Issue 53: Allow to use treeAttribute as external key

[mikehaertl]

Fix for issue #53

[mikehaertl]

@creocoder Before merging, maybe you can explain why at this point there is no tree with tree Id 10 here. It is created in the previous test and the DB is not cleared as far as I can see. I'm still quite confused about the state of the DB for each test. I only came up with this solution by trial and error.

[mikehaertl]

@creocoder Any news on this?

[creocoder]

@mikehaertl In general fix is much more complex than introduced in that PR. So i'll take this ticket myself. I'll explain why. There should be special "externalRoots" mode. And behavior should take that into account. For example roots() method should throw exception when used in that mode, etc. So "externalRoots" mode is not so trivial to implement. On other side its not general direction of that behavior. This is why this ticket in low priority.

[mikehaertl]

Hmm, can you explain that a bit more? I don't think that such a mode is really neccessary. And why do you think that roots() should throw an exception in that mode? How would you obtain the root nodes in that case?

The only problem i can see is something like this:

$root1 = new Category(['tree' => 2, 'name' => 'test']);
$root1->makeRoot();
echo $root1->id; // == 1

$root2 = new Category(['name' => 'test2']);
$root2->makeRoot();
echo $root2->id; // == 2

The last call to makeRoot() would mess things up, as it now would use the same tree id 2 for $root2. But this could also be solved if we'd just insert another check in afterInsert(), whether the tree id does already exist.

[creocoder]

@mikehaertl Seems you misunderstood original issue. External roots mean roots in different table. So no roots in NS table at all. Your PR is about allow use custom ids for roots, but roots is still exists inside NS table. This not solve original issue.

[mikehaertl]

Hmm. But won't you always need a root node to have a consistent tree in your node table?