Custom comparison function for `expect_snapshot_file()`

[nbenn]

I have been struggling a bit with trying to get some tests on plot outputs to pass on gh action runs. With a bit of trial and error, I believe I managed to get both my local machine and the actions VM produce images that are identical on a pixel level, however non-identical wrt the current comparison function

compare_file_binary <- function(old, new) {
  old <- brio::read_file_raw(old)
  new <- brio::read_file_raw(new)
  identical(old, new)
}

A potentially worthwhile addition to the expect_snapshot_file() API could be an argument for specifying a custom comparison function. For my use-case for example

compare_img_data <- function(old, new) {
  old <- magick::image_read(old)
  new <- magick::image_read(new)
  identical(magick::image_data(new), magick::image_data(old))
}

would do the trick. This could make expect_snapshot_file() much more powerful than it currently is. For example, again using the magick package, a user could test for image similarity by thresholding the result of magick::image_compare_dist(), etc.

My suggestion would be to add a minor change to expect_snapshot_file(), something along the lines of nbenn@9f91064. I'm happy to expand this to a proper PR if this is of interest to others as well.

[lionel-]

This suggestion makes sense but I would perhaps deprecate the binary argument and export the comparison functions instead. Otherwise binary is ignored when compare is supplied, which is not ideal.

[nbenn]

Sure, even better! I made my suggestion to not break with the current API. But given that this is clearly marked experimental, this makes total sense.

Thanks for considering my suggestion.

Let me know if you want me to change and submit a PR.

[hadley]

Yeah, that sounds reasonable to me. It'd be worth a quick github search or expect_snapshot_file() just to double check that it isn't going to affect a bunch of existing code (which I doubt).

[lionel-]

These packages are using binary = : rmarkdown, vmr, dm, pkglite, reportfactory. I'll deprecate the argument so they can update whenever they are working on a new release.

@nbenn I have cherry-picked your commit in #1400 so I can go forward with the release. Thanks for this!

[nbenn]

Thanks!