Update rlang use

[michaelquinn32]

This PR standardizes calls to rlang in the package. It

• removes the need to use lazyeval
• fixes uses of rlang that weren't complete

The most notable change is that verify now works within other functions. This lets users wrap it in their own helper, (for example) setting the default error or success function.

[tonyfischetti]

Thank you very much for this :)

[tonyfischetti]

The tests are now not passing on my machine. For example...

test-assertions.R:755: failure: all assertions work with .data pronoun without chains
`assert(test.df, within_bounds(-Inf, 2), .data$y, error_fun = just.show.error)` did not throw an error.

Do you know why this is?

[michaelquinn32]

Sorry about that. It looks like dplyr/rlang won't error if it doesn't find a column when using the masking pronoun from that test. It just returns what it can.

test_fun <- function(...) {
   columns <- rlang::quos(...)
   dplyr::select(iris, !!!columns)
}

test_fun(.data$not_here)
# data frame with 0 columns and 150 rows

I can fix this. There needs to be a check that subframe has columns. I'll send it soon.