Standard evaluation version of functions

[richierocks]

Standard evaluation is often better for when you want to include code in your own packages. Do you plan to include standard evaluation versions of the functions?

I'm thinking of the dplyr style, for example filter vs. filter_. So you could include add_variable_, set_obejctive_, etc.

I think that most of the logic is handled by the lazyeval package, so it shouldn't be too tricky to implement.

[dirkschumacher]

Sounds like a good idea. How would that api look exactly? Pass in a lazyeval object or a character string?

[richierocks]

I think you want something like this:

setGeneric("set_objective", function(model, expression, direction = "max") {
  set_objective_(model, expression = lazyeval::lazy(expression), direction = direction)
}

setGeneric("set_objective_", function(model, expression, direction = "max") {
  # As the current set_objective fn, but without calls to substitute
}

expression can be passed to set_objective_ as a formula, string, or quoted expression.

BTW, if direction can only be "min" or "max", then you can check it by changing the signature to

function(model, expression, direction = c("max", "min"))

then calling direction <- match.arg(direction) in the function body.

[dirkschumacher]

Cool thanks. Will look into this. Indeed, maybe I should use lazyeval instead of base R's eval/substitute constructs.

[dirkschumacher]

@richierocks Added a first version. If you have any comments on the code or other ideas let me know

[dirkschumacher]

I close the issue for now.

[dirkschumacher]

Still needs a solution for variable and constraint quantifiers