嘗試做一個.NET簡單、高效、避免OOM的Excel工具 MiniExcel


園友好,最近晚輩嘗試做一個.NET簡單、高效、避免OOM的Excel工具 MiniExcel

主要目前主流框架大多將資料全載入到內存方便操作,但這會導致內存消耗問題,MiniExcel 嘗試以 Stream 角度寫底層算法邏輯,能讓原本1000多MB占用降低到幾MB,避免內存不夠情況。適合像是低規格 azure app service 或是讀取大文件等情境。

image

特點

  • 低內存耗用,避免OOM(out of memoery)、頻繁 Full GC 情況
  • 支持即時操作每行數據
    miniexcel_lazy_load
  • 兼具搭配 LINQ 延遲查詢特性,能辦到低消耗、快速分頁等復雜查詢
  • 輕量,不依賴任何套件,DLL小於100KB
  • 簡便操作的 API 風格

Get Started

Demo

安裝

請查看 from NuGet

更新日志

請查看 Release Notes

TODO

請查看 TODO

性能測試

Test1,000,000x10.xlsx 做基准與主流框架做性能測試,總共 1千萬筆 "HelloWorld",文件大小 23 MB

Benchmarks 邏輯可以在 MiniExcel.Benchmarks 查看或是提交 PR,運行指令

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

最后一次運行結果 :

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 最大內存耗用 平均時間 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 最大內存耗用 平均時間 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

讀 Excel

1. Query 查詢 Excel 返回強型別 IEnumerable 數據 [Try it]

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>();

image

2. Query 查詢 Excel 返回Dynamic IEnumerable 數據 [Try it]

  • Key 系統預設為 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);
}

3. 查詢數據以第一行數據當Key [Try it]

注意 : 同名以右邊數據為准

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);
}

4. Query 查詢支援延遲加載(Deferred Execution),能配合LINQ First/Take/Skip辦到低消耗、高效率復雜查詢

舉例 : 查詢第一筆數據

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);
}

與其他框架效率比較 :

queryfirst

5. 查詢指定 Sheet 名稱

MiniExcel.Query(path, sheetName: "SheetName");
//or
stream.Query(sheetName: "SheetName");

6. 查詢所有 Sheet 名稱跟數據

var sheetNames = MiniExcel.GetSheetNames(path).ToList();
foreach (var sheetName in sheetNames)
{
    var rows = MiniExcel.Query(path, sheetName: sheetName);
}

7. 查詢所有欄(列)

var columns = MiniExcel.GetColumns(path); // e.g result : ["A","B"...]

var cnt = columns.Count;  // get column count

8. Dynamic Query 轉成 IDictionary<string,object> 數據

foreach(IDictionary<string,object> row in MiniExcel.Query(path))
{
    //..
}

寫 Excel

  1. 必須是非abstract 類別有公開無參數構造函數
  2. MiniExcel SaveAs 支援 IEnumerable參數延遲查詢,除非必要請不要使用 ToList 等方法讀取全部數據到內存

圖片 : 是否呼叫 ToList 的內存差別

image1. Anonymous or strongly type [Try it]

var path = Path.Combine(Path.GetTempPath(), $"{Guid.NewGuid()}.xlsx");
MiniExcel.SaveAs(path, new[] {
    new { Column1 = "MiniExcel", Column2 = 1 },
    new { Column1 = "Github", Column2 = 2}
});

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);

3. 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);
}

4. 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

5. SaveAs 支援 Stream [Try it]

using (var stream = File.Create(path))
{
    stream.SaveAs(values);
}

Excel Column Name/Index/Ignore Attribute

e.g

input excel :

image

public class ExcelAttributeDemo
{
    [ExcelColumnName("Column1")]
    public string Test1 { get; set; }
    [ExcelColumnName("Column2")]
    public string Test2 { get; set; }
    [ExcelIgnore]
    public string Test3 { get; set; }
    [ExcelColumnIndex("I")] // system will convert "I" to 8 index
    public string Test4 { get; set; } 
    public string Test5 { get; } //wihout set will ignore
    public string Test6 { get; private set; } //un-public set will ignore
    [ExcelColumnIndex(3)] // start with 0
    public string Test7 { get; 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("Test7", rows[0].Test4);
Assert.Null(rows[0].Test5);
Assert.Null(rows[0].Test6);
Assert.Equal("Test4", rows[0].Test7);

