Julia: HDF5与JLD2两个库的比较

直接上码：下面是REPL代码，不是完整的脚本工程！

using JLD2;
using DelimitedFiles;

csv_file = "C:\\Users\\rustr\\Desktop\\000001.XSHE.csv";
jld_file = "C:\\Users\\rustr\\Desktop\\000001.XSHE2.h5";

# readdlm  header=true,=>tuple
csv_data,~ = DelimitedFiles.readdlm(csv_file,',',header =true);
# @time :6.5 seconds
s= size(csv_data); # 59w,14col

#close(f)
csv = csv_data[:,2:end];
new_csv = drop_array_null(csv);

# data=>
data = turn_larger(new_csv,20);
group_dataset_name = "000001.XSHE/000002.XSHE";
# write
#@time write_jld_h5(jld_file,data,group_dataset_name) # 4s
# read
for i =1:10
    @time data = read_jld_h5(jld_file,group_dataset_name) #0.5-1s
end

#test
function turn_larger(data,n)
    _data = data
    for i =1: n
        _data = vcat(_data,data)
    end
    return _data
end

function drop_array_null(csv_data)
    s = size(csv_data)
    arrdata = Array{Float64,2}(undef,s...)
    for i = 1:length(csv_data)
        if csv_data[i] == ""
            arrdata[i]= 0.0
        else
            arrdata[i] =csv_data[i]
        end
    end
    return arrdata
end

function write_jld_h5(file,data,group_dataset_name)
    f = JLD2.jldopen(file, "w")
    try
        JLD2.write(f, group_dataset_name, data)#
    finally
        JLD2.close(f)
    end
end

function read_jld_h5(file,group_dataset_name)
    f = JLD2.jldopen(file, "r")
    data =[]
    try
        data = JLD2.read(f,group_dataset_name)
    finally   
        JLD2.close(f)        
    end
    return data
end
#
@time write_h5(jld_file,data,group_dataset_name)
function write_h5(file,data,group_dataset_name)
    HDF5.h5write(file,group_dataset_name,data)
end
@time data = read_h5(jld_file,group_dataset_name)
function read_h5(file,group_dataset_name)
    fid = HDF5.h5open(file,"r")
    data =[]
    try
        data = HDF5.read(fid,group_dataset_name)
    finally
        HDF5.close(fid)
    end
    return data
end

结论：
1、两个库效率差不多，至少JLD2没感觉和HDF5有什么明显的区别。

由于JLD2库更加有活力，一直在更新，是Julia的主力库；HDF5库则不同，已经不更新了。

建议转JLD2吧。