New handling of expressions in data_modify(), introducing as_expr() helper function

[strengejacke]

library(datawizard)

data_modify(iris, as_expression("sepwid = 2 * Sepal.Width")) |> head()

data_modify(
    iris,
    as_expression(c("sepwid = 2 * Sepal.Width", "seplen = 5 * Sepal.Length"))
) |> head()

data_modify(iris, sepwid = as_expression("2 * Sepal.Width")) |> head()

data_modify(
    iris,
    sepwid = as_expression("2 * Sepal.Width"),
    seplen = {"5 * Sepal.Length"}
) |> head()

e <- "sepwid = 2 * Sepal.Width"
data_modify(iris, as_expression(e)) |> head()

e <- c("sepwid = 2 * Sepal.Width", "seplen = 5 * Sepal.Length")
data_modify(iris, as_expression(e)) |> head()

e <- "2 * Sepal.Width"
data_modify(iris, sepwid = as_expression(e)) |> head()

e <- "2 * Sepal.Width"
data_modify(iris, sepwid = as_expression(e), seplen = 5 * Sepal.Length) |> head()

e <- "2 * Sepal.Width"
f <- "half_petal = 0.5 * Petal.Length"
a <- "string"
num <- 1:5
data_modify(
    iris,
    sepwid = as_expression(e),
    seplen = 5 * Sepal.Length,
    {f},
    new_var = a,
    new_num = num,
    new_var2 = "ho",
    new_num2 = 4:6,
    Sepal.Length = NULL,
    Petal.Length = NULL,
    Sepal.Width = NULL,
    Petal.Width = NULL
) |> head()


d <- data.frame()
for (param in letters[c(1, 2, 5)]) {
  out <- data.frame(x = as.numeric(as.factor(param)))
  out <- data_modify(out, Parameter = param)
  d <- rbind(out, d)
}
d

data_modify(iris, sepwid = "2 * Sepal.Widht") |> head()

# errors

e <- "sepwid = 2 * Sepal.Widht"
data_modify(iris, as_expression(e))

data_modify(iris, as_expression("sepwid = 2 * Sepal.Widht"))

data_modify(iris, sepwid = 2 * Sepal.Widht)

[strengejacke]

Thanks for the review! All comments are resolved, hence now we have additional features included. Tests should be added for all cases.

[etiennebacher]

Thanks, almost there for me. The implementation is quite complex but the test suite looks complete and I like the fact that one can still build and evaluate expressions from strings.

@DominiqueMakowski does the current behavior look ok to you?

[strengejacke]

I guess we're going in this direction now, but let's keep it for another PR, this one is already very dense.

Maybe let's wait a bit. In other functions, by is unique, but here it could be a new variable name, conflicting with the by argument. I guess we need to think more about this for data_modify().

[strengejacke]

Ok, I think recent comments are also resolved.

[etiennebacher]

Alright, thanks for addressing all the comments, I'm happier with the current implementation than the old one. We have 3 weeks to fix CRAN errors so let's merge this now and revisit later if some things must be improved.

[strengejacke]

CRAN errors should already be fixed, I think. You can submit any time, but I think you had some reminders/todos in #615 before submission.

Let me wait for checks, then I'll merge.,