Indexing the GRIB file

The FileIndex object is used to keep the values of some of the GRIB keys into memory. The keys to be kept are defined with the index_keys keyword argument (by default, GRIBDatasets.ALL_KEYS). It is also possible to filter on specific values for the keys with the filter_by_values keyword argument. This can significantly improve the performance for reading the file:

using GRIBDatasets
using BenchmarkTools

dir_tests = abspath(joinpath(dirname(pathof(GRIBDatasets)), "..", "test"))
grib_path = abspath(joinpath(dir_tests, "sample-data", "era5-levels-members.grib"))

julia> only_geopotential_on_500_hPa  = Dict(
           "cfVarName" => "z",
           "level" => 500,
       )Dict{String, Any} with 2 entries:
  "level"     => 500
  "cfVarName" => "z"
julia> @btime filtered_index = FileIndex(grib_path; filter_by_values = only_geopotential_on_500_hPa);  302.550 ms (17920 allocations: 2.78 MiB)
julia> @btime index = FileIndex(grib_path);  708.187 ms (70654 allocations: 10.59 MiB)