Massive var-array fix (CAUTION: affects all vars)

[sorlok]

This branch contains a lot of changes to arrays-inside-variables. It brings arrays closer to GM feature parity.

1. Removed placement new; this will (hopefully) make the code more readable and make memory leaks easier to trace.
2. Implemented the various array length functions.
3. Added printing for arrays; if a var's value type is multi-dimensional, it is printed as a 2-D or 3-D array.
4. Fixed (x,y) storage for arrays; GM does it backwards.

Please be careful with this branch; I am fairly confident in it, but it does affect the "var" class extensively.

[RobertBColton]

@sorlok Excellent work, as usual.

I can confirm your changes conform to ECMA standards and so do GM: Studio's with the following test.

myarr[1] = 50;
myarr[56] = 29;
myarr2[56] = 29;
myarr2[1] = 50;
show_message(string(array_length_1d(myarr))); // should be 57
show_message(string(array_length_1d(myarr2))); // should be 57

However I must note two discrepancies. If it is not too much to ask, can you resolve these two issues before I merge?

1. You forgot is_array, like is_real, and is_string
2. Additional warnings now appear in the console.

Universal_System/lua_table.h:78:10: warning: 'lua_table<variant>::mx_size' will be initialized after [-Wreorder]
   size_t mx_size; //Actual max_element+1
          ^
Universal_System/lua_table.h:72:10: warning:   'size_t lua_table<variant>::dn_reserve' [-Wreorder]
   size_t dn_reserve; //Total available units.
          ^
Universal_System/lua_table.h:160:3: warning:   when initialized here [-Wreorder]
   lua_table<T>() : dense(new T[1]), mx_size(1), dn_reserve(1) {
   ^

Otherwise your pull request is excellent and again I am ecstatic about having you as a contributor here sorlok. I can also confirm that several games continue to work as expected with your changes, including Project Mario where Project K is utilized to load collision data with arrays from d3d model files.

I have also taken the liberty of documenting the new functions on the Wiki for you.
http://enigma-dev.org/docs/Wiki/Data_Structure_Functions

[sorlok]

Excellent feedback; I'll fix the warnings and add the is_array function.

[sorlok]

I've added is_array(), and fixed the reorder warnings.

A note about is_array(); due to the way that we handle arrays, the following won't work right:

a[0] = 10; //is_array() prints "false", since arrays with one element are considered to be scalar.
a[1] = 10; //Now is_array() prints "true"
a = 10; //is_array() still prints "true".

I can't think of any clever ways of fixing this right now, but normal array usage probably won't encounter these errors.

[RobertBColton]

@sorlok I expanded your test to the following, I got false, true, true in ENIGMA and true, true, false in Studio.

a[0] = 10; //is_array() prints "false", since arrays with one element are considered to be scalar.
show_message(string(is_array(a)));
a[1] = 10; //Now is_array() prints "true"
show_message(string(is_array(a)));
a = 10; //is_array() still prints "true".
show_message(string(is_array(a)));

I can confirm the compile errors have been resolved. If there is anything else you would like to add please do so, otherwise I am prepared to merge this pull request, the is_array() function at least partially works. Josh says ECMA compliance will side with YoYo on this one, but that's fine for now.

[RobertBColton]

After testing on several games, I am approving this commit because it meets our expectations and improves the functionality of the engine. Great work @sorlok!

[sorlok]

Excellent! If we ever decide to do a massive rewrite of variables (probably with the new parser), let me know and I'll see about fixing is_array().

[RobertBColton]

@sorlok Alright! I don't exactly understand much about it, but yeah bets are on the new parser.

[TheExDeus]

When I have 2D array, I can't get correct size. Like so:

inputs[0,0] = "Number of images";
inputs[0,1] = 0;
inputs[0,2] = 0;
inputs[0,3] = -1;

inputs[1,0] = "Directory";
inputs[1,1] = 0;
inputs[1,2] = 0;
inputs[1,3] = -1;

All combinations in array_length_2d(inputs) always either return 4 or 0.

[sorlok]

Interesting; I'll have a look. It's possible something got mixed up in the length code.

[TheExDeus]

Just to be more specific - I need it to return 2 at some point. I think it would at array_length_2d(inputs,0) or array_length_2d(inputs, 1).

[sorlok]

Just figured it out; you actually want to use the function "array_height_2d(inputs)". The function "array_length_2d()" is used for the length of the second column.

Confirmed this behavior in GM:S. So we're doing the right thing already.