html_document
---

```{r,results='asis'}
require(plotly)
data(iris)
b <- lapply(setdiff(names(iris), c("Sepal.Length","Species")),
            function(x) {
              plot_ly(iris, 
                      x = iris[["Sepal.Length"]],
                      y = iris[[x]], 
                      mode = "markers")
            })
print(b)
```

I tried replacing print(b) with a combination of lapply eval and parse but only the last figure was displayed.

I suspect a scoping/environment issue but I cannot find any solution.

Answer 1

Instead of print(b), put b in htmltools::tagList(), e.g.

```{r}
library(plotly)

b <- lapply(
  setdiff(names(iris),
          c("Sepal.Length","Species")),
  function(x) {
    plot_ly(iris, 
            x = iris[["Sepal.Length"]],
            y = iris[[x]], 
            mode = "markers")
  }
)

htmltools::tagList(b)
```

Note: Before Plotly v4 it was necessary to convert the Plotly objects to htmlwidgets using Plotly's as.widget() function. As of Plotly v4 they are htmlwiget objects by default.

For people who are interested in the technical background, you may see this blog post of mine. In short, only top-level expressions get printed.

Answer 2

I found a "dirty" solution by using temp file and kniting it :

```{r,echo=FALSE}
mytempfile<-tempfile()
write("```{r graphlist,echo=FALSE}\n",file=mytempfile)
write(paste("p[[",1:length(p),"]]"),file=mytempfile,append = TRUE)
write("\n```",file=mytempfile,append = TRUE)
```

`r knit_child(mytempfile, quiet=T)`

But it's unsatisfactory.

Answer 3

For anyone struggling with a loop, here's what worked for me.

p=list()
for (n in 1:3){
  p[[n]] <- plot_ly(x = 1:100, y = rnorm(100)+n, mode = 'lines', type="scatter")
}
htmltools::tagList(p)

I.e. it doesn't matter if the list p is created in a loop or lapply, etc. as long as you call htmltools::tagList outside the loop.

Thanks to Yihui for helping me get there and for immense work developing and helping with these tools.