netcdf time axis gets irregular values

[felixcremer]

When I open the data from the xarray tutorial with Rasters I get very irregular time axis values.
I would have expected to get clear hourly DateTime objects every six hours.
The time axis is saved in the netcdf as an integer value hours since 1800-01-01.

It seems to me as if the data is converted to Float and then there is some strange float interaction.

using Rasters, NCDatasets
path = "air_temperature.nc"
# First we download the data locally if needed.
if !isfile(path)
    download("https://github.com/pydata/xarray-data/blob/master/air_temperature.nc", "air_temperature.nc")
end

ds = RasterStack("air_temperature.nc")
dims(ds, Ti) # Rasters opening the data
Ti Sampled{DateTime} ForwardOrdered Irregular DimensionalData.Dimensions.LookupArrays.Points
wrapping: 2920-element Vector{DateTime}:
 2013-01-01T00:04:02.688
 2013-01-01T06:02:18.496
 2013-01-01T12:00:34.304
 2013-01-01T17:58:50.112
 2013-01-01T23:57:05.920
 2013-01-02T06:04:06.016
 2013-01-02T12:02:21.824
 2013-01-02T18:00:37.632
 2013-01-02T23:58:53.440
 2013-01-03T05:57:09.248
 2013-01-03T12:04:09.344
 2013-01-03T18:02:25.152
 2013-01-04T00:00:40.960
 ⋮
 2014-12-28T23:56:06.528
 2014-12-29T06:03:06.624
 2014-12-29T12:01:22.432
 2014-12-29T17:59:38.240
 2014-12-29T23:57:54.048
 2014-12-30T05:56:09.856
 2014-12-30T12:03:09.952
 2014-12-30T18:01:25.760
 2014-12-30T23:59:41.568
 2014-12-31T05:57:57.376
 2014-12-31T11:56:13.184
 2014-12-31T18:03:13.280
julia> dims(yax,Ti) # This is what I would have expected.
Ti Sampled{DateTime} ForwardOrdered Irregular DimensionalData.Dimensions.LookupArrays.Points
wrapping: 2920-element Vector{DateTime}:
 2013-01-01T00:00:00
 2013-01-01T06:00:00
 2013-01-01T12:00:00
 2013-01-01T18:00:00
 2013-01-02T00:00:00
 2013-01-02T06:00:00
 2013-01-02T12:00:00
 2013-01-02T18:00:00
 ⋮
 2014-12-30T06:00:00
 2014-12-30T12:00:00
 2014-12-30T18:00:00
 2014-12-31T00:00:00
 2014-12-31T06:00:00
 2014-12-31T12:00:00
 2014-12-31T18:00:00

[rafaqz]

Weird. But I can't actually reproduce it:

julia> lookup(RasterStack(path), Ti)
Sampled{Dates.DateTime} ForwardOrdered Irregular DimensionalData.Dimensions.Lookups.Points
wrapping: 2920-element Vector{Dates.DateTime}:
 2013-01-01T00:00:00
 2013-01-01T06:00:00
 2013-01-01T12:00:00
 2013-01-01T18:00:00
 2013-01-02T00:00:00
 2013-01-02T06:00:00
 2013-01-02T12:00:00
 2013-01-02T18:00:00
 2013-01-03T00:00:00
 2013-01-03T06:00:00
 2013-01-03T12:00:00
 ⋮
 2014-12-29T06:00:00
 2014-12-29T12:00:00
 2014-12-29T18:00:00
 2014-12-30T00:00:00
 2014-12-30T06:00:00
 2014-12-30T12:00:00
 2014-12-30T18:00:00
 2014-12-31T00:00:00
 2014-12-31T06:00:00
 2014-12-31T12:00:00
 2014-12-31T18:00:00

Maybe try main?

[felixcremer]

It is fixed on main.