Remove duplicate coordinates with indexes on sub-trees #9531
Conversation
This is a _partial_ solution to the duplicate coordinates issue from pydata#9475. Here we remove all duplicate coordinates between parent and child nodes with an index (these are already checked for equality via the alignment check). Other repeated coordinates (which we cannot automatically check for equality) are still allowed for now. We will need an alternative solution for these, as discussed in pydata#9475, but it is less obvious what the right solution is so I'm holding off on it for now.
TomNicholas
added
the
topic-DataTree
There was a problem hiding this comment.
Does this completely supercede #9510? If so how exactly does it get around the problem mentioned in #9510 (comment)?
This PR looks good, but I'm not sure we should merge it until we know what the rest of the solution to #9475 looks like.
This PR intentionally punts on the issue of duplicate inherited non-index coordinates for now, preserving the current behavior (keeping them duplicated). I think it is an incremental improvement, and any full solution in #9475 will include this. The behavior of duplicate non-indexed coordinates will also need another fix. In my mind, there are two leading contenders, but I would like to discuss it a bit more:
|
|
@TomNicholas are you comfortable merging this PR for now? I think it will be a part of whichever final solution we pick |
This is a partial solution to the duplicate coordinates issue from #9475.
Here we remove all duplicate coordinates between parent and child nodes with an index (these are already checked for equality via the alignment check).
Other repeated coordinates (which we cannot automatically check for equality) are still allowed for now. We will need an alternative solution for these, as discussed in #9475, but it is less obvious what the right solution is so I'm holding off on it for now.