Excel 類別自動判斷

MiniExcel 預設會根據擴展名或是 Stream 類別判斷是 xlsx 還是 csv,但會有失准時候,請自行指定。

stream.SaveAs(excelType:ExcelType.CSV);
//or
stream.SaveAs(excelType:ExcelType.XLSX);
//or
stream.Query(excelType:ExcelType.CSV);
//or
stream.Query(excelType:ExcelType.XLSX);

范例

1. SQLite & Dapper 讀取大數據新增到數據庫

note : 請不要呼叫 call ToList/ToArray 等方法,這會將所有數據讀到內存內

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:
image

2. ASP.NET Core 3.1 or MVC 5 下載 Excel Xlsx API Demo Try it

public class ExcelController : Controller
{
    public IActionResult DownloadExcel()
    {
        var values = new[] {
            new { Column1 = "MiniExcel", Column2 = 1 },
            new { Column1 = "Github", Column2 = 2}
        };
        var memoryStream = new MemoryStream();
        memoryStream.SaveAs(values);
        memoryStream.Seek(0, SeekOrigin.Begin);
        return new FileStreamResult(memoryStream, "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet")
        {
            FileDownloadName = "demo.xlsx"
        };
    }

    public IActionResult DownloadExcelFromTmplate()
    {
        var templatePath = "TestTemplateComplex.xlsx";
        var value = new Dictionary<string, object>()
        {
            ["title"] = "FooCompany",
            ["managers"] = new[] {
                new {name="Jack",department="HR"},
                new {name="Loan",department="IT"}
            },
            ["employees"] = new[] {
                new {name="Wade",department="HR"},
                new {name="Felix",department="HR"},
                new {name="Eric",department="IT"},
                new {name="Keaton",department="IT"}
            }
        };
        var memoryStream = new MemoryStream();
        memoryStream.SaveAsByTemplate(templatePath, value);
        memoryStream.Seek(0, SeekOrigin.Begin);
        return new FileStreamResult(memoryStream, "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet")
        {
            FileDownloadName = "demo.xlsx"
        };
    }

    public IActionResult DownloadExcelFromTmplate_StremVersion()
    {
        var templatePath = "TestTemplateComplex.xlsx";
        var tytes = System.IO.File.ReadAllBytes(templatePath);
        var value = new Dictionary<string, object>()
        {
            ["title"] = "FooCompany",
            ["managers"] = new[] {
                new {name="Jack",department="HR"},
                new {name="Loan",department="IT"}
            },
            ["employees"] = new[] {
                new {name="Wade",department="HR"},
                new {name="Felix",department="HR"},
                new {name="Eric",department="IT"},
                new {name="Keaton",department="IT"}
            }
        };
        var memoryStream = new MemoryStream();
        memoryStream.SaveAsByTemplate(tytes, value);
        memoryStream.Seek(0, SeekOrigin.Begin);
        return new FileStreamResult(memoryStream, "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet")
        {
            FileDownloadName = "demo.xlsx"
        };
    }
}

3. 分頁查詢

void Main()
{
	var rows = MiniExcel.Query(path);
	
	Console.WriteLine("==== No.1 Page ====");
	Console.WriteLine(Page(rows,pageSize:3,page:1));
	Console.WriteLine("==== No.50 Page ====");
	Console.WriteLine(Page(rows,pageSize:3,page:50));
	Console.WriteLine("==== No.5000 Page ====");
	Console.WriteLine(Page(rows,pageSize:3,page:5000));
}

public static IEnumerable<T> Page<T>(IEnumerable<T> en, int pageSize, int page)
{
	return en.Skip(page * pageSize).Take(pageSize);
}

20210419

FQA

Q: 如何將查詢結果轉為 DataTable

提醒 : 不建議使用,因為DataTable會將數據全載入內存,失去MiniExcel低內存消耗功能。

public static DataTable QueryAsDataTable(string path)
{
	var rows = MiniExcel.Query(path, true);
	var dt = new DataTable();
	var first = true;
	foreach (IDictionary<string, object> row in rows)
	{
		if (first)
		{
			foreach (var key in row.Keys)
			{
				var type = row[key]?.GetType() ?? typeof(string);
				dt.Columns.Add(key, type);
			}

			first = false;
		}
		dt.Rows.Add(row.Values.ToArray());
	}
	return dt;
}

image

局限與警告

  • 目前不支援 xls (97-2003) 或是加密文件。


免責聲明!

本站轉載的文章為個人學習借鑒使用,本站對版權不負任何法律責任。如果侵犯了您的隱私權益,請聯系本站郵箱yoyou2525@163.com刪除。



 
粵ICP備18138465號   © 2018-2025 CODEPRJ.COM