Update vendored renv; update script to allow taking tagged versions

[toph-allen]

This PR updates the embedded version of renv used by Packrat. Since the current HEAD on the renv repo is not passing CI, I took the previous version. To do this, I modified the tools-sync-renv.R script to allow passing a single argument, like so:

R -f tools/tools-sync-renv.R --args 0.15.5-48

If there are zero arguments, the script will proceed as before. If there is one argument, it uses that in a modified git checkout command. If there is more than one argument, it stops and emits a warning.

QA notes

These are taken from the last "bump renv" PR.

We want to ensure that this change doesn't break dependency detection, which is currently the only place we use renv. One way to exercise this is to deploy a simple Shiny app to an RStudio Connect instance.

After installing the version of Packrat in this PR, deploy the following Shiny app to a Connect instance and ensure that it runs.

library(shiny)

ui <- fluidPage(
    titlePanel("Old Faithful Geyser Data"),
    sidebarLayout(
        sidebarPanel(
            sliderInput("bins",
                        "Number of bins:",
                        min = 1,
                        max = 50,
                        value = 30),
            div("username", verbatimTextOutput("user")),
            div("groups", verbatimTextOutput("groups"))
        ),

        mainPanel(
            plotOutput("plot")
        )
    )
)

server <- function(input, output, session) {
  output$plot <- renderPlot({
    x    <- faithful[, 2]
    bins <- seq(min(x), max(x), length.out = input$bins + 1)

    hist(x, breaks = bins, col = 'darkgray', border = 'white')
  })

  output$user <- renderText(session$user)
  output$groups <- renderText(session$groups)
}

shinyApp(ui = ui, server = server)

[aronatkins]

Idle musing about the renv-updater script.

[aronatkins]

How do you feel about requiring a ref? That could be a commit or a SHA, but would force us to always be explicit about what renv code we want.

[toph-allen]

I don't feel that strongly. I made this choice for two reasons:

1. It looks like it's more complex to shallow clone a ref than a tag or branch
2. The old TODO commend indicated that the author was thinking about branches ¯_(ツ)_/¯

Also

3. renv has about one tag per commit on main, so it's functionally not too different

[aronatkins]

The header comment indicates the current commit hash; should it also include the tag/ref (which may be a commit)?

[toph-allen]

Good point! I'll include the argument that was used to clone a specific tag if one was provided

[aronatkins]

Are renv tags ever overwritten? It appears as if @kevinushey generally bumps the version, but I don't know if that's true of every commit.

Should we suggest using commit references rather than tags?

[toph-allen]

I'll see what it would be like to use commit references.

[toph-allen]

To do this would require changing the conditional flow to the following, and would only accept full shas, not abbreviations.

if (is.null(ref)) {
  # check out the latest renv sources
  system("git clone --depth 1 https://github.com/rstudio/renv")
  twd <- setwd("renv")
} else if (length(ref) == 1) {
  # check out the tagged version
  dir.create("renv")
  twd <- setwd("renv")
  system("git init")
  system("git remote add origin https://github.com/rstudio/renv")
  system(paste("git fetch --depth 1 origin", ref))
  system("git checkout FETCH_HEAD")
}

gitCommitHash <- system("git rev-parse HEAD", intern = TRUE)
setwd(twd)

I'm going to leave this here for now, and not make this change unless you feel strongly that this is better. I feel comfortable treating tags on renv as reasonable ways to find commits, and as you wrote in another comment, the header comment in the resulting renv.R file mentions the commit that was used, so… I don't think this causes too much of a problem.

[toph-allen]

@aronatkins I updated logic to remove one of the conditional statements by just capturing the command to use. Now the message at the top of renv.R includes the git clone command used and the timestamp at which it was called. The message also includes the renv version and the hash.

[aronatkins]

No need to adjust further. We can continue to iterate on the renv sync script as we take additional updates.

[aronatkins]

I don't feel like the timestamp adds information and could do without it.

[ChaitaC]

verified on packrat branch toph-update-renv

RemoteRef: toph-update-renv
RemoteSha: 7d423359b75a1fd0214ccbfb2078d7a16dee812e
GithubRepo: packrat
GithubUsername: rstudio
GithubRef: toph-update-renv
GithubSHA1: 7d423359b75a1fd0214ccbfb2078d7a16dee812e
NeedsCompilation: no
Packaged: 2022-08-10 14:37:32 UTC; chaitamacpro
Built: R 4.1.1; ; 2022-08-10 14:37:33 UTC; unix