search

briandk / granovaGG

Bob Pruzek and Jim Helmreich's implementation of Elemental Graphics for Analysis of Variance
Other
15 stars 4 forks source link

How closely should granova.ds.ggplot formal parameters match those of granova.ds? #7

Closed WilDoane closed 13 years ago

WilDoane commented 13 years ago

This is the public API. Anyone using the package already would have to change all their calls to granova.ds to use your new formal parameter names.

briandk commented 13 years ago

I only see two options:

  1. Make my signature match the original, which propagates confusing variable names like ne and sw
  2. Force all users to redo their code

What am I missing?

WilDoane commented 13 years ago

Option 2 tends not to be an option in programming.

Option 1 + an adapter would be my recommendation: accept "ne", but then

northeastShift <- ne

early in the code, to adapt the input parameters to your more expressive naming convention.

briandk commented 13 years ago

The adapter proposal is good, but it still means "ne" appears in the signature. (It was something I was thinking of as a option 3, but it seemed identical to option 1 in that when calling the function the user sees no difference from the old granova: the adapter is within the body of the function.

WilDoane commented 13 years ago

That's the price of exposing a public API. You never want to break code that people have already created.

WilDoane commented 13 years ago

And, yes, I do worry about the public API. It's hard not to. Anything public runs the risk of not being able to be changed in future releases.

briandk commented 13 years ago

Just for the record here, if granova.ds.ggplot() becomes the new default granova.ds(), then:

I think a third option might be to match on as many parameters from the old API as we can, then use . . . in the function signature as a way to handle any non-matched parameters, even if we might discard them in the function body.

So, a signature like:

granova.ds.ggplot <- function( data                      = NULL, 
                               plotTitle                 = "Dependent Sample Scatterplot",
                               conf.level                = 0.95,
                               plotTheme                 = theme_granova()
                     )

Means you could call our function with a legacy parameter (like jj):

granova.ds.ggplot(data = rat, jj = 5)

And the function would take the jj parameter, but just never use it and otherwise execute normally.