On Wed, 11 Sep 2019 at 15:10, ducasse <stepharo@netcourrier.com> wrote:

> On 11 Sep 2019, at 04:07, James Foster <Smalltalk@JGFoster.net> wrote:
>
> Would use of ? and ! in unary/keyword selectors be convention or somehow required? If simply convention, then we should start with renaming testing methods to be named is* or has*.
>�� flag1 := anInteger even.�� not good"

Agreed.�� That could "almost" be construed�� as converting a 3 to a 2 or 4.

��

>�� flag2 := anInteger isEven.�� better"

Agreed. It reads well.

��

>�� flag3 := anInteger even?.�� how much better?��

For me, this doesn't read as well as flag2, but even though there is some redundancy, for me a combination reads well...

�� flag3a := anInteger isEven?

Perhaps if "?"==>Boolean was a strong convention then there could be a check when the value is returned rather than when it is used (or would that complicate other things?)��

��

>�� flag4 := #(1 2 3) includes?: 2. ��how much better?��

My first impression is "yuck!", but then I think... "maybe" if there was a definite benefit (i.e. optimization) from strong guarantees about the return value being Boolean.

��

I think that I would use ? mainly for unary message

Now I��m sure that if you look carefully some people use

�� include
�� for the action
�� includes
�� for the tests

I took include as an example and this is super not intention revealing.

>> lineUpBlockBrackets

�� lineUpBlockBrackets?
�� Now I will rewrite them all as shouldLineUpBlockBrackets or isLineUpBlockBrackets and to me for unary message ? makes it a lot better.

btw, some alternatives...��

�� doBlockBracketsLineUp�� - reads well but "do" is already a loaded word

�� areBlockBracketsLinedUp - reads well with the past-tense phrasing

cheers -ben