POINTER NULL CHECK

[jj2007]

'Len (string)' is neither true nor false, but conversely 'Len (string) = 0' is true or false.

I suspect it's just a habit. In one of my older sources I find 200 instances of IF LEN(some$), only two have IF LEN(some$)>0.
One is

Code: Select all

IF LEN(a$)>0 AND LEN(b$)>0 THEN ...

Actually, omitting the >0 might pose a problem here (and only here) if the Basic dialect treats AND bitwise, e.g. with len(a$)=2 but len(b$)=1:

Code: Select all

IF LEN(a$) AND LEN(b$) THEN ...

where (2 and 1) is false.

[dodicat]

You've lost me here.
true or false are boolean.
len(s) and len(s)>0 are both integers.
Do I miss something?

[Munair]

You've lost me here.
true or false are boolean.
len(s) and len(s)>0 are both integers.
Do I miss something?

IF tests are boolean by definition. They are either TRUE or FALSE, regardless of the values being tested. Strictly speaking, checking IF LEN(string) THEN returns TRUE if the length is anything but NULL. Checking IF LEN(string) > 0 THEN returns TRUE only if the length is greater than NULL. The latter is the correct way to test since string lengths can never be negative.

[dodicat]

-1 and 0 are the two possible returns.

If tests are Boolean (Capital B)
I think they were part of C.Bools thesis.

boolean datatypes are a new freebasic addition.
true/false ... 1/0
Thus my mix up.

[fxm]

In my previous post, true and false are not the intrinsic Boolean constants but only qualifiers for the tested condition (verified and not verified).

[Munair]

-1 and 0 are the two possible returns.

If tests are Boolean (Capital B)
I think they were part of C.Bools thesis.

boolean datatypes are a new freebasic addition.
true/false ... 1/0
Thus my mix up.

TRUE and FALSE are outcomes while data type Boolean is merely a convient way to hold either of them. Remember QuickBASIC where you could define

Code: Select all

TRUE = -1
FALSE = NOT TRUE

Yet, QuickBASIC did not have a Boolean data type.

Since the beginning of the computer age, it was generally accepted that TRUE is anything non-zero, while FALSE is always zero. So testing IF LEN(string) THEN returns TRUE if the outcome is non-zero. It is up to a compiler to translate that to -1 or anyting other than zero. ;)

[jj2007]

So testing IF LEN(string) THEN returns TRUE if the outcome is non-zero. It is up to a compiler to translate that to -1 or anyting other than zero. ;)

As the disassembly above shows, the compiler doesn't bother to translate that to a boolean or integer value; it translates to

Code: Select all

jz skipthis   ; jump if zero for the if len() then case
jle skipthis  ; jump if less or equal for the if len()>0 then case

[Munair]

As the disassembly above shows, the compiler doesn't bother to translate that to a boolean or integer value; it translates to

Code: Select all

jz skipthis   ; jump if zero for the if len() then case
jle skipthis  ; jump if less or equal for the if len()>0 then case

It does when a return value from the evaluation is requested. Here, the ASM instructions test for exactly the opposite. So if you test for LEN(string) (<>0) then JZ jumps if not true. Likewise, when testing for LEN(string) > 0 JLE jumps if not true.

But:

Code: Select all

print 1 > 2 ' output: 0
print 1 > 0 ' output -1

return value requested.

[jj2007]

As the disassembly above shows, the compiler doesn't bother to translate that to a boolean or integer value

It does when a return value from the evaluation is requested.

That is correct (avoiding to write 'true'), but what does it return: An integer or a boolean? Take a little test...

Code: Select all

Dim result as boolean

Print 1<2
Print 1>2

result=(1<2)
print "1<2=", result
result=(1>2)
print "1>2=", result

print true
print false

[fxm]

It always returns an Integer:

Code: Select all

#print typeof(1<2)

INTEGER

[Munair]

It looks like the compiler does an evaluation before producing ASM code. In case of PRINT (1 < 2) the value -1 is returned immediately:

Code: Select all

  mov	edx, 1
  mov	rsi, -1
  mov	edi, 0
  call	fb_PrintLongint@PLT

When trying to evaluate IF TRUE = FALSE THEN it seems clear to the compiler that no operation (NOP) is necessary.

[jj2007]

It always returns an Integer

Yep! Funny, isn't it? Not true or false but rather a number...

[Munair]

It always returns an Integer

Yep! Funny, isn't it? Not true or false but rather a number...

Computers are very stupid. They only recognize on or off: 10011011000101 :)

[fxm]

It always returns an Integer

except all the same if all the terms of the condition are already Boolean.

[D.J.Peters]

Code: Select all

#define CBOOLEAN(expression) iif((expression),1,0)
var BASIC  =         (1>2) shl 1 or         (1=2) shl 2 or         (1<2) shl 3
var C_Code = CBOOLEAN(1>2) shl 1 or CBOOLEAN(1=2) shl 2 or CBOOLEAN(1<2) shl 3 
print BASIC,C_CODE
sleep