Better Handling of Formatting Anonymous Functions that are Piped to?

[emstruong]

Hello,

I was wondering if there air could better format the following piece of code?

correction_factor <- fitted.values(mod) |>
          {\(y_hat) (exp(y_hat) - 1)}() |>
          {\(y_hat_back_transformed) {
              lm(df$deaths ~ y_hat_back_transformed - 1)
            }}() |>
          {\(fitted_model) coef(fitted_model)}() |>
          {\(named_coef) unname(named_coef)}()

I'm running air 0.7.0 and it formats it this way, which I think is arguably harder to read because there's so much clutter separating the informative pieces of information

correction_factor <- fitted.values(mod) |>
  {
    \(y_hat) (exp(y_hat) - 1)
  }() |>
  {
    \(y_hat_back_transformed) {
      lm(df$deaths ~ y_hat_back_transformed - 1)
    }
  }() |>
  {
    \(fitted_model) coef(fitted_model)
  }() |>
  {
    \(named_coef) unname(named_coef)
  }()

[lionel-]

Since this is official R syntax, I guess we could have better support for it. But that's not my favourite part of the base pipe, in terms of legibility.

I'd just create temporary variables personally (they already have a name via the lambda parameter).

[DavisVaughan]

Honestly I don't think we should encourage the pipe being used this way, so I vote that we don't change anything

[emstruong]

Hello,

@lionel- Thank you for the suggestion. Could you please elaborate on what you mean by temporary variables? I'm not sure I understand what you mean... I was hoping to avoid too many variable assignments ( meaning lots of <-)...

@DavisVaughan Yes, I agree that this code is a bit cursed. Although, for my purposes, I think I need to do this because of the way another R package is coded. Furthermore, I think a special 'perk' of this way of doing things is that the pipes let the reader knows what the start and end points of a long non-tidyverse and non-data-processing statistical computation is.

In my mind, the only other three alternatives would be:

• A very long single-line computation, which is great for a mathematical formula, but hard on the eyes in R
• Multiple assignments to intermediate computational variables, which obfuscates the start/end points and how the different pieces of information are brought together
• A specially coded function just for this purpose, but the internals of this function still need to do one of the above two.

[lionel-]

@emstruong yep I meant successive <- with well-named variables. Since you already have the names in the lambda parameters, it seems more readable to have a regular <- syntax.

You can also use the magrittr pipe with the more convenient . pronoun.