FIX only exclude SiteTree subclasses if there are any set.

[fspringveldt]

LumberJack::liveChildren() and LumberJack::stageChildren() should check if count(LumberJack::getExcludedSiteTreeClassNames()) > 0 to avoid an InvalidArgumentException exception being thrown by DataList::exclude().

[robbieaverill]

I think we should make this PR for the SS3 compatible version and merge up.

Have requested a review from @micmania1 for the logic

[robbieaverill]

We need to update the travis config - logged as #61

[fspringveldt]

Okay, rebased off 1.3 branch instead.

[micmania1]

I think the check can just be added to the shouldFilter() function.

The reason being; it's a bit misleading when shouldFilter() returns true because that implies that we should filter, but then we don't because count() returns 0.

[fspringveldt]

Thanks @micmania1

I think the check can just be added to the shouldFilter() function.

I considered this (after @robbieaverill suggested it too), but decided against.

The single responsibility of shouldFilter() is "Checks if we're on a controller where we should filter".

This additional check (::excludeSiteTreeClassNames()) mutates a DataList based on whether we shouldFilter() and if there any configured hidden subclasses (result of ::getExcludedSiteTreeClassNames()).

Thus, implementing this on LumberJack::shouldFilter() would mean updating everywhere it's used (e.g. BlogFilter::stageChildren()), which doesn't necessarily care about hidden subclasses.

Thoughts?

[micmania1]

I can see how it would affect blog. In that case could you just make the new method excludeSiteTreeClassNames protected since its only use is internal? If we need to make it public later we can do that easily without introducing breaking changes.

Could you also add further notes to the shouldFilter docblocks so that its absolutely clear that it only checks the current controller and maybe link back to this discussion?

[robbieaverill]

We can't do this in 1.3 any more, will need to go back to targetting master

I only read one commit, sorry