ENH Convert scalars and arrays to viewable data

[emteknetnz]

Issue #11196

[GuySartorelli]

ArrayList converts associative arrays to ArrayData - should we do the same here?

[GuySartorelli]

Why is this conditional on the casting helper being identical to the default cast?

[GuySartorelli]

Suggested change

	// Cast arrays to ArrayList by default
	// Wrap arrays in ArrayList by default

[GuySartorelli]

I don't think we should be changing the way ArrayList works - people use ArrayList for purposes other than just use in templates, and this could break some logic people have that assumes the current logic will remain as-is.

[GuySartorelli]

What about other list types? We have to assume someone has made other implementations of SS_List.

[GuySartorelli]

Making these changes in obj() is going to have wider ranging effects than just making things work in the templating system. These changes should all be done inside the template parser.

[emteknetnz]

The way things are setup I really don't know how to do this.

My PageController has this

    public function ArrayTest()
    {
        return ['item3', 'item4'];
    }

My template has this

		<% loop $ArrayTest %>
			$Me
		<% end_loop %>

The generated template is this

$scope->locally()->obj('ArrayTest', null, true); $scope->pushScope(); while (($key = $scope->next()) !== false) {
$val .= '
			';

$val .= $scope->locally()->XML_val('Me', null, true);
$val .= '
		';


}; $scope->popScope(); 

$scope->locally()->obj('ArrayTest', null, true); will call SSViewer_DataPresenter::obj() which will call SSViewer_Scope::obj() which calls SSViewer_DataPresenter::getObj() which calls SSViewer_Scope::getObj() which call ViewableData::obj() which is where the code changes are. ViewableData::obj() is what actually gets stuff, in this case PageController::ArrayTest() AND it wraps the results in a cache.

When ViewableData::obj() encounters an array, before this PR, it would convert it to a DBText because that's the ViewableData.default_cast. So all I get to work with in SSViewer_Scope::getObj() is a DBText which is obviously no usable

While this recipe-sink CI has gone green with this PR, it seems fair enough being concerned about ViewableData::obj() having potential side-effects. Given that, we could target this to CMS 6 instead?

[GuySartorelli]

I had to dive in and try to get this working myself to wrap my head around it. I ended up with these commits which handle allowing primitives to be used in templates, and allowing iterators such as ArrayIterator to work natively in templates.

I cam to the same conclusion as you regarding arrays though - even though SSViewer_Scope::next() has a comment saying the item could be an array, and handles that, there's no way to actually have an array there unless we update ViewableData::obj() to return them.

What should we do?

I think we need to do two things:

1. In CMS 5 allow primitives other than array. This is an improvement for simple use cases and doesn't require any BC breaking changes as my attempt shows. Looking at the commit history in this PR you had something very similar originally as well.
2. In CMS 6 make whatever changes are necessary to get arrays working natively in templates. This could mean:
   1. Update ViewableData::obj() to return arrays directly if there are any
   2. Update ViewableData::obj() to wrap arrays in either a list or ArrayData depending on what type of array they are
   3. Update ArrayList to either never wrap arrays, or to always wrap them in either a list or ArrayData depending on what type of array they are
   4. Some combination of the above
   5. Decoupling the logic from ViewableData - though that may be more effort than it's worth
   6. Something else neither of us considered yet, if someone else has any ideas.

I'd say update this PR to handle point 1 above, and raise a new card to look at point 2 since that'll need to go through refinement.

Considering there's a ton of ways point 2 could go, that may even be worth discussing in an architecture catchup.

[emteknetnz]

I'll close the PR to retain the diff for the CMS 6 card and open a new one to do just the scalars