Re: [X10-core] RFC: Integer literals

[Josh M. <jos...@an...>]

I realise that the case is closed and the jury has already rendered its 
verdict,  but is there a reason we can't use 'z' (= integer) rather than 
'n' (= natural number, non-negative integer) for a signed 32-bit int?

------------------------------------------
Josh Milthorpe

Postdoctoral Fellow, Research School of Computer Science
Australian National University, Building 108
Canberra, ACT 0200
Australia
Phone:  + 61 (0)2 61254478
Mobile: + 61 (0)407 940743
E-mail: jos...@an...
Web:    http://cs.anu.edu.au/~Josh.Milthorpe/

On 03/08/13 00:07, Jonathan Brezin wrote:
>
> Dave et al,
>
> As an application programmer mostly working in very weakly typed 
> languages like JavaScript, I was very sceptical when I first started 
> working in X10 -- how long ago?  5 or 6 years?  It took a while, but I 
> have become a real fan of being forced to say explicitly what each 
> identifier is, either in terms of the one value it will ever have, or 
> by writing out an honest, all i's dotted type expression.  This causes 
> me to be more sympathetic to the "no default" for literals than I 
> might otherwise be.
>
> Looking back to my early computing days--C and Bell Labs in the late 
> 1970's and early 80's--C's choice of being glib about what size things 
> were ("int" meant "the natural size integer for the machine in 
> question") led to no end of #ifdefs as machines made the leap from 16 
> to 32 bit integers, which led to code very easy to misread.  C's 
> semantics for numeric operators are still enormously complicated, as 
> anyone who has looked at the literature on random number generators 
> will agree, where that forest of #ifdefs rears its ugly head for the 
> sake only of machine independence.   Goodbye int, hello int32.
>
> So here we are again.  32 to 64 bits this time, but so what?  What was 
> "natural" is a year ago is no longer "natural" now.  The only 
> difference between our situation and C's is the relatively modest 
> amount of legacy code we are saddled with, and it's only literals that 
> need tweaking.  So bottom line:
>
> I think that "n" is better than "i" precisely because mathematicians 
> for two hundred years have used "i" differently, and "n" is a choice 
> that is mathematically in good taste (natural numbers).  I also think 
> (and few will agree with me) that numeric precision should always be 
> made explicit.  I am not even all that pleased with float versus 
> double:  how long until long doubles are the precision of choice for 
> HPC numeric problems, and are we going to go through this all over again?
>
> Jonathan
>
>
> Jonathan Brezin
> Research Staff Member
> IBM TJ Watson Research Labs
> 1101 KITCHAWAN RD
> ROUTE 134 / PO BOX 218
> YORKTOWN HEIGHTS NY 10598
>
> Phone: (914) 784-6728 (Tie: 863-6728)
>