[whatwg] number-related feedback

[Ian Hickson]

On Fri, 22 Aug 2008, Shannon wrote:
>
> I was going to suggest the spec define unsigned integers as a value 
> between 1 and MAX_INT inclusive when I realised the spec doesn't define 
> the range of integer values at all. Is this deliberate?

Yes. The range is likely to differ based on the platform used -- today's 
32 bit systems will likely have different ranges than next decade's 128 
bit systems.


> Either way I would recommend making a decision on minimum and maximum 
> integer values an using them consistently. If not I can imagine the 
> rapid adoption of 64-bit systems will cause unexpected errors when the 
> same code is run on older 32-bit systems. There are valid arguments for 
> letting each system use its native integer but if this is the case then 
> perhaps the spec should require MIN_INT and MAX_INT be made available as 
> constants.

ECMAScript does define a range, and the limits of that range are exposed 
to scripts. Are there cases where there are non-script limits that would 
benefit from being exposed? Use cases would be helpful here.


> Also the spec interchanges the terms "non-negative integer", "positive 
> integer" and "unsigned integer". I suggest defining one of these clearly 
> and then using it everywhere.

"non-negative integer" is used to mean zero-or-more. I got rid of all 
cases of "positive integers" I could find. I changed the heading of the 
section on non-negative integers to say "non-negative" instead of 
"unsigned" but now the sections are inconsistently named, which may annoy 
me enough that I'll change it back at some point.


> This is a very minor point but is it necessary to say "valid integer". 
> Given that there appears to be no defined min/max range when is 
> something both an integer and at the same time invalid? Isn't an invalid 
> integer a string?

A valid integer is a string also. I don't understand what you mean.


> Finally I wasn't aware Javascript made a distinction between signed and 
> unsigned integers. Is the distinction really necessary? Can we just make 
> everything signed integers and consistently call the full range 
> "integer" and the positive range "integer greater than 0"?

JavaScript doesn't make the distinction, but other languages do, and 
WebIDL defines behavior for unsigned integers that is different from 
signed integers.

Cheers,
-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'