Allow manual seeding of leaves

[ChrisBeaumont]

A feature I would find useful is the ability to pass a list of seed pixels to compute. The dendrogram would then be constructed in such a way that each leaf contains exactly one seed pixel. This would be helpful for, e.g., resolution studies.

[astrofrog]

Actually, this could be done e.g. by a custom function as suggested in #7

[ChrisBeaumont]

Reading #7 reminded me of this. Perhaps both can be addressed using callback functions

[astrofrog]

@ChrisBeaumont - yes, in fact what I was suggesting is that all one needs to do in the scenario of #7 and the implementation of #10 is pass a function that will only accept a leaf as independent if it contains one of a list of seed pixels. However, I do agree that it would be nice to make this a little easier to use that requiring the user to have to write a custom function. We could either:

• add a hard-coded initialization option when creating the dendrograms for the list of seed pixels

• add a method that will prune a tree after the fact based on 'seed' pixels

• add a helper function that could be passed to Example custom merge test function #10 and which just takes the list of seed pixels, e.g.:

  from astrodendro.helpers import seed_leaves
  d = Dendrogram.compute(image, merge_test_function=seed_leaves(x, y, v))

[ChrisBeaumont]

I think #7 and #8 both have their uses, and they don't really interfere with each other. Your second suggestion seems like the best way to achieve what I was talking about, without complicating the __init__ or __compute__ signatures further.

[astrofrog]

Do we still want to go with the second option now we've said (in other issues) that re-computing is preferable for now to pruning?

[ChrisBeaumont]

Maybe not. Option 3 would be the simplest interface, since we want to use custom merger functions anyways. I think we should merge that code, and then make sure it's easy to use a custom merge test function to build a dendrogram with manually-specified seeds

[ChrisBeaumont]

Closed via #10