plotting of a DimArray leads to a Segmentation Fault

[felixcremer]

I am trying to plot a DimArray that I got converted from a yaxarray and then I run into a segmentation fault this is an example cube, that leads to the segfault with the following code. You need to unzip the zarr file before opening it in julia
s1_hidalgo_deforestation_zoom.zarr.zip

julia
julia> using DimensionalData

julia> using YAXArrays

julia> using Zarr

julia> cube = Cube("pathto/s1_hidalgo_deforestation_zoom.zarr")
YAXArray with the following dimensions
lon                 Axis with 82 Elements from -98.62577203503945 to -98.61121932743671
lat                 Axis with 51 Elements from 20.619481135755912 to 20.61049798291472
time                Axis with 1098 Elements from 2014-01-02T00:00:00 to 2020-01-01T00:00:00
Variable            Axis with 4 elements: sentinel1_vv_d sentinel1_vh_a sentinel1_vh_d sentinel1_vv_a 
Total size: 70.07 MB


julia> using YAXArrayBase


julia> dimsub = yaxconvert(DimArray, cube);

julia> typeof(dimsub)
DimArray{Union{Missing, Float32}, 4, Tuple{Dim{:lon, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}}, Sampled{Ordered{ForwardIndex, ForwardArray, ForwardRelation}, Regular{Float64}, Points}, NoMetadata}, Dim{:lat, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}}, Sampled{Ordered{ReverseIndex, ForwardArray, ForwardRelation}, Regular{Float64}, Points}, NoMetadata}, Dim{:time, Vector{DateTime}, Sampled{Ordered{ForwardIndex, ForwardArray, ForwardRelation}, Irregular{DimensionalData.AutoBounds}, Points}, NoMetadata}, Dim{:Variable, Vector{String}, Categorical{Unordered{ForwardRelation}}, NoMetadata}}, Tuple{}, DiskArrayTools.DiskArrayStack{Union{Missing, Float32}, 4, ZArray{Union{Missing, Float32}, 3, Zarr.BloscCompressor, DirectoryStore}, 1}, NoName, Dict{String, Any}}

julia> plot(dimsub)
[1]    512 segmentation fault  julia

This is on DimensionalData version 0.17.10

[rafaqz]

Plotting a DimArray wont give you a heatmap, but whatever a generic plot() does to underlying data, with explicit axis values. I would guess you are breaking GR with a 3d array plot with three axis ranges? I doubt the segmentation fault is in Julia code or in DimensionalData.jl.

Testing it without DD, this doesn't work for me either:

plot(rand(50, 80, 1000))

Do you actually want to take a heatmap of a single time slice?

This use case is pretty much what GeoData.jl is for, and has specific plot recipes for. Although here it will still try to tile 1098 time slices in a single plot... but you can select sections of the timespan to do that. GeoData.jl just needs a zarr source, see https://github.com/rafaqz/GeoData.jl/issues/42.

Or you can probably do GeoArray(dimsub) to use GeoData.jl recipes.

[felixcremer]

Thanks, yes, I actually was using GeoData for that but I got a segmentation fault there as well.
You are right, that I should select the proper time slices and variable before plotting and this works now.with GeoArray.
I just tested the above code with the plotly backend and there I also get an error, but at least it doesn't segfault so the segfault seems to happen somewhere in GR or in the Plots backend of GR.