I believe you are considering (or actively) deprecating the info argument to many expect functions. I understand the value of using NSE for label, but it has a length limitation I'm butting my head against.

Working example: a function that takes an input file, does something to it (silently), and returns a data.frame. Because the processing is non-trivial, I want to do it only once per file, so I capture the return value after testing for silence (ergo expect_silence2). I need to run it twice on each file (diff arguments).

Taking a cue from your tests chapter, I believe a way to do this is something like:

my_complex_function <- function(fn, add1 = FALSE) data.frame(x = seq_len(nchar(fn)) + add1)
expect_silent2 <- function(obj) { expect_silent(o <- obj) ; o ; }
expect_mytests <- function(fn) {
  out0 <- expect_silent2(my_complex_function(fn, add1 = FALSE))
  out1 <- expect_silent2(my_complex_function(fn, add1 = TRUE))
  nr <- function(a, b) nrow(b)
  eval(bquote(expect_lt(nr(.(fn), out0), 10)))
  eval(bquote(expect_lt(nr(.(fn), out1), 10)))
}

Ideally, I'd run it as for (fn in allfiles) expect_mytests(fn), but for the sake of this issue:

expect_mytests(strrep("b", 9))
expect_mytests(strrep("b", 53))
# Error: nr("bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb", out0) is not strictly less than 10. Difference: 43
expect_mytests(strrep("b", 54))
# Error: nr(...) is not strictly less than 10. Difference: 89

First: we've lost the NSE info on the last call (not dependent on options("width")). Filenames can be (and in my use, "are often") over 53 characters, and this complete loss is unexpected. I can strategically truncate the string, though this is prone to "works best for me" syndrome (e.g., first 50? last 50? first/last 25?).

Second: the only way I can imagine providing that file-specificity with the failure messages is to add it as an unused argument to my nr function. This works, but it seems a bit kludgy. Is this the only way to include the filename (in this example) within the failure message?

(Third: is this really the best/only way to get the info stuff into the failure message?)

As I'm writing this ... if I switch the order of the arguments within nr, this goes away:

expect_mytests <- function(fn) {
  out0 <- expect_silent2(my_complex_function(fn, add1 = FALSE))
  out1 <- expect_silent2(my_complex_function(fn, add1 = TRUE))
  nr <- function(a, ...) nrow(a) # arguably better-looking, anyway
  eval(bquote(expect_lt(nr(out0, .(fn)), 10)))
  eval(bquote(expect_lt(nr(out1, .(fn)), 10)))
}
expect_mytests(strrep("b", 99))
# Error: nr(out0, "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb") is not strictly less than 10. Difference: 89

Is this a limitation of eval or bquote, or is this something testthat is imposing? (I could not find "54" ...)

Related issue refs:

• Deprecate label & info arguments #218 Deprecate label & info arguments
• Add info, label arguments to expect_silent #332 Add info, label arguments to expect_silent
• Use conditions for reporting results #360 Use conditions for reporting results
• consider un-deprecating info (and label?) for certain expectations #392 consider un-deprecating info (and label?) for certain expectations
• expect_silent missing info field? #429 expect_silent missing info field?
• expect_{lt,gt} don't have 'info' argument #476 expect_{lt,gt} don't have 'info' argument