Allow trailing commas in more places

[hadley]

I grepped for list(...) and replaced with rlang::list2(...). This also enables !!! which is generally not important for Shiny because it automatically splices lists/tagLists, but I doubt it will affect any existing code.

Let me know if you want me to document this, or add a news bullet.

[cpsievert]

Since there was some overlap with messaging for #3315 I took a stab at a NEWS item (and fixed the merge conflict).

[hadley]

Suggested change

* All uses of `list(...)` have been replaced with `rlang::list2(...)`. By effect, `{rlang}`'s `!!!` operator may now be used to "splice" a list of argument values into `...`. We think this'll be particularly useful for passing a list of `tabPanel()` to their consumers (i.e., `tabsetPanel()`, `navbarPage()`, etc). For example, `tabs <- list(tabPanel("A", "a"), tabPanel("B", "b")); navbarPage(!!!tabs)`. (#3315 and #3328)

* All uses of `list(...)` have been replaced with `rlang::list2(...)`. This means that you can use trailing `,` without error and use rlang's `!!!` operator to "splice" a list of argument values into `...`. We think this'll be particularly useful for passing a list of `tabPanel()` to their consumers (i.e., `tabsetPanel()`, `navbarPage()`, etc). For example, `tabs <- list(tabPanel("A", "a"), tabPanel("B", "b")); navbarPage(!!!tabs)`. (#3315 and #3328)

Is that true about splicing though? I thought tagList() already did auto-splicing?

[cpsievert]

Yea, I believe all of htmltools uses dots_list() now. I couldn't find any mention of it in the NEWS though so I figured it wouldn't hurt to say it just generally works now with any ...?

[cpsievert]

Also, shouldn't the suggestion be without error instead of with out warning?

[hadley]

Ooops yeah. But I meant doesn't splicing already happen without !!!?

[cpsievert]

Ohh, technically no, but in practice the difference usually isn't important (at least as far as renderTags() is concerned)

> str(tagList(list(tags$a(), tags$div())))
List of 1
 $ :List of 2
  ..$ :List of 3
  .. ..$ name    : chr "a"
  .. ..$ attribs : Named list()
  .. ..$ children: list()
  .. ..- attr(*, "class")= chr "shiny.tag"
  ..$ :List of 3
  .. ..$ name    : chr "div"
  .. ..$ attribs : Named list()
  .. ..$ children: list()
  .. ..- attr(*, "class")= chr "shiny.tag"
 - attr(*, "class")= chr [1:2] "shiny.tag.list" "list"
> str(tagList(!!!list(tags$a(), tags$div())))
List of 2
 $ :List of 3
  ..$ name    : chr "a"
  ..$ attribs : Named list()
  ..$ children: list()
  ..- attr(*, "class")= chr "shiny.tag"
 $ :List of 3
  ..$ name    : chr "div"
  ..$ attribs : Named list()
  ..$ children: list()
  ..- attr(*, "class")= chr "shiny.tag"
 - attr(*, "class")= chr [1:2] "shiny.tag.list" "list"

[hadley]

Suggested change

* All uses of `list(...)` have been replaced with `rlang::list2(...)`. This means that you can use trailing `,` without error and use rlang's `!!!` operator to "splice" a list of argument values into `...`. We think this'll be particularly useful for passing a list of `tabPanel()` to their consumers (i.e., `tabsetPanel()`, `navbarPage()`, etc). For example, `tabs <- list(tabPanel("A", "a"), tabPanel("B", "b")); navbarPage(!!!tabs)`. (#3315 and #3328)

* All uses of `list(...)` have been replaced with `rlang::list2(...)`. This means that you can use trailing `,` without error and use rlang's `!!!` operator to "splice" a list of argument values into `...`. (#3315 and #3328)

Since it doesn't provide any new benefit, I think it's better to not over advertise it.

[cpsievert]

!!! inside tabsetPanel() is new (was merged earlier today), but !!! inside tagList() isn't.

That said, I'd be OK with removing it, but I'll leave that up to @wch