尝试做一个.NET简单、高效、避免OOM的Excel工具

Github : https://github.com/shps951023/MiniExcel

简介

我尝试做一个.NET简单、高效、避免OOM的Excel工具

目前主流框架大多将资料全载入到记忆体方便操作，但这会导致记忆体消耗问题，MiniExcel 尝试以 Stream 角度写底层算法逻辑，能让原本1000多MB占用降低到几MB，避免记忆体不够情况。适合像是低规格 azure app service 或是读取大档案等情境。

特点

低内存耗用，避免OOM(out of memoery)、频繁 Full GC 情况
支持即时操作每行数据
兼具搭配 LINQ 延迟查询特性，能办到低消耗、快速分页等复杂查询

图片:与主流框架对比的消耗、效率差
轻量，不依赖任何套件，DLL小于100KB
简便操作的 Dapper API 风格

性能测试

以 Test1,000,000x10.xlsx 做基准与主流框架做性能测试，总共 1千万笔 "HelloWorld"，文件大小 23 MB

Benchmarks 逻辑可以在 MiniExcel.Benchmarks 查看或是提交 PR，运行指令

dotnet run -p .\benchmarks\MiniExcel.Benchmarks\ -c Release -f netcoreapp3.1 -- -f * --join

最后一次运行结果 :

BenchmarkDotNet=v0.12.1, OS=Windows 10.0.19042

Intel Core i7-7700 CPU 3.60GHz (Kaby Lake), 1 CPU, 8 logical and 4 physical cores

  [Host]     : .NET Framework 4.8 (4.8.4341.0), X64 RyuJIT

  Job-ZYYABG : .NET Framework 4.8 (4.8.4341.0), X64 RyuJIT

IterationCount=3  LaunchCount=3  WarmupCount=3

Method	最大内存耗用	平均时间	Gen 0	Gen 1	Gen 2
'MiniExcel QueryFirst'	0.109 MB	726.4 us	-	-	-
'ExcelDataReader QueryFirst'	15.24 MB	10,664,238.2 us	566000.0000	1000.0000	-
'MiniExcel Query'	17.3 MB	14,179,334.8 us	367000.0000	96000.0000	7000.0000
'ExcelDataReader Query'	17.3 MB	22,565,088.7 us	1210000.0000	2000.0000	-
'Epplus QueryFirst'	1,452 MB	18,198,015.4 us	535000.0000	132000.0000	9000.0000
'Epplus Query'	1,451 MB	23,647,471.1 us	1451000.0000	133000.0000	9000.0000
'OpenXmlSDK Query'	1,412 MB	52,003,270.1 us	978000.0000	353000.0000	11000.0000
'OpenXmlSDK QueryFirst'	1,413 MB	52,348,659.1 us	978000.0000	353000.0000	11000.0000
'ClosedXml QueryFirst'	2,158 MB	66,188,979.6 us	2156000.0000	575000.0000	9000.0000
'ClosedXml Query'	2,184 MB	191,434,126.6 us	2165000.0000	577000.0000	10000.0000

Method	最大内存耗用	平均时间	Gen 0	Gen 1	Gen 2
'MiniExcel Create Xlsx'	15 MB	11,531,819.8 us	1020000.0000	-	-
'Epplus Create Xlsx'	1,204 MB	22,509,717.7 us	1370000.0000	60000.0000	30000.0000
'OpenXmlSdk Create Xlsx'	2,621 MB	42,473,998.9 us	1370000.0000	460000.0000	50000.0000
'ClosedXml Create Xlsx'	7,141 MB	140,939,928.6 us	5520000.0000	1500000.0000	80000.0000

Query 查询 Excel 返回`强型别` IEnumerable 数据 [Try it]

推荐使用 Stream.Query 效率会相对较好。

public class UserAccount

{

    public Guid ID { get; set; }

    public string Name { get; set; }

    public DateTime BoD { get; set; }

    public int Age { get; set; }

    public bool VIP { get; set; }

    public decimal Points { get; set; }

}

var rows = MiniExcel.Query<UserAccount>(path);

// or

using (var stream = File.OpenRead(path))

    var rows = stream.Query<UserAccount>();

Query 查询 Excel 返回`Dynamic` IEnumerable 数据 [Try it]

Key 系统预设为 A,B,C,D...Z

MiniExcel	1
Github	2

var rows = MiniExcel.Query(path).ToList();

// or

using (var stream = File.OpenRead(path))

{

    var rows = stream.Query().ToList();

    Assert.Equal("MiniExcel", rows[0].A);

    Assert.Equal(1, rows[0].B);

    Assert.Equal("Github", rows[1].A);

    Assert.Equal(2, rows[1].B);

}

查询数据以第一行数据当Key [Try it]

note : 同名以右边数据为准

Input Excel :

Column1	Column2
MiniExcel	1
Github	2

var rows = MiniExcel.Query(useHeaderRow:true).ToList();

// or

using (var stream = File.OpenRead(path))

{

    var rows = stream.Query(useHeaderRow:true).ToList();

    Assert.Equal("MiniExcel", rows[0].Column1);

    Assert.Equal(1, rows[0].Column2);

