Function parameter order

[Raab, Donald]

You were able to fill in the blanks I left nicely.  :)

I built in the short forms for the most common cases which are of course all Object.  Ref might be a good alternative.  I started out trying Obj, as Object is taken, but hated it quickly.  I wanted consistency, but felt brevity in the Object cases would probably eventually be more appreciated.  There is the possibility of just having Function.ObjectToObject, without the separator which would leave the others able to be specified with the separator.  I hadn't thought of that until now.

I tried some fun combinations like this along the way for the "Object" type references as well:

Function.A.ToB
Function.AB.ToC
Function.ABC.ToD

After trying some other funny variations with _, I threw them all away.  Naming is hard but I think having a full set of tests around the names you're considering helps you get the "feel" of the names.

What I really liked about the ToXYZ forms is that if I hit Ctrl-N in IntelliJ and type ToObject or ToInt, I get the full list of all function types that support those as return types.

Anyway, not sure how helpful this is to the matter at hand, but I wanted to at least get my thought experiments out there.  I certainly appreciate the thoughts and feedback.

From: tpeierls at gmail.com [mailto:tpeierls at gmail.com] On Behalf Of Tim Peierls
Sent: Tuesday, November 13, 2012 11:24 AM
To: Sam Pullara
Cc: Raab, Donald [Tech]; lambda-libs-spec-experts at openjdk.java.net
Subject: Re: Function parameter order

On Mon, Nov 12, 2012 at 10:34 PM, Sam Pullara <spullara at gmail.com<mailto:spullara at gmail.com>> wrote:
This proposal makes a lot of sense to me. Not sure how nice it will look in the APIs that use them but it will be obvious what is going on and matches my expectations on type ordering.

Mine, too, mostly. It took me a little while to understand that T2 is short for ObjectObject, and that Object marks a slot to be specified via the type parameters, in left-to-right order. I was also a little confused initially by the omission of a ToObject when no primitive types are involved.

What would lower the entry barrier, I think, would be to use a name for slots that isn't already a valid actual type parameter. "Ref" would work nicely; it's short and accurate.

So I'd prefer Function.RefRef.ToRef<String, Pattern, Iterable<String>> rather than the shorter Function.T2<String, Pattern, Iterable<String>>.

For very common cases, you could have Function<F, T> extend Function.Ref.ToRef<F, T> (currently Function.T1<F, T>) and Predicate<T> extend Predicate.T1<T>. (I have to confess that my main reason for wanting this is to avoid having to change my existing code. :-)  )

--tim