Bug? long division result casted to int saturates to max in value instead of rolling over to the negative number

[ema01]

It took me a while.

dim a as long = 239939944437253
dim b as int = a / 65536
log(b)

I am primarly a C developer so what i would expect: perform the division as long and then cast the result to the appropriate size so - following c rules - it should become 3661192999 -> 0xDA395F27 -> -633774297, but i don't care about the sign because i'm checking individual bits.
however the result is 2147483647 -> 0x7FFFFFFF aka the maximum value for an integer.

Is this expected? As i said i am primarly a C developer and maybe the C mindset is fooling me in this case. If it's expected, where can i read up about this?

 

[Peter Simpson]

What happens if you declare the Int as a Long???

dim b as int = a / 65536

I'm just wondering...

 

[agraham]

It's a Java narrrowing primitive conversion which is much smarter than that of C which basically does nothing.
Chapter 5. Conversions and Promotions (oracle.com)

The generated Java code looks like this

 //BA.debugLineNum = 40;BA.debugLine="Dim a As Long = 239939944437253";
_a = (long) (239939944437253L);
 //BA.debugLineNum = 41;BA.debugLine="Dim b As Int = a / 65536";
_b = (int) (_a/(double)65536);
 //BA.debugLineNum = 42;BA.debugLine="Log(b)";

Which results in a double (you'll have to ask Erel why the double cast is there) that is narrowed to an int on assignment. My boldening

Otherwise, one of the following two cases must be true: a. The value must be too small and the result is the smallest representable value of type int or b. The value must be too large and the result is the largest representable value of type int.

 

[ema01]

Thanks for the explanation. I had a feeling it would be a java thing.
(The cast to double... i suspect it is to conform to BASIC? doesn't it use floating point for every computation then cast to the appropriate type?)

 

[agraham]

i suspect it is to conform to BASIC?

B4X doesn't try to conform to any other dialect of Basic.

doesn't it use floating point for every computation then cast to the appropriate type?

No, though it seems to occasionally do so. I don't know the rules for this. Its predecessor, Basic4ppc, did do all calculations as doubles but Erel changed this for Basic4Android which became B4A and then the more general B4X.

 

[Erel]

About the implicit cast, in B4X the result of 5 / 2 is 2.5, even when both values are integers.

 

[agraham]

About the implicit cast,

Ah. I've just looked at some code and it seems it is just for division. I didn't realise that before.