support for reindex

[Beforerr]

I am wondering if we would like something like what xarray.reindex does https://docs.xarray.dev/en/stable/generated/xarray.DataArray.reindex.html.

I have working codes that support something like this, but I am not sure if I should make a PR.

The interface is something like this below

using DimensionalData, Dates

times = DateTime(2020, 1, 1):Day(1):DateTime(2020, 1, 10)
da = DimArray(1:10, Ti(times))

# Reindex to new times
new_times = DateTime(2020, 1, 1, 12):Day(1):DateTime(2020, 1, 10, 12)
reindex(da, Ti(new_times); method=:bfill)

One thing is that Base has reindex method for handling subarray.

[rafaqz]

Sounds good to me!

[rafaqz]

Sounds good to me!

[tiemvanderdeure]

Would this be similar to extend in Rasters? https://rafaqz.github.io/Rasters.jl/stable/api#Rasters.extend

[rafaqz]

I think partly, except it could e.g. increase the resolution but fill the missing lookup values with missings, rather than just extending for the same resolution.

I think it could be done in a few lines using DimSelectors:

# Probably needs `Missing` in the eltype, and maybe tweaks to dims
new_da = similar(old_da, dims(template_da, old_da)...)
new_da .= missing
new_da[DimSelectors(old_da)] .= old_da

I'm also not sure whats meant to happen when old lookup values dont exist in the template. It seems bad that it could silently lose data. In the above, it will error.

(The function can also pass a tolerance through to DimSelectors for when the tempalte doesn't totally match)

[Beforerr]

I felt like reindex should be more flexible than extend (and I tried to mimic what xarray does).
My use case is to handle missing data gaps and nonuniform coordinates, as well as tweak coordinate orders.

I am not sure I can follow what istemplate_da and the idea why silently losing data is bad (I felt like that reindex is kinda of different from index)

[rafaqz]

template_da is some dimarray/dimstack or just tuple of dims that holds the lookups to reindex too.

I guess my question is that it seems easy to return an array completely full of missing with no complaints at all. I guess you just change the atol until it works?

Is it required that any of the old lookup values match the template?

[Beforerr]

I felt like it is the user responsibility to specify tolerance/atol (Also this is tricky when the coordinate is not numbers, like datetime, as the tolerance should be something like Millisecond(int)).

Is it required that any of the old lookup values match the template?

No, it is not required.

[rafaqz]

I think this is something we should have, if you want to write up a PR

[Beforerr]

oh the PR is already here #1148 El El dom, dic 14, 2025 a la(s) 8:11 p.m., Rafael Schouten < ***@***.***> escribió:

…

*rafaqz* left a comment (rafaqz/DimensionalData.jl#1147) <#1147 (comment)> I think this is something we should have, if you want to write up a PR — Reply to this email directly, view it on GitHub <#1147 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOAN2QIWDLSKTGPSQD63GH34BYKFZAVCNFSM6AAAAACOIA2LHWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTMNJSGYYDQNJYGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>