[KissLick]

I have another question, will you "empty" handle variable on closing handle in new API?

like:

PHP Code:

new Handle:hndl = INVALID_HANDLE;
hndl = CreateArray();

/*
.
.
.
*/

CloseHandle(hndl); // hndl = INVALID_HANDLE 


Or at least add native to check handle validity like

PHP Code:

bool:IsHandleValid(Handle:hndl) 


[BAILOPAN]

You can use delete on a null handle. For example,

PHP Code:

ArrayList array = CreateArray();
...
delete array; 


You can use CloseHandle on a null handle as well, but new-syntax prefers delete.

[friagram]

Are we keeping the requirement for array declaration? It becomes pointless after 3 anyways
[] [][] [][][]

Isn't
new foo[2][2] essentially the same as new foo[2*2]
or new foo[3][1] and new foo[3]

[BAILOPAN]

@friagram: No, they are quite different. int[2][2] is a T[2] where T=int[2]. An int[2*2] is a T[4] where T=int.

It is true that you can use N*M cells to simulate coordinates of a 2d-array, but that does not make it an array of arrays, since arrays are a special builtin type.

[friagram]

Aye

Oh, there are some weird preprocessor bugs, and also some dumb ones.

Not sure if this should be bothered with or not but a

#define HAMMER HAMMER Will stall the compiler forever

Also, if/elseif/endif skips the loose indentation warnings. Not sure if it is intentional.

Also, types in switches, is there a way around this double colon case problem? If I have something of an enum type that I cast like Blahtype:84 it won't play nice in a switch, i have to use the int without the tag. I suppose the new system will help with this.

[BAILOPAN]

@friagram - the preprocessor will be phased out over time, so I'm not as concerned about bugs in it (the recursion bug is known).

You can put the case in parenthesis, like this:

  case (BlahType:84):

Pawn should be taking the type of the switch() and checking it against each case. (If it's not, I can fix that for 1.7.) So, if you have to write code like above, it would seem that the initial switch() expression is incorrectly typed.

[friagram]

Iirc the tf conds cause problems sometimes, like if i do

Switch(condition)

case TFCond_Taunting:
case TFCond:5:

So, if parenthesis around it works, that's fine. I could just use a 5 without the type, but its not always convenient for readability. Usually I just use the preprocessor to give the variable a friendly name and go back and fix it down the road I suppose I could make a static const of that type and name it something decent until the new conds get added in a later update, like static const TFCond:TFCond_Ubercharged = 5

[psychonic]

Originally Posted by friagram Iirc the tf conds cause problems sometimes, like if i do Switch(condition) case TFCond_Taunting: case TFCond:5: So, if parenthesis around it works, that's fine. I could just use a 5 without the type, but its not always convenient for readability. Usually I just use the preprocessor to give the variable a friendly name and go back and fix it down the road I suppose I could make a static const of that type and name it something decent until the new conds get added in a later update, like static const TFCond:TFCond_Ubercharged = 5

Why wouldn't you just always use the named value anyway rather than tagging the bare integer?

[nergal]

Question on the new Syntax+API.

Is there Classes and/or Data Structures?
If not, will there be in the future?

[BAILOPAN]

The first post answers this question. The syntax is a transitional, syntactic transformation. It does not introduce any new semantics. It's intended to pave the way for cleaning up code to make newer semantics possible.