Provide more informative/helpful condition messages in Abort, Warn(, Inform?)

[brookslogan]

rlang conditions allow us to assign custom classes (allowing selective filtering), pack in relevant information within the condition object, and provide more informative traces than traceback. However, some of this functionality is not at all obvious for abort, and most/all of it is not obvious or activated for warn and inform. Plus, packing information inside the object rather than the message may actually obscure warn and inform and require an extra step for abort without some duplication inside the main message.

Help users unfamiliar with rlang conditions by:

• Providing a short tutorial periodically when we use rlang conditions, overviewing how to view condition fields, how to get an improved trace, and how to enable these things for warn and inform when they aren't available by default (on some/all R distributions) (requiring global_entrace() first instead). Try this using inform with a .frequency.
• Making it easy in Abort, Warn, to package in some informative objects that should be printed, and others whose names should be listed, placing them either with appropriate package-name prefixes as condition fields, or as a list condition field.
• Consider an analogous Inform function.
• In Abort, provide the caller's function name rather than "Error in Abort", using call parameter.
• Tweak init to adjust the wrapping based on the actual prefixes used by rlang, e.g., "! " for abort. (Might need to know if global_entrace is active)
• Make sure there's a way to use recover; look into what happened with this Issue.
• Make condition class names more readily available for selective recovery? Also, maybe have length 2 or 3 subclasses that will allow to catch all errors or all conditions from this package?

[brookslogan]

It looks like using recover is not currently possible via options(error=recover), but is possible via withCallingHandlers. The interactive session below gives an example.

R version 4.0.5 (2021-03-31) -- "Shake and Throw"
Copyright (C) 2021 The R Foundation for Statistical Computing
Platform: x86_64-redhat-linux-gnu (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

  Natural language support but running in an English locale

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(rlang)
> 
> 
> 
> inner_fn_using_stop = function() { result = 42; stop("stop message") }
> outer_fn_using_stop = function() { inner_fn_using_stop() }
> options(error=recover)
> outer_fn_using_stop()
Error in inner_fn_using_stop() : stop message

Enter a frame number, or 0 to exit   

1: outer_fn_using_stop()
2: #1: inner_fn_using_stop()

Selection: 2
Called from: outer_fn_using_stop()
Browse[1]> ls(all.names=TRUE)
[1] "result"
Browse[1]> result
[1] 42
Browse[1]> 

Enter a frame number, or 0 to exit   

1: outer_fn_using_stop()
2: #1: inner_fn_using_stop()

Selection: 0
> 
> 
> 
> options(error=NULL) # (recover wouldn't work automatically below if trying the same thing, at least in current development environment; just deactivate it to avoid confusion)
> 
> inner_fn_using_abort = function() { result = 42; abort("abort message") }
> outer_fn_using_abort = function() { inner_fn_using_abort() }
> withCallingHandlers(outer_fn_using_abort(), error = function(e) { recover() })

Enter a frame number, or 0 to exit   

1: withCallingHandlers(outer_fn_using_abort(), error = function(e) {
    recov
2: outer_fn_using_abort()
3: #1: inner_fn_using_abort()
4: #1: abort("abort message")
5: signal_abort(cnd, .file)
6: signalCondition(cnd)
7: (function (e) 
{
    recover()
})(list(message = "abort message", trace = l

Selection: 3
Called from: signalCondition(cnd)
Browse[1]> ls(all.names=TRUE)
[1] "result"
Browse[1]> result
[1] 42
Browse[1]> 

Enter a frame number, or 0 to exit   

1: withCallingHandlers(outer_fn_using_abort(), error = function(e) {
    recov
2: outer_fn_using_abort()
3: #1: inner_fn_using_abort()
4: #1: abort("abort message")
5: signal_abort(cnd, .file)
6: signalCondition(cnd)
7: (function (e) 
{
    recover()
})(list(message = "abort message", trace = l

Selection: 0
Error in `inner_fn_using_abort()`:
! abort message
Run `rlang::last_error()` to see where the error occurred.
> 
> 
> 
> 
> sessionInfo()
R version 4.0.5 (2021-03-31)
Platform: x86_64-redhat-linux-gnu (64-bit)
Running under: Fedora 34 (Workstation Edition)

Matrix products: default
BLAS/LAPACK: /usr/lib64/libflexiblas.so.3.1

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] rlang_1.0.2

loaded via a namespace (and not attached):
[1] compiler_4.0.5 cli_3.3.0      glue_1.6.2     crayon_1.5.1  

Some more investigation is likely required when dealing with

• rlang condition chains (using parent args)
• noninteractive sessions where we may want to dump and use debugger afterward
• interaction with parallel::mclapply, bettermc::mclapply, furrr::future_*map*

[brookslogan]

Current work left:

• Check exactly how we want text wrapping to work, make sure cli interacts reasonably if user has it, make sure tutorial and errors are readable under these settings.
• Check debug frames functionality
• Check interaction with parallelism libraries
• Add tests

[brookslogan]

(@dsweber2 the call = change might be nested in this ancient PR WIP though it's simple to reimplement... Anything else from here you have been wishing for?)

[dsweber2]

hard to answer, there's a lot in here. I guess we just hadn't gotten to this in the triage? Is the list of todos in the actual PR the only thing that's preventing this from getting merged?