    Assert.Equal("Github", rows[1].Column1);

    Assert.Equal(2, rows[1].Column2);

}

Query 查询支援延迟加载(Deferred Execution)，能配合LINQ First/Take/Skip办到低消耗、高效率复杂查询

Query First

var row = MiniExcel.Query(path).First();

Assert.Equal("HelloWorld", row.A);

// or

using (var stream = File.OpenRead(path))

{

    var row = stream.Query().First();

    Assert.Equal("HelloWorld", row.A);

}

建立 Excel 文件 [Try it]

必须是 non-abstract 类别有 public parameterless constructor
MiniExcel SaveAs 支援 IEnumerable参数``延迟查询，除非必要请不要使用 ToList 等方法读取全部数据到内存

图片 : 是否呼叫 ToList 的内存差别

Anonymous or strongly type:

var path = Path.Combine(Path.GetTempPath(), $"{Guid.NewGuid()}.xlsx");

MiniExcel.SaveAs(path, new[] {

    new { Column1 = "MiniExcel", Column2 = 1 },

    new { Column1 = "Github", Column2 = 2}

});

Datatable:

var path = Path.Combine(Path.GetTempPath(), $"{Guid.NewGuid()}.xlsx");

var table = new DataTable();

{

    table.Columns.Add("Column1", typeof(string));

    table.Columns.Add("Column2", typeof(decimal));

    table.Rows.Add("MiniExcel", 1);

    table.Rows.Add("Github", 2);

}

MiniExcel.SaveAs(path, table);

Dapper:

using (var connection = GetConnection(connectionString))

{

    var rows = connection.Query(@"select 'MiniExcel' as Column1,1 as Column2 union all select 'Github',2");

    MiniExcel.SaveAs(path, rows);

}

IEnumerable<IDictionary<string, object>>

var values = new List<Dictionary<string, object>>()

{

    new Dictionary<string,object>{{ "Column1", "MiniExcel" }, { "Column2", 1 } },

    new Dictionary<string,object>{{ "Column1", "Github" }, { "Column2", 2 } }

};

MiniExcel.SaveAs(path, values);

output :

Column1	Column2
MiniExcel	1
Github	2

SaveAs 支援 Stream [Try it]

using (var stream = File.Create(path))

{

    stream.SaveAs(values);

}

Excel Column Name/Ignore Attribute

e.g

input excel :

Test1	Test2	Test3	Test4	Test5	Test6	Column1	Column2
Test1	Test2	Test3	Test4	Test5	Test6	Column1	Column2

public class ExcelAttributeDemo

{

    [ExcelColumnName("Column1")]

    public string Test1 { get; set; }

    [ExcelColumnName("Column2")]

    public string Test2 { get; set; }

    [ExcelIgnore]

    public string Test3 { get; set; }

    public string Test4 { get; set; }

    public string Test5 { get; }

    public string Test6 { get; private set; }

}

var rows = MiniExcel.Query<ExcelAttributeDemo>(path).ToList();

Assert.Equal("Column1", rows[0].Test1);

Assert.Equal("Column2", rows[0].Test2);

Assert.Null(rows[0].Test3);

Assert.Equal("Test4", rows[0].Test4);

Assert.Null(rows[0].Test5);

Assert.Null(rows[0].Test6);

例子 : SQLite & Dapper 读取大数据新增到数据库

note : 请不要呼叫 call ToList/ToArray 等方法，这会将所有数据读到内存内

using (var connection = new SQLiteConnection(connectionString))

{

    connection.Open();

    using (var transaction = connection.BeginTransaction())

    using (var stream = File.OpenRead(path))

    {

	   var rows = stream.Query();

	   foreach (var row in rows)

			 connection.Execute("insert into T (A,B) values (@A,@B)", new { row.A, row.B }, transaction: transaction);

	   transaction.Commit();

    }

}

效能:

例子 : ASP.NET Core 3.1 or MVC 5 下载 Excel Xlsx API Demo

public class ExcelController : Controller

{

    public IActionResult Download()

    {

        var values = new[] {

            new { Column1 = "MiniExcel", Column2 = 1 },

            new { Column1 = "Github", Column2 = 2}

        };

        var stream = new MemoryStream();

        stream.SaveAs(values);

        return File(stream,

            "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",

            "demo.xlsx");

    }

}

Excel 类别自动判断

MiniExcel 预设会根据扩展名或是 Stream 类别判断是 xlsx 还是 csv，但会有失准时候，请自行指定。

stream.SaveAs(excelType:ExcelType.CSV);

//or

stream.SaveAs(excelType:ExcelType.XLSX);

//or

stream.Query(excelType:ExcelType.CSV);

//or

stream.Query(excelType:ExcelType.XLSX);

Dynamic Query 转换 `IDictionary<string,object>` 数据

foreach(IDictionary<string,object> row = MiniExcel.Query(path))

{

    //..

}

局限与警告

目前不支援 xls (97-2003) 或是加密文件。
不支援样式、字体、宽度等修改，因为 MiniExcel 概念是只专注于值数据，借此降低内存消耗跟提升效率。

参考

读取逻辑 : ExcelDataReader
API 设计方式 :　StackExchange/Dapper