Use global option for R package verbosity

[rcurtin]

This fixes #3692 in just the way that @cgiachalis suggested.

Instead of the verbose option being printed with a default of FALSE (e.g. verbose = FALSE), it is now printed as verbose = getOption("mlpack.verbose", FALSE). That is, it defaults to false unless mlpack.verbose has been set (at least to the best of my limited R understanding).

I also added tests to make sure that it operates correctly.

@eddelbuettel @coatless let me know what you think here, a double-check would be much appreciated. 👍

[eddelbuettel]

Using verbose = getOption("mlpack.verbose", FALSE) is indeed a good pattern and somewhat idiomatic ... but I do not see it in the diff for this PR. Did I miss it? It's been a long day...

[coatless]

@eddelbuettel yes; it's here:

https://github.com/mlpack/mlpack/pull/3706/files#diff-0fd950ea62a1bd092169e153e05f92bedfe37985aebfef1f2512cc4493229d8eR34-R43

@rcurtin modified how function signatures are created. So, now when the bindings detect verbose it prints:

knn <- function(.,.,..., verbose = getOption("mlpack.verbose", FALSE) )

instead of:

knn <- function(.,.,..., verbose = FALSE)

[rcurtin]

I realized I also need to update the Markdown documentation; in essence I changed the code so that the table of parameters that gets listed for each binding lists the default of the verbose option as its new value, instead of FALSE.
e.g. this table: https://www.mlpack.org/doc/stable/r_documentation.html#input-options-19

Now, the line of Markdown in that table will instead be:

| `verbose` | [`logical`](#doc_r_logical) | Display informational messages and the full list of parameters and timers at the end of execution. | `getOption("mlpack.verbose", FALSE)` |

[rcurtin]

Using verbose = getOption("mlpack.verbose", FALSE) is indeed a good pattern and somewhat idiomatic ... but I do not see it in the diff for this PR. Did I miss it? It's been a long day...

Sorry about that, I should have been more clear in the description. Here is an example of the newly generated binding:

kfn <- function(algorithm=NA,                                                                                                                                                              
                epsilon=NA,                                                                                                                                                                
                input_model=NA,                                                                                                                                                            
                k=NA,                                                                                                                                                                      
                leaf_size=NA,                                                                                                                                                              
                percentage=NA,                                                                                                                                                             
                query=NA,                                                                                                                                                                  
                random_basis=FALSE,                                                                                                                                                        
                reference=NA,                                                                                                                                                              
                seed=NA,                                                                                                                                                                   
                tree_type=NA,                                                                                                                                                              
                true_distances=NA,                                                                                                                                                         
                true_neighbors=NA,                                                                                                                                                         
                verbose=getOption("mlpack.verbose", FALSE)) {   

I also noticed that the documentation for the parameter verbose in the generated .R files was wrong. So now after the commit I just pushed, verbose is documented as:

#' @param verbose Display informational messages and the full list of                                                                                                                      
#'   parameters and timers at the end of execution.  Default value                                                                                                                         
#'   "getOption("mlpack.verbose", FALSE)" (logical).   

I think I didn't forget any other place documentation should be?

[coatless]

@rcurtin this looks great. I'm going to wait until the R build passes R CMD check/the testthat suite returns an all clear 😄

[rcurtin]

Sounds great, fingers crossed 😄

[coatless]

LGTM!

Though, need to work on dropping the compile time of the package.

[rcurtin]

The test failure looks like an unrelated unlucky failure; I don't think the changes here caused it. Someday I will figure out how to isolate that test, but I still have yet to manage to reproduce the issue locally.

[rcurtin]

Though, need to work on dropping the compile time of the package.

Totally agree. It is getting painful (even more so than it was before, which was already a considerable amount of pain!) and needs some investigation.

[mlpack-bot]

Second approval provided automatically after 24 hours. 👍

[eddelbuettel]

@rcurtin Just thinking out loud here: maybe ccache if we can persist the cache layer? (I know GHA has some caching logic, but I never used it. There is also sccache which someone at work looking into / uses (?) but I have no simple example to point at.)

What else were you thinking of?

[rcurtin]

ccache would probably be helpful, but I think that there is some lower-level improvements that could be made, like compilation flag improvements, or maybe some CMake improvements, I'm not sure. I need to put a few hours into looking into it...