OpenEuphoria

2. Re: Testing for NULL

• Posted by SDPringle Sep 02, 2015
• 1994 views

The addition of testing EUPHORIA variables as being either not set or not is something that is relatively new. Some of us, like myself, say the user should have a set dummy value as values can be either atoms or sequences.

You can use object rather than sequence or use a custom type.

type sequence_or_0(object x) 
   return equal(x,0) or sequence(x) 
end type 
 
sequence_or_0 x = 0 -- no fatal error here. 

The variables should always be set to something. Here 0 means that there is no string yet.

public function fname(sequence_or_0 x) 
    if compare(x,0) then 
	self[FNAME] = x 
    end if 
    return self[FNAME] 
end function 

Values that might not be atoms (numbers) must not be used or you may get a fatal error. If you use x=0, then you will get a sequence of 0s and 1s depending on how the individual characters compare to 0. You can use compare(), and equal(). These always return atoms. Only atoms are valid, non-zero atoms are true, and 0 is is false.

Shawn

4. Re: Testing for NULL

• Posted by petelomax Sep 02, 2015
• 1961 views

I could show you how to do this in Phix, using a grubby bit of inline assembly, which would need to be duplicated on a per variable basis. Oh, alright, you won't like it, but go on then:

-- WARNING: DO NOT USE, SIMPLY BE UTTERLY HORRIFIED! 
object x 
procedure is_x_assigned() 
    ?object(x) 
end procedure 
is_x_assigned() 
x = "hello" 
-- note: after this point the compiler starts optimising away the "not assigned" handling for x... hence the use of is_x_assigned() 
is_x_assigned() 
#ilASM{ 
    [32] 
        mov edx,[x] 
        mov [x],h4 
        cmp edx,h4 
        jle @f 
            sub dword[ebx+edx*4-8],1 
    [64] 
        mov rdx,[x] 
        mov r15,h4 
        mov [x],r15 
        cmp rdx,r15 
        jle @f 
            sub qword[rbx+rdx*8-16],1 
    [] 
            jne @f 
            call :%pDealloc 
      @@: 
       } 
is_x_assigned() 
 
-- output is: 0 1 0 

However, it is obviously far easier just to make x always assigned to, say, {is_valid,actual_value}.

Pete

6. Re: Testing for NULL

• Posted by andi49 Sep 02, 2015
• 1941 views

#!/usr/bin/env eui 
 
sequence x 
-- The next executable line is an error as expected.   
-- But, how can I test/defend against this error in advance? 
-- ie 
-- 
-- if x = ??? then 
--   -- do something 
-- end if 
printf(1, "x is %s", {x}) -- How to test for this error? 
 
-- No problem here.  Works as I expected 
x = "some sequence" 
printf(1, "x is %s\n", {x}) 
 
-- I don't know how to express this in Euphoria either. 
-- What I hoped for was a way to set x to some value  
-- after valid assignment that would cause the error at the  
-- start of the code. 
x = 0 -- This causes a type error.  But how do you unset a variable? 
printf(1, "x is %s\n", {x}) 

public function fname(sequence self, sequence x = 0) 
    if x then 
        self[FNAME] = x 
    end if 
    return self[FNAME] 
end function 

Hi, i'am not really sure about what you are like todo.

But what about this? (Windows code)

-- #!/usr/bin/env eui  
include std/console.e -- for anykey() 
  
object x = 0 
-- x="test" 
-- The next executable line is an error as expected.    
-- But, how can I test/defend against this error in advance?  
-- ie  
--  
-- if x = ??? then  
--   -- do something  
-- end if  
if atom(x) then 
	printf(1, "x is %d\n", {x}) -- How to test for this error? 
else 
	printf(1, "x is %s\n", {x}) -- How to test for this error? 
end if 
  
-- No problem here.  Works as I expected  
x = "some sequence"  
printf(1, "x is %s\n", {x})  
  
-- I don't know how to express this in Euphoria either.  
-- What I hoped for was a way to set x to some value   
-- after valid assignment that would cause the error at the   
-- start of the code.  
x = 0 -- This causes a type error.  But how do you unset a variable?  
-- printf(1, "x is %s\n", {x})  
if atom(x) then 
	printf(1, "x is %d\n", {x}) -- How to test for this error? 
else 
	printf(1, "x is %s\n", {x}) -- How to test for this error? 
end if 
any_key() 
 

8. Re: Testing for NULL

• Posted by ghaberek (admin) Sep 04, 2015
• 1906 views

I propose naming these new types with CamelCase, vis-a-vis the string type and String class in C#.

IMHO, this looks cleaner than adding "_or_0" to the type name.

It would also be helpful to have "is null or empty" functions for these new types.

include std/dll.e -- includes NULL 
include std/types.e -- includes string type 
 
-- 
-- returns 1 if x is NULL 
-- 
public function IsNull( object x ) 
    return equal( x, NULL ) 
end function 
 
-- 
-- returns 1 if length(x) = 0 
-- 
public function IsEmpty( object x ) 
    return sequence( x ) and length( x ) = 0 
end function 
 
-- 
-- returns 1 if a x is NULL or length(x) = 0 
-- 
public function IsNullOrEmpty( object x ) 
    return IsNull( x ) or IsEmpty( x ) 
end function 
 
-- 
-- a sequence type that allows NULL 
-- 
public type Sequence( object x ) 
    return sequence( x ) or equal( x, NULL ) 
end type 
 
-- 
-- a string type that allows NULL 
-- 
public type String( object x ) 
    return string( x ) or equal( x, NULL ) 
end type 

Example:

String myName = NULL  -- valid 
 
if IsNullOrEmpty( myName ) then 
    myName = "John Smith" -- valid 
end if 
 
myName = 3.14159      -- invalid 

-Greg

10. Re: Testing for NULL

• Posted by ghaberek (admin) Sep 04, 2015
• 1864 views
• Last edited Sep 05, 2015

Yeah, that's probably a better idea. C# allows for this with Nullable Types, so "string?" and "string" are two different things (nullable and non-nullable, respectively).

Too bad "?" isn't a valid type name character in Euphoria.

public type String( object x ) 
    return string(x) 
end type 
 
public type StringOrNull( object x ) 
    return string(x) or equal(x, NULL) 
end type 

-Greg

P.S. I'm not trying to compare Euphoria to C#, just using it as an example to show that such conventions are already established in other languages.