[Bug]: `saveRDS` being used in reporter creates large reports & bookmark folder

[averissimo]

What happened?

Problems:

• Size of report and bookmarks folders is very large on the current implementation of teal_reportable
• Time it takes to generate the report is affected negatively

This happens due to the way saveRDS saves reference objects (for example: environments), we can indirectly observe this by comparing the space in memory and the estimated size via {lobstr} package

Minimal example:

sample_plot <- function() ggplot2::ggplot(iris) + ggplot2::geom_histogram(ggplot2::aes(x = Sepal.Length), binwidth = .1)

path_bare <- withr::local_tempfile()
obj0 <- sample_plot()
{
  tictoc::tic("With namespace")
  saveRDS(obj0, file = path_bare)
  tictoc::toc()
}

{
  glue::glue("obj0: (obj.si) ", object.size(obj0) |> format(units = "auto")) |> print()
  glue::glue("obj0: (lobstr) ", lobstr::obj_size(obj0) |> format(units = "auto")) |> print()
  glue::glue("obj0: (disk)   ", structure(file.size(path_bare), class = "object_size") |> format(units = "auto")) |> print()
}
#> With namespace: 0.129 sec elapsed
#> obj0: (obj.si) 12.7 Kb
#> obj0: (lobstr) 4.59 MB
#> obj0: (disk)   205 Kb

sessionInfo()

Relevant log output

Code of Conduct

• I agree to follow this project's Code of Conduct.

Contribution Guidelines

• I agree to follow this project's Contribution Guidelines.

Security Policy

• I agree to follow this project's Security Policy.

[averissimo]

ℹ️ note that the saved rds here is not that large, but when called from a teal app, it is at least 1 order of magnitude larger

[averissimo]

Proposed solution after call with @gogonzo:

• Using recordedplot as object that is stored in place of ggplots
  • From our limited tests it is still scalable once re-imported

See the small minimal example below, where the file size on disk (on a clean session) is 19Kb vs 424Kb (recordedplot vs ggplot)

The size discrepancy becomes even bigger once inside a teal app 19Kb vs 1100kb

    sample_plot <- function() ggplot2::ggplot(iris) + ggplot2::geom_histogram(ggplot2::aes(x = Sepal.Length), 
        binwidth = 0.1)
    g <- sample_plot()
    res <- evaluate::evaluate("print(g)", output_handler = evaluate::new_output_handler(value = identity))
    recorded_path <- withr::local_tempfile()
    ggplot_path <- withr::local_tempfile()
    res
#> <evaluation>
#> Source code: 
#>   print(g)
#> Plot [3]:
#>   <base> palette2()
#>   <grid> requireNamespace("ggplot2", quietly = TRUE)
#>   <grid> drawGTree(x)
    class(res[[2]])
#> [1] "recordedplot"
    saveRDS(res[[2]], file = recorded_path)
    saveRDS(g, file = ggplot_path)
    cat(format(structure(file.info(recorded_path)$size, class = "object_size"), 
        units = "auto"), "\n")
#> 19.1 Kb
    cat(format(structure(file.info(ggplot_path)$size, class = "object_size"), 
        units = "auto"), "\n")
#> 424.3 Kb

[gogonzo]

@averissimo I think we can close this?