Create html and rds treeviews by default

[russHyde]

Fix: #14

Adds a output_format parameter for both treeview() and save_trees() that allows the user to specify whether they would like to have the interactive ggtree objects saved in .rds files (as ggtree objects) or in .html files (as htmlwidgets).

By default the argument for output_format is c("rds", "html"), in which case it creates both .rds and .html versions of the interactive plots (as requested by Erik in meeting 2023-01-12).

The sina_output_format parameter that was previously included in the treeview() signature has been removed. Now the new output_format parameter determines the output format for the interactive versions of both sina plots and tree plots.

Tests have been added to ensure that the correct output files are created when save_trees() is called: a noninteractive plot is always made, and an interactive html and/or interactive rds plot is made based on user arguments to save_trees()

[russHyde]

The initial version of this added a output_format to the treeview() parameters.
In the "sinaplot-rds" branch (now merged into "staging-wp5" but not yet "master"), a sina_output_format parameter was added.
It seems appropriate to rename the output_format as tree_output_format, or to use a single output_format parameter to control both tree output and sina output.

WDYT @jr-nicola

[jr-nicola]

I'd lean towards having a single output_format that controls both. For the app, if we want the .rds for one, we'd always want the .rds for the other. There might be times when Erik only wants the .html for the treeview but that seems less likely after the app is deployed.

[russHyde]

Thanks. I'll sort that out.

[russHyde]

These tests have been constructed so that they look virtually identical. They could reasonably be rewritten as a single testthat block parameterised by the different arguments to save_trees using {patrick}. Adding the extra dependency seemed over the top to reduce 4 simple tests to 1, though.

[jr-nicola]

Happy to have a little duplication to avoid extra dependencies

[russHyde]

@jr-nicola I've changed a dependency. The package now requires {htmlwidgets} >= 1.6.0.
I found, while running the tests associated with this PR, that with htmlwidgets < 1.6.0, I was getting an error when using system R and lots of pandoc warnings when using RStudio.

# System R
Error in `system(paste(shQuote(pandoc_path), "--version"), intern = TRUE)`: error in running command
Backtrace:
  1. tfpscanner:::save_trees(...)
       at test-plot_tree.R:101:4
  2. htmlwidgets::saveWidget(...)
       at tfpscanner/R/plot_tree.R:150:4
  3. htmlwidgets:::pandoc_available()
  4. htmlwidgets:::find_pandoc()
  5. base::lapply(...)
  6. htmlwidgets (local) FUN(X[[i]], ...)
  7. htmlwidgets:::get_pandoc_version(src)
 10. base::system(paste(shQuote(pandoc_path), "--version"), intern = TRUE)

# RStudio
[WARNING] Deprecated: --self-contained. use --embed-resources --standalone

These originated from the discovery and use of different versions of pandoc in {htmlwidgets} and {rmarkdown}. In brief, htmlwidgets was discovering the 'system' pandoc (2.9* on my machine) and creating the appropriate arguments for that version of pandoc but then sending those arguments to an 'RStudio'-embedded pandoc (2.19* on my machine; this is the one that rmarkdown uses when inside RStudio) for which they are invalid.

As of htmlwidgets 1.6.0, the method for discovering/using pandoc is aligned between htmlwidgets and rmarkdown.

Pinning against htmlwidgets >= 1.6.0 should prevent these clashes, and removed the errors and warnings on my machine.

[jr-nicola]

LGTM