Multi-Root

[JaySunSyn]

Merged
https://github.com/Neovici/cosmoz-tree/tree/mat-tests
https://github.com/Neovici/cosmoz-tree/tree/multi-root

And added some fixes & tests.

Please check.

[nomego]

Yes and the same conditions are still present.

The biggest change is that we now support pathLocators that address ancestors that we don't have access to. And then we had a few discussions on other cases with other types of trees for wider support.

A lot of our discussions are spread out in PR 1, but I think the biggest one on the subject is #1 (comment) .

I'm not sure if it's there or somewhere else but we discussed this point and I do agree that for our purposes, we can simplify the logic by saying "it's fine to lack parts of that pathLocator in the beginning, but invalidate any path that can't find a needed path part in the middle or end".
What we're doing now is basically saying "if we can't find any valid path on any node, should we return the best candidate we could find?"
I actually think we shouldn't, just return undefined/null since "best candidate" can be anything, but it's still not right. This would simplify _getPathNodes a bit and enable some performance improvements.

What I'm not sure about though is if I request the pathNodes for 1.2.3.4.5 and only have permission from 3 (recursively), should we return an array with 5 elements, first 2 undefined, or just the last three that we actually had permissions too ?
I would expect to receive as many elements back as I had parts in my pathLocator, but on the other hand that just puts additional error-checking on the caller, likely every time.

Thoughts @plequang @JaySunSyn ?

[nomego]

Sorry, yes your implementation also supports missing ancestors but adds a dependency on the "pathLocator" property.

[plequang]

adds a dependency on the "pathLocator" property.
I don't understand. The getPathNodes function modified by the PR takes pathLocator as an argument.

Also I keep thinking that :

the function that returns a path should return an array of nodes that start at one of the root, and without any undefined.

If no such path exists, this means you don't have access.

I think you over complicated the test cases in missing ancestor tree tests:

1. The id used as the key for path locator is unique. So id "301" can appear only below only one path.
2. The test for path 1.2.7.8 should return null, as that node does not exists/not visible. There is no point in returning the beginning of the path.

What I'm not sure about though is if I request the pathNodes for 1.2.3.4.5 and only have permission from 3 (recursively), should we return an array with 5 elements, first 2 undefined

I think it's useless, we won't display anything regarding the 2 undefined nodes. 3 will appear as if it was a root. Think of network shares, or the "shared with me" folder in Google drive.

[plequang]

Anyway, I made some tests and this implementation works in real situation scenario.
Will merge it and we can take this discussion later on.