[Rd] rgl/webGL complains about Javascript, even in recent online docs?

[Dominick Samperi]

Thanks for the pointers and the quick fix.

Perhaps the generated HTML code should issue a
message like "Javascript load problem" instead of
"You must enable Javascript to view this page properly,"
because the latter can be misleading.


On Sat, Sep 12, 2015 at 12:42 PM, Duncan Murdoch
<murdoch.duncan at gmail.com> wrote:
> On 12/09/2015 7:37 AM, Duncan Murdoch wrote:
>> On 11/09/2015 10:14 PM, Dominick Samperi wrote:
>>> Hello,
>>>
>>> The recently created online "rgl Overview" at
>>> https://cran.r-project.org/web/packages/rgl/vignettes/rgl.html
>>> illustrates a problem that I am trying to resolve.
>>>
>>> At the bottom of each image block on that page appears the
>>> advisory: You must enable Javascript to view this page properly.
>>>
>>> I am using Safari under MacOS with Javascript and WebGL
>>> both enabled, so it must be the detection algorithm that is
>>> mistaken. It is not clear to me how this works after looking
>>> over the generated HTML code (generated by knit2html
>>> in my case).
>>>
>>> I found this after I ran into this problem in a different context,
>>> namely, in the process of creating a wordpress page that
>>> contains interactive rgl content. I tried simply placing
>>> the code generated by knit2html into a wordpress page,
>>> but the result is not interactive and that message about
>>> Javascript not being enabled appears.
>>>
>>> On the other hand, if I open the generated code in a
>>> browser directly, interaction works and there is no
>>> Javascript message.
>>>
>>> In summary, interactive rgl code works stand-alone, but
>>> not in wordpress, and not in the "rgl Overview" page.
>>>
>>
>> I see the same in Firefox, it's not just Safari.
>>
>> If I look at the browser console, I see errors indicating that rglClass
>> is not defined, and I don't see a block of Javascript code that should
>> have been inserted.
>>
>> If I install rgl from the source on CRAN, I see the same problem, so it
>> looks like an rgl bug.  I'll see if I can fix it.
>
> Yes, it was an rgl bug.  During my testing I always had knitr loaded
> before rgl, but when building the vignettes for the package, the load
> order is different, and the initialization was silently skipped.
>
> I'll soon be committing changes to R-forge to fix this, and eventually
> it will end up on CRAN.  At the moment it looks like I need to add this
> function:
>
> setupKnitr <- function() {
>   if (requireNamespace("knitr")) {
>     knitr::knit_hooks$set(webgl = hook_webgl)
>     knitr::knit_hooks$set(rgl = hook_rgl)
>     environment(hook_webgl)$commonParts <- TRUE
>     environment(hook_webgl)$reuse <- TRUE
>   }
> }
>
> and that should be called (just once) near the start of the vignette.
>
> Duncan Murdoch
>