OpenEuphoria

2. Re: Request: Integer Division

• Posted by jimcbrown (admin) Jan 02, 2013
• 1436 views

This really should be in a Request-for-Feature ticket, but...

integer Y = 3
integer Z = 2 
integer X = Y \ Z

integer X = trunc(X/Z)

integer X = 0
while Y >= Z do 
X += 1 
Y -= Z 
end while

integer Y = 3
integer Z = 2 
atom YY = Y 
atom ZZ = Z 
atom XX = YY / ZZ 
integer X = trunc(XX)

I think the syntax is the most contriversal part here.

Adding a new builtin or machine func, like so:

integer X = integer_division(Y,Z) 

which is equivalent to

integer X = 0 
while Y >= Z do 
X += 1 
Y -= Z 
end while 

but implemented using integer division at the assembler level is pretty easy to do, and it's certainly a lot faster and more efficient than using intdiv.

I think however, changing the syntax such that

integer X = Y \ Z 

does integer division, may break backwards compatibility and could be confusing, considering the way that the current and older versions of Euphoria handled this.

4. Re: Request: Integer Division

• Posted by jimcbrown (admin) Jan 02, 2013
• 1382 views

Here: http://openeuphoria.org/ticket/index.wc?per_page=20&page=1&type_id=2&category_id=-1&milestone=&severity_id=-1&status_id=-1&user_id=-1&actiontype=New+Ticket

I did.

Nope. It was just a simple typo. (Unlike a similar example in one of your recent posts, http://openeuphoria.org/forum/m/120272.wc - which I consider hypocritical considering your reaction in http://openeuphoria.org/forum/m/117914.wc )

integer X = integer_division(Y,Z) 

The point of adding a new integer_division() would be to take less cpu time to set up and execute (as compared to the existing intdiv() in std/math.e which is slower as it involves conversion to and operations with floating point numbers).

I wanted to just use the name intdiv() (since that's less typing than integer_division) which saves human time too - but that was already taken. I'm open to alternatives, though. idiv() perhaps?

6. Re: Request: Integer Division

• Posted by jimcbrown (admin) Jan 02, 2013
• 1369 views

Ah, I didn't think about using the backslash as the low-level fast intdiv operator. I think this is the best approach.

8. Re: Request: Integer Division

• Posted by jimcbrown (admin) Jan 02, 2013
• 1371 views

integer Y = 3
integer Z = 2 
integer X = Y \ Z

integer X = trunc(X/Z)

Doh! I read that as the forward slash.

10. Re: Request: Integer Division

• Posted by CoJaBo2 Jan 02, 2013
• 1374 views

Its a pretty standard term http://en.wikipedia.org/wiki/Ticket

You complain on almost every thread here that "everyone is picking a fight with you". How, exactly, is this not you picking a fight with everyone else?

Note that this thread is about adding an operator for integer division. It is not about the meaning of the word "ticket". It is not about the spelling of the word "controversial". It is not a historical survey of every past flamewar on this forum.

This kind of nonsense, that unfailingly pops up on every single thread, is undoubtedly driving away newcomers, who ask a simple question only to see that 90% of the thread becomes an argument on whether it KB means kilobits or was a typo for megabits or something equally pointless.

Can we drop this now? IMVHO, the mods should start simply deleting all off-topic posts (like this one) instead of replying to them and dragging the thread further and further off topic.

12. Re: Request: Integer Division

• Posted by mattlewis (admin) Jan 02, 2013
• 1345 views

Actually, there are (and have been for as long as I'm aware) 4 division opcodes:

• DIVIDE
• DIV2
• FLOOR_DIV
• FLOOR_DIV2

The DIV2 versions attempt to use bitwise optimizations. The FLOOR versions use C integer division if the divisor and dividend are euphoria integer values (we may not know until runtime). One potential difference with explicit integer division is how we would deal with fractional values.

I can think of two ways to deal with this:

1. floor( a / b )
2. floor( floor(a) / floor(b) )

The second option would be similar to what you'd get in, say, C, where you'd do casts to int instead of calling floor() in order to invoke integer division on FP values.

Matt

14. Re: Request: Integer Division

• Posted by mattlewis (admin) Jan 02, 2013
• 1329 views

If you declare something to be an integer, then we do know. But you may instead be using an atom. Or a value stored in a sequence, or the return value from a function. Or a sequence full of...well, who knows?

In general, declaring things that you know will always be integers to be integers does offer speed benefits, and I do this as much as possible, too. There is a lot of stuff in both the interpreter and translator to optimize this sort of thing. However, other cases where we can't determine the storage type at parse / compile time have to be dealt with, too.

Matt

16. Re: Request: Integer Division

• Posted by jimcbrown (admin) Jan 03, 2013
• 1286 views

No. Try out the following code:

integer x, y, z 
y = 1 
z = 2 
x = y/z 
? x 

I'm not sure if this is explicitly mentioned anywhere, but it has been this way since Euphoria 1.5 (if not earlier).

18. Re: Request: Integer Division

• Posted by jaygade Jan 03, 2013
• 1213 views

Ah, fair enough.

I'm neither in favor nor opposed to further increasing Euphoria's complexity by introducing a new operator.

Since I come from the old-school BASICs, I never had much problem with typing in

100 x = int(y/z) 

Or the Euphoria equivalent. I can only say that explicitly specifying a function makes it clear to the interpreter and the reader that the intent is to do integer division. In that respect, it is preferable to the way that other languages handle this case.

20. Re: Request: Integer Division

• Posted by jaygade Jan 04, 2013
• 1219 views

DerekParnell said...

Have you ever coded in COBOL?

   DIVIDE y BY z WITH ROUNDING GIVING x. 

No function calls and no operators

Heh, no, but I remember seeing it in my Dad's college textbooks.

The nightmares didn't last more than a couple of nights.