OpenEuphoria

2. Re: Create new socket

• Posted by jeremy (admin) Aug 25, 2009
• 1003 views

The docs are wrong. Originally it returned an atom, but that was later changed. The first element of that sequence is actually the socket, the second is a pointer to the sockets address. The internals, however, shouldn't need to be of a concern and may change in the future w/o notice. The important thing to know is that on failure an atom is returned giving the error code, otherwise, it's a valid socket object.

That socket object can be passed to any of the socket functions. They will use the "sequence" correctly.

Jeremy

4. Re: Create new socket

• Posted by euphoric (admin) Aug 25, 2009
• 940 views

Why not return a sequence on FAIL as well, such as {-1,0}?

6. Re: Create new socket

• Posted by euphoric (admin) Aug 25, 2009
• 939 views

Because then I don't have to use the ambiguous "object" type identifier.

Instead of

x = thefunc() 
 
if not atom(x) then... 

I can use

x = thefunc() 
 
if x[1] > 0 then... 

This assumes that a sequence type is more efficient than using an object type.

But it also just seems more logical and reasonable to return a consistent result.

Throughout BBCMF, function results are returned as follows:

{ 1, result } for success, where result can then be in any format
{ 0, result } for fail, where result is usually an error message or code

if result[1] then 
  -- process result[2] 
else 
  -- display result[2] 
end if 

It has made things quite efficient, especially during development.

The primary thing is, I get rid of an ambiguous and inefficient "object" type declaration.

8. Re: Create new socket

• Posted by euphoric (admin) Aug 25, 2009
• 934 views

socket s = create() 
if atom(s) then 
    printf(1, "Could not create socket, error number: %d\n", { s }) 
end if 

Or:

socket s = create() 
if not valid(s) then -- a sockets.e function to validate your socket* 
   printf(1, "Error %d:%s", { get_error_num( s ), get_error_msg( s ) }) -- sockets.e error management 
-- or 
   printf(1, "Error %d:%s", get_error(s) ) -- or get_error() could return {errnum,errmsg} 
-- or 
   printf(1, "Error %d:%s", s[2] ) -- on fail, s is {0,{errnum,errmsg}} 
else 
   -- process normally 
end if 

-- in sockets.e 
public function valid(object s) 
   return sequence(s) and length(s) > 0 and s[1] -- hopefully this short-circuits...  
end function 

10. Re: Create new socket

• Posted by AndyDrummond Aug 26, 2009
• 913 views

One good reason is that if you forget to check for errors, and the error return is a sequence, then the program could go on quite a way before the error gets discovered, and the failure would be totally confusing. If the error return is an atom, then the first time the error value was used as if valid then it would fail and you would quickly see why. Then you would add your proper test for error at the right place! IMHO, data which is in error should not be usable as if it was correct.

If you read data from a file, you get a sequence - but at EoF you get an atom. Easily identified, but impossible to use as file data. This is a good and safe standard technique.

Andy

12. Re: Create new socket

• Posted by AndyDrummond Aug 26, 2009
• 914 views

I blush with pleasure ...