# =========================================================================
# 脚本名称: generate_dataset.jl (全量增强版)
# 功能: 自动化生成包含 case14, case30, case2383wp 的全量 MIP 优化模型
# 说明: 生成的 .mps 文件包含大量二进制变量，专用于混合整数规划 (MIP) 研究
# 路径: 输出结果将导出至上级目录的 UnitCommitment_Trajectory_Dataset 文件夹
# =========================================================================

using UnitCommitment
using JuMP
using CodecZlib

# 配置路径映射
config = Dict(
    "case14"     => "instances/matpower/case14",
    "case30"     => "instances/matpower/case30",
    "case2383wp" => "testdata/case2383wp"
)

output_root = "../UnitCommitment_Trajectory_Dataset"
variants = ["hourly_noline", "hourly_withline", "subhourly_noline", "subhourly_withline"]

function run_full_generation()
    println("🚀 开始全量数据集生成任务 (包含大型算例)...")
    
    # 确保根目录存在
    mkpath(output_root)

    for (case_name, case_path) in config
        println("\n>>> 正在准备算例: $case_name")
        
        # 创建目录
        for v in variants
            mkpath(joinpath(output_root, case_name, v))
        end
        
        # 获取目录下所有的 json.gz 文件
        all_files = filter(f -> endswith(f, ".json.gz"), readdir(case_path))
        
        for (i, file_name) in enumerate(all_files)
            date_tag = split(file_name, ".")[1]
            src_path = joinpath(case_path, file_name)
            println("    [$case_name] 处理进度: $i/$(length(all_files)) ($date_tag)")
            
            # --- 建模逻辑 ---
            
            # 1. Hourly No-Line
            instance_v1 = UnitCommitment.read(src_path)
            empty!(instance_v1.scenarios[1].lines)
            model_v1 = UnitCommitment.build_model(
                instance=instance_v1, 
                formulation=UnitCommitment.Formulation(
                    transmission=UnitCommitment.ShiftFactorsFormulation(precomputed_isf=zeros(0,0), precomputed_lodf=zeros(0,0))
                ),
                variable_names=true
            )
            JuMP.write_to_file(model_v1, joinpath(output_root, case_name, "hourly_noline", "$(case_name)_$(date_tag)_h_noline.mps"))

            # 2. Hourly With-Line
            instance_v2 = UnitCommitment.read(src_path)
            model_v2 = UnitCommitment.build_model(instance=instance_v2, variable_names=true)
            JuMP.write_to_file(model_v2, joinpath(output_root, case_name, "hourly_withline", "$(case_name)_$(date_tag)_h_withline.mps"))

            # 3. Subhourly No-Line
            instance_v3_base = UnitCommitment.read(src_path)
            instance_v3 = UnitCommitment.convert_to_subhourly(instance_v3_base, instance_v3_base)
            empty!(instance_v3.scenarios[1].lines)
            model_v3 = UnitCommitment.build_model(
                instance=instance_v3, 
                formulation=UnitCommitment.Formulation(
                    transmission=UnitCommitment.ShiftFactorsFormulation(precomputed_isf=zeros(0,0), precomputed_lodf=zeros(0,0))
                ),
                variable_names=true
            )
            JuMP.write_to_file(model_v3, joinpath(output_root, case_name, "subhourly_noline", "$(case_name)_$(date_tag)_s_noline.mps"))

            # 4. Subhourly With-Line
            instance_v4_base = UnitCommitment.read(src_path)
            instance_v4 = UnitCommitment.convert_to_subhourly(instance_v4_base, instance_v4_base)
            model_v4 = UnitCommitment.build_model(instance=instance_v4, variable_names=true)
            JuMP.write_to_file(model_v4, joinpath(output_root, case_name, "subhourly_withline", "$(case_name)_$(date_tag)_s_withline.mps"))
            
            # 及时释放内存 (Julia 的 GC 有时会有延迟)
            model_v1 = model_v2 = model_v3 = model_v4 = nothing
        end
    end
    println("\n✅ 所有算例生成完毕！生成的 MPS 文件存放在: $output_root")
end

run_full_generation()