Zugg Software :: View topic

Posted: Wed Sep 27, 2006 1:51 am

In C you can do fancy stuff in IF statements such as variable assignment. I have always found this really hard to read and support over time. Doing:

GURU Joined: 23 Dec 2000 Posts: 4834 Location: USA

This would make things a lot simpler. One thing that might prove to be a minor issue is mixing #IF and #WHILE on the same nest level:

#while (stuff) {
#if (other stuff) {#if (yet more stuff) {}} {unrelated stuff}
#forall (still more stuff) {}
}

Would the "other stuff" #IF have the same variable as the #FORALL (%j), or would they increment (%j for the #IF, maybe %k for the second #IF, and %L for the #FORALL)?

Wizard Joined: 15 Jan 2002 Posts: 1644 Location: USA

*drool* I think you just made my day!

f I understand correctly I should also be able to do something like this?

Posted: Wed Sep 27, 2006 3:04 am

Hmm I'd say no... but thats from the description he has given so far.

The %i relates only to the result of the bracketed #IF statement... so for that one you just posted it would only come back as true... since the actual result is from the AND.

You Could do...

Posted: Wed Sep 27, 2006 6:16 am

Namsar is correct. And that's a nice trick for it. But yeah, CMUD can't get stuff from "subexpressions", only from the full IF expression.

Matt: mixing IF, WHILE and loops is no problem. Just start at the outer-most block and use %i, then move to inner blocks and increment to %j, %k, etc. They act like local variables within the scope of their respective statements.

Actually, Matt's comment just made me realize that this actually might break some stuff. When a loop is nested within a IF like this:

SubAdmin Joined: 18 Nov 2001 Posts: 5182

How about using ^i..^n. Right now the caret isn't used for anything that I can think of.

Wizard Joined: 14 Aug 2004 Posts: 1269

Using carets just for this might seem like another wierdness.

How about having something like %exprresult(1), %exprresult(2), ... , %exprresult(n) ?

Or, if it should be a variable and not a function:
%exprresult1, %exprresult2, ... , %exprresult99

Posted: Wed Sep 27, 2006 4:59 pm

I'm leaning towards something like Seb's suggestion. I definitely don't want to add any other special characters. I'd like something that fits into the current language scheme.

However, I also think there might be a way to capture multiple expressions into variables. Like the post that Nexela made. I want to add some code to optimize expression evaluation so that when you do X AND Y then it never evaluates Y if X is already false, and X OR Y where it doesnt evaluate Y if X is already true.

When adding this kind of optimization, it is possible to capture the individual subexpressions (X, Y). So in this case, something like %expresult(1) would refer to X and %expresult(2) would refer to Y, with just %expresult referring to the entire expression.

The problem with this is that you can access the %expresult of any outer statements. However, I think this might actually be OK. After all, if you want to access the value of an outer statement, you can always assign %expresult to a local variable of your choice, and that probably results in a more readable script.

Posted: Wed Sep 27, 2006 6:43 pm

Vijilante,
Wouldn't using ^ break the anchors for triggers then?

GURU Joined: 23 Dec 2000 Posts: 4834 Location: USA

I think that's a different parsing engine, Vitae. Although maybe it might generate a syntax error, using ^ doesn't similarly interfere with trigger anchors.

SubAdmin Joined: 18 Nov 2001 Posts: 5182

I posted during my first cup of coffee in the morning, and reflecting on it I have to say it is much better not to add another special character. I like the idea of treating it as a function or other new naming. It is rather easily documented and very understandable.

Zugg, if you do manage to implement it in such a way that portions of the full expression are available to scripts I would suggest assigning those portions based on the use of parenthesis. Thereby a user that does:
#IF (a=1 and b=2) {
would only have 1 expression result for the entire thing
#IF (a=1 and (b=2)) {
would have results 1:a=1 and (b=2);2:b=2
#IF ((a=1) and (b=2)) {
would have results 1:(a=1) and (b=2);2a=1;3:b=2

This should match the REGEX method for determining returned values from parenthesis.

Posted: Fri Sep 29, 2006 4:02 am

I forgot to remove this "feature" in the 1.08 release. So unfortunately, if you have a Loop within an IF or WHILE statement, then %i is going to return the IF expression rather than the Loop variable. You can temporarily change %i to %j to get this working. I'll fix this new feature in the next version. Sorry I forgot about this :(

Posted: Fri Sep 29, 2006 8:28 pm

There appears to be a somewhat interesting side effect of this, the %i variables take account of all if and loop statements involved above it, which can potentially cause problems. To demonstrate consider the following (utterly contrived) code:

SubAdmin Joined: 18 Nov 2001 Posts: 5182

This seems to indicate a problem with properly removing the references from the script stack, in the current incomplete implementation. I think from the discussion in this thread Zugg will be seperating them and further improving the functionality of this new feature.

Magician Joined: 29 Nov 2005 Posts: 482

didnt zugg create the switch command instead of using if after if after if or nested ifs?

GURU Joined: 23 Dec 2000 Posts: 4834 Location: USA

Only for cases where the same expression was being used. If each #IF had a different expression, then just recently he created the #SWITCH command.

Still, there are always cases where nested or multiple #IFs will have to be used either because the logic is short enough to do so or because there's simply no better way.

Posted: Tue Oct 03, 2006 12:24 am

OK, I have this implemented and working properly in v1.09 now.

Here is what I did: I implemented a new function called %exp. The first argument is the subexpression index that you want to return. Subexpressions start at 0 and get incremented for each parenthesis, much like with () in patterns. The reason is starts at zero is so the outermost main expression is always subexpression 0. An optional second argument specifies which "stack level" to return the expression from. If omitted, the current stack is used (same as specifying zero). A one will access the parent-stack, 2 the parent of the parent, etc.

Here is an example:

Wizard Joined: 03 Mar 2001 Posts: 1127 Location: London

Nifty, look forward to playing with it

Posted: Tue Oct 03, 2006 1:50 am

Btw, I also implemented the boolean expression optimization. In most modern programming languages, the compiler will optimize your AND/OR boolean tests and avoid executing any code that isn't needed. For example: