ENIGMA Forums
Contributing to ENIGMA => Proposals => Topic started by: Josh @ Dreamland on September 13, 2012, 02:11:37 pm
-
I doubt anyone would disagree that ENIGMA needs arrays. What I'm proposing is that rather than inherit array syntax from C++, we take a page out of the JavaScript playbook and denote them with [].
I want to use [] for a couple reasons. First, of course, is that [1, 2, 3][1] looks neater than {1,2,3}[1]. That was a joke; I don't give a shit about that. What I want to allow is this sort of syntax:
[x, y] = [y, x] // Switches the values of x and y
[x, y] = get_some_coordinates(); // Fetches a length-2 array of coordinates, assign
Using {} for arrays, {x,y} would be ambiguous; it could be a new scope which uses x and y, or it could be our assignment array. There would be next to no way to distinguish the two, so it's out of the question if we want the assignment array.
As for the dynamics, when compiling to C, this is what it would look like:
In the [a,b,c]=array_func() case:
{
Array tmp = array_func(); // This is copied verbatim
switch (tmp.length) { default: // Array size is size_t; must be >= 0
case 3: c = tmp[3]; // Waterfall
case 2: b = tmp[2];
case 1: a = tmp[1];
case 0: ;
}
}
In the [a,b,c] = scalar_func() special case:
c = b = a = scalar_func();
As for building an array, that's easy. Old ENIGMA can do that. [expression1, expression2, expression3] simply becomes this:
[snip=cpp](Array(3).put(0,expression1).put(1,expression2).put(2,expression3))[/snip]
Where [snip]Array(int N)[/snip] reserves N variants, and [snip]Array& Array::put(int, variant)[/snip] is the same as [snip]Array::operator[](int)::operator=(variant)[/snip].
-
What about this though
a = b
[x, y] = get_some_coordinates();
-
This is an EDL feature; the expectation is that the user will distinguish those two with the use of a semicolon. However, if you have an idea to eliminate that, let's hear it.
-
Agree with
[x, y] = array_func();
Disagree with
[x, y] = scalar_func();
-
What would you have [] = scalar do?
-
Error/warn :-p
Seriously, though, what would the user expect [a,b] = 3; to do when a = b = 3 is just as easy to write.
Also, what happens with:
[a,b,c] = [1,2];
-
This is also cool, but what if I only want to set one of the variables? Like have:
[0, y] = get_some_coordinates();That wouldn't work would it?
-
No, that would throw a syntax error to the tune of "Assignment array elements must be variables."
-
So would there be a possibility to return only N'th result? From your previous example I see that this would probably work:
[x] = get_some_coordinates();As that would just return the first variable to x, but if I wanted the second one?
-
[snip=edl]y = get_some_coordinates()[2][/snip].
But I see the point you're trying to make. I guess we can allow using zero or maybe the null keyword to indicate that you don't want the field; otherwise, you'd do something like this:
int discard;
[discard,y] = get_some_coordinates();
-
y = get_some_coordinates()[2]Also works. I just didn't think of that for some reason (though I do it all the time in C++ with OpenCV).
-
Here's my proposal.
1. If function returns a scalar, forbid [] syntax. Completely. It's worthless anyway.
2. If function returns Array, allow [] syntax but force the number of parameters to match.
3. If function returns Array but you don't want to use all parameters, use
[a, b, c, ...] = func();
Still, the array would be forced to have at least 3 elements.
4. If this is *still* not acceptable, you could have this:
[a, b?] = func();
or
[a, b=2]=func();
That code would work for func() returning both 1 or 2 parameters. It'd still be an error for 0 or more than 2 parameters (for 0, use a?. For more than 2, add the "...")
[a?, b?, ...] = func();
This is inspired by function arguments syntax.
As a final note, after an optional variable, all following variables must be optional. So this would still be an error:
[a?, b] = func(); //ERROR!
As an extra note, __unused could be used for worthless variables:
[__unused, x, __unused, __unused?, y?] = func();
cout << __unused << endl; //ERROR: __unused is not defined
If you dislike this syntax, this could also be used (although it might be more error-prone):
[, x, , ?, y?] = func();
-
I don't want to use ?, since ? is ternary. [snip=EDL](boolean_expression ? x : y) = 10;[/snip] is already valid EDL. It will remain so in the arrays, I imagine. I do like your [snip]...[/snip] to denote that the array may be larger. However, I like HaRRi's 0 idea better than __unused.
Perhaps as a prefix, ? could denote optionality.
[snip=EDL][0, x, y, ?z] = get_wxyz();[/snip]
-
Prefix, postfix. I really don't care. It was just a proposal.
For further headaches:
int& x = ...;
int& y = ...;
[ x > 2 ? x : y] = func();
(Basically, what if instead of just variables it actually allowed any reference? Non-references would still be forbidden)
-
I was just going on about how that'd be fine. All I really need is to make sure JDI includes the & in variable types it coerces, then I can ask JDI for the type of ENIGMA's AST.
-
[0, x] = func();
Seems to imply that the first element should be 0. Just like this:
Array res = func();
assert(res[0] == 0);
x = res[1];
-
Then your code would also assert that the second element should be x, but it doesn't. You implying it should be 0 is just illogical. So I do like the 0 way to do this.
-
These things look like a mess to me.
-
I am fine with just it being: x = func()[2]
-
Agreed with HaRRi, again. Zero seems like an elegant solution.
I'd just make them use array subscripts, except it may not always be the case that the function returns an array of only two elements. It may return five, and the user needs only four.
-
Then maybe this would also work:
[y, z]=func()[1,2]That would also allow changing the order of the return like:
[y, z]=func()[2,1]This would set the second param to z and third to y while leaving the first unused.
But that I think is unneeded. So using [] for a specific value and 0 for an array of values seems the best way.
-
Unfortunately, GML ambiguates [1,2]. It means the same as [1][2]. Otherwise we could do all sorts of neat array tricks.
Perhaps array[[1,2]] could work.
-
Hmmm...
[x,y,z] = func()[1..3]
-
If we allow arbitrary element access by using an array as an array subscript, and we let 1..5 be the same as [1, 2, 3, 4, 5] (and maybe let 1..2..9 be [1,3,5,7,9]), then that will work.