[J3] [EXTERNAL] Re: Why is += missing?

[Clune, Thomas L. (GSFC-6101)]

Ondrej,

I pretty much agree with the metric you introduce here, but with a couple of caveats.

1.  The "number of uses"  may not be a linear factor.     For the moment I think it is fair though.
2. "how much it saves" is very subjective.     

Note:  I suspect the opportunities for the ? case with N=2 options is probably not very rare, but the savings are also not particularly strong in that case.   And you may not ever have had the N=4 case like Van.   I've only had that 2-3 times in my career.   But the workaround without ? is dreadful and arguably quite difficult to understand/maintain.    And I'm sure there are (rare) cases out there with N > 4.      A person could very easily oppose the feature on those grounds.     As a former president once said though,  votes matter.    WG5 decided that this feature should be pursued.

Cheers,

- Tom



On 9/2/21, 4:28 PM, "Ondřej Čertík via J3" <j3 at mailman.j3-fortran.org> wrote:

    Hi Malcolm,

    On Thu, Sep 2, 2021, at 1:10 AM, Malcolm Cohen via J3 wrote:
    > Ondřej Čertík writes:
    > > Given the number of lines that could use +=, it seems the number of lines that could use ? is orders of magnitude less, or is that not the case?
    >  
    > Others have replied to this, but anyway:
    >  
    > Conditional expressions turn multiple statements into one expression.
    > As Van said, the alternatives to op= are trivial.
    > They are so far apart as to be not comparable in functionality.
    >  
    > Some might even disagree that “var op=expr” is clearer than “var=var op expr”.

    It's a good point that the alternative to "+=" is simpler than the alternative to "?". Van made the same point.

    So a better metric might be: (number of uses) * (how much it saves)

    "+=" saves less, but is used a lot.
    "?" saves more, but is used less.

    In one of my codes I ran the "grep" command from the thread above and obtained ~600 cases where it could be used, consistent with others who posted. I can't think of a single case where I would use "?", so for me the metric above clearly wins for "+=".

    I have no doubts that others would find more uses for "?", so perhaps "?" can win in the above metric for them.

    >  
    > > The new ? operator is exactly like in C. As a committee we have decided that was the best approach from all the alternatives.
    >  
    > Did I not mention how much I disliked the syntax the committee voted 
    > for? Not sufficiently to change my vote, but I was quite disappointed.

    Fair enough.

    >  
    > > I would encourage us to be open minded about adding +=.
    > 
    > I try to be open-minded about all new features, with the caveats
    >  1. All new features start with minus 100 points.

    I do the same.

    >  2. I try to think of as many obvious problems as I can at the start, 
    > so we’re not surprised by them later. (Even so, I frequently miss 
    > problems that ought to have been obvious!)
    >  
    > Fortunately, there’s loads of time for people to think about this 
    > particular suggestion before we get to formal submission of a proposal.

    Thank you,
    Ondrej