Int64 overflow checks

Int64 overflow checks

[Raj B]

Hi

I am writing an implementation of the Python programming language in  
OCaml and ran into an interesting issue.

Python allows the programmer to implicitly perform arbitrary-sized  
integer operations by switching internally between its 'int' and  
'long' types. (which seems to translate to OCaml's int64 and BigInt).

I found an OCaml library on a mailing list which checks for overflow  
in 'normal' 32-bit integer operations. How can I check for overflows  
in int64 operations so I can switch to big-int if that happens?


Thanks!
Raj

Re: [Caml-list] Int64 overflow checks

[skaller]

On Mon, 2007-04-30 at 12:24 -0500, Raj B wrote:
> Hi
> 
> I am writing an implementation of the Python programming language in  
> OCaml and ran into an interesting issue.
> 
> Python allows the programmer to implicitly perform arbitrary-sized  
> integer operations by switching internally between its 'int' and  
> 'long' types. (which seems to translate to OCaml's int64 and BigInt).
> 
> I found an OCaml library on a mailing list which checks for overflow  
> in 'normal' 32-bit integer operations. How can I check for overflows  
> in int64 operations so I can switch to big-int if that happens?

you use basic maths. For example on addition, you expect 

	a + b > a
	a + b > b

and one of these will be false if you got overflow
(something like that).


-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net

Re: [Caml-list] Int64 overflow checks

[Jon Harrop]

On Monday 30 April 2007 18:24, Raj B wrote:
> Hi
>
> I am writing an implementation of the Python programming language in
> OCaml and ran into an interesting issue.
>
> Python allows the programmer to implicitly perform arbitrary-sized
> integer operations by switching internally between its 'int' and
> 'long' types. (which seems to translate to OCaml's int64 and BigInt).
>
> I found an OCaml library on a mailing list which checks for overflow
> in 'normal' 32-bit integer operations. How can I check for overflows
> in int64 operations so I can switch to big-int if that happens?

Sounds like a premature optimization. Just use Big_int for everything to start 
with...

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
The F#.NET Journal
http://www.ffconsultancy.com/products/fsharp_journal/?e