如何跳过在htmlwidgets::saveWidget()中编写依赖项？

Question

当使用plotly可视化数据时，我希望将小部件编写为html-documents，而不是每次都使用htmlwidgets::saveWidget编写依赖项，假设这些都已经就位，以节省处理时间。小部件需要是自包含的，以节省磁盘空间。

library(plotly)
t <- Sys.time()
p <- plot_ly(ggplot2::diamonds, y = ~price, color = ~cut, type = "box")
htmlwidgets::saveWidget(as_widget(p), "test.html", selfcontained = F, libdir = NULL)
print(Sys.time() - t)

我机器上的Time difference of 4.303076 secs。

这仅在依赖中产生约6MB的数据(串音-1.0.0，htmlwidgets-1.2，jquery-1.11.3，plotly-binding-4.7.1.9000，plotly-htmlwidgets-css-1.38.3，plotly-main-1.38.3，typedarray-0.1)

htmlwidgets::saveWidget写入依赖关系，尽管这些文件已经存在。这是可以防止的吗？

Answer 1

问得好。我尝试在代码中以内联注释的形式回答问题。htmlwidgets依赖项有两个来源：htmlwidgets::getDependency()和小部件列表中的dependencies元素。将dependencies中的src元素更改为href而不是file意味着不会复制这些依赖项。然而，来自htmlwidgets::getDependency()的依赖项很难覆盖，但在这种情况下，将只复制htmlwidgets.js和plotly-binding.js，与其他四个相比，它们相当小。

library(plotly)

p <- plot_ly(ggplot2::diamonds, y = ~price, color = ~cut, type = "box")

# let's inspect our p htmlwidget list for clues
p$dependencies

# if the src argument for htmltools::htmlDependency
#   is file then the file will be copied
#   but if it is href then the file will not be copied

# start by making a copy of your htmlwidget
#   this is not necessary but we'll do to demonstrate the difference
p2 <- p

p2$dependencies <- lapply(
  p$dependencies,
  function(dep) {
    # I use "" below but guessing that is not really the location
    dep$src$href = "" # directory of your already saved dependency
    dep$src$file = NULL
    
    return(dep)
  }
)

# note this will still copy htmlwidgets and plotly-binding
#  requires a much bigger hack to htmlwidgets::getDependency() to change

t <- Sys.time()
htmlwidgets::saveWidget(as_widget(p), "test.html", selfcontained = F, libdir = NULL)
print(Sys.time() - t)

t <- Sys.time()
htmlwidgets::saveWidget(as_widget(p2), "test.html", selfcontained = F, libdir = NULL)
print(Sys.time() - t)