rafaqz / Rasters.jl

Raster manipulation for the Julia language
MIT License
197 stars 34 forks source link

slow speed in write nc #642

Closed kongdd closed 2 months ago

kongdd commented 2 months ago

It costs about 80s to save a Raster with a dimension of 3600*1500 into nc with Rasters.jl. While, in comparison, NCDatasets only costs 2s.

using Rasters
using NCDatasets

f = "./LAI_mean_Global_WHIT_2000-2023.nc"
@time ra = Raster(f, lazy=false)

@time write("temp.nc", ra; force=true)
@time write("temp.nc", ra; force=true)

# 105.611572 seconds (272.79 M allocations: 11.790 GiB, 1.76% gc time)
# 77.975531 seconds (272.79 M allocations: 11.790 GiB, 2.06% gc time)

LAI_mean_Global_WHIT_2000-2023.zip

rafaqz commented 2 months ago

Should be fixed already on main

kongdd commented 2 months ago

Any release plan for the latest version?

rafaqz commented 2 months ago

We have a breaking change, so I'm trying to lump them all.

It will be released whenever I can get this passing: https://github.com/rafaqz/Rasters.jl/pull/612

And also get up to date with DD again...

rafaqz commented 2 months ago

Just work on main for now ;)

Hopefully before next week there will be a new version, but I've been thinking that for a month now so who knows

rafaqz commented 2 months ago

0.11.0 is registered now