Add evaluating environment to tidy eval call. Closes #2835

[clauswilke]

This PR adds the calling environment to the transformation evaluation in the secondary axis. This closes #2835.

[lionel-]

Can you add a unit test please? This would be a good start:

n <- 100
p <- ggplot() + scale_y_continuous(sec_axis(trans = ~ . / n))
expect_error(regexp = NA, ggplot_build(p))

No need to wrap in local() because all testthat blocks are local. It'd be even better to test that the axis has been transformed correctly.

Edit: Or would it be good practice to wrap in local() anyway, so you can evaluate line by line interactively?

[lionel-]

@hadley I think we should provide a way of supporting quasiquotation for this type of transformation. There's a bug in @IndrajeetPatil's function: whenever the data frame contains a column total_obs, the column has precedence over the environment. Quasiquotation would help solving the precedence.

We could support quosures:

sec_axis(trans = quo(. / !!total_obs))

But that feels like a strange UI. It might be better to provide a formula constructor in rlang:

sec_axis(trans = f(. / !!total_obs))

Which could also be useful for statistical models.

[clauswilke]

@lionel- Thanks, I will add the test case you suggest and also think about making one that tests scaling. Does the proposed fix to the bug look correct to you?

[lionel-]

Does the proposed fix to the bug look correct to you?

Yup, but if we'd like to support quosures it'll need a slightly different approach. If we go for a formula constructor with qq support then there's no further change needed.

[clauswilke]

There was a unit test already that I could adapt with very minor change.

@lionel- Does this change look good to you or do we need another one?

@dpseidel I think you wrote that test function. Could you look over my modification also?

[lionel-]

Looks good!

[dpseidel]

Yep! Your modification shouldn't change the original intent, looks good to me.

[clauswilke]

Great, thanks! If somebody can approve I'll merge.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/