issues: 2019645081

This data as json

id node_id number title user state locked assignee milestone comments created_at updated_at closed_at author_association active_lock_reason draft pull_request body reactions performed_via_github_app state_reason repo type
2019645081 I_kwDOAMm_X854YVaZ 8498 Allow some notion of ordering in Dataset dims 5635139 closed 0     5 2023-11-30T22:57:23Z 2023-12-08T19:22:56Z 2023-12-08T19:22:55Z MEMBER      

What is your issue?

Currently a DataArray's dims are ordered, while a Dataset's are not.

Do we gain anything from have unordered dims in a Dataset? Could we have an ordering without enforcing it on every variable?

Here's one proposal, with fairly wide error-bars: - Datasets have a dim order, which is set at construction time or through .transpose - Currently .transpose changes the order of each variable's dims, but not the dataset's - If dims aren't supplied, we can just use the first variable's - Variables don't have to conform to that order — .assign(foo=differently_ordered) maintains the differently ordered dims. So this doesn't limit any current functionality. - When there are transformations which change dim ordering, Xarray is "allowed" to transpose variables to the dataset's ordering. Currently Xarray is "allowed" to change dim order arbitrarily — for example to put a core dim last. IIUC, we'd prefer to set a non-arbitrary order, but we don't have one to reference. - This would remove a bunch of boilerplate from methods that save the ordering, run .apply_ufunc and then reorder in the original order[^1]

What do folks think?

[^1]: though also we could do this in .apply_ufunc

 
{
    "url": "https://api.github.com/repos/pydata/xarray/issues/8498/reactions",
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   not_planned 13221727 issue

Links from other tables