前言

最近经历了一次数据量比较大的导出，也做了各种优化尝试，这里稍记录一下

一、DataTable =>EXCEL

using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Runtime.InteropServices;
using System.Text;
using System.Threading.Tasks;
using DataTable = System.Data.DataTable;
using Microsoft.Office.Interop.Excel;namespace ClassLibrary
{public class ExcelHelper{/// <summary>/// dt输出到新的Excel/// </summary>public static void OutPut(DataTable dt,string Path, string Filename){// 初始化 Excel 应用程序Application excelApp = new Application();Workbook workbook = null;Worksheet worksheet = null;try{// 设置可见性和默认路径excelApp.Visible = false; // 隐藏应用程序窗口if (!Directory.Exists(Path)) // 验证文件夹是否存在{Directory.CreateDirectory(Path); // 若不存在，则创建新文件夹}// 添加一个新的工作簿workbook = excelApp.Workbooks.Add();// 获取第一个工作表worksheet = (Worksheet)workbook.Sheets[1];//列名for (int i = 0; i < dt.Columns.Count; i++){worksheet.Cells[1, i + 1] = dt.Columns[i].ColumnName;}int rows = 2;//当前行// 获取 DataTable 列数和行数int rowCount = dt.Rows.Count;int colCount = dt.Columns.Count;// 构建数据部分的二维数组if (dt.Rows.Count > 0){object[,] dataValues = new object[rowCount, colCount];for (int r = 0; r < rowCount; r++){for (int c = 0; c < colCount; c++){dataValues[r, c] = dt.Rows[r][c];}}// 定义目标范围（从 (2, 2) 开始）Range startCell = (Range)worksheet.Cells[2, 1]; // 起始单元格Range endCell = (Range)worksheet.Cells[2 + rowCount - 1, 1 + colCount - 1]; // 结束单元格Range targetRange = worksheet.get_Range(startCell, endCell);// 写入数据targetRange.Value = dataValues;// 自动调整列宽worksheet.Columns.AutoFit();}//宽度自适应workbook.SaveAs(Filename);workbook.Close(false);}catch (Exception ex){return;}finally{if (worksheet != null) Marshal.ReleaseComObject(worksheet);if (workbook != null) Marshal.ReleaseComObject(workbook);if (excelApp != null){excelApp.Quit();Marshal.ReleaseComObject(excelApp);}GC.Collect();GC.WaitForPendingFinalizers();}}}
}

.netframework4.8,支持的Microsoft.Office.Interop.Excel。
将DataTable转换为二维数组，划定范围，一次性塞进去。
比逐单元格赋值效率更高

二、DBReader =>Excel （NPOI）

using NPOI.SS.UserModel;
using NPOI.XSSF.Streaming;
using System.Data;
using System.Data.Common;
public class ExcelHelper
{public static async Task<bool> OutPut(DbDataReader reader, string filePath){try{const int maxRowsPerSheet = 1_000_000;  // 每个Sheet最大行数const int bufferSize = 1000;            // 行缓存大小（优化内存）// 初始化流式工作簿（关键内存优化）using var workbook = new SXSSFWorkbook(null, bufferSize, true);int sheetIndex = 1;ISheet currentSheet = workbook.CreateSheet($"Sheet_{sheetIndex}");// 创建标题行IRow headerRow = currentSheet.CreateRow(0);for (int i = 0; i < reader.FieldCount; i++){headerRow.CreateCell(i).SetCellValue(reader.GetName(i));}int rowIndex = 1;  // 数据行从1开始（0是标题行）while (reader.Read()){// 达到最大行数时切换Sheetif (rowIndex >= maxRowsPerSheet){// 正确调用无参数FlushRows（NPOI 2.7.3+）((SXSSFSheet)currentSheet).FlushRows();  // 刷新当前Sheet[^1]sheetIndex++;currentSheet = workbook.CreateSheet($"Sheet_{sheetIndex}");rowIndex = 1;  // 新Sheet重置行索引}// 创建数据行IRow dataRow = currentSheet.CreateRow(rowIndex);// 写入所有列数据（类型安全处理）for (int col = 0; col < reader.FieldCount; col++){var cell = dataRow.CreateCell(col);// 根据数据类型安全写入if (reader.IsDBNull(col)){cell.SetCellValue((string)null);}else{switch (Type.GetTypeCode(reader.GetFieldType(col))){case TypeCode.String:cell.SetCellValue(reader.GetString(col));break;case TypeCode.DateTime:cell.SetCellValue(reader.GetDateTime(col));break;case TypeCode.Int16:cell.SetCellValue(reader.GetInt16(col));break;case TypeCode.Int32:cell.SetCellValue(reader.GetInt32(col));break;case TypeCode.Int64:cell.SetCellValue(reader.GetInt64(col));break;// case TypeCode.Decimal:// cell.SetCellValue(reader.GetDecimal(col));// break;case TypeCode.Double:cell.SetCellValue(reader.GetDouble(col));break;case TypeCode.Single:cell.SetCellValue(reader.GetDouble(col));break;case TypeCode.Boolean:cell.SetCellValue(reader.GetBoolean(col));break;default:cell.SetCellValue(reader.GetValue(col).ToString());break;}}}rowIndex++;}// 最终写入文件（使用异步提升性能）using var fs = new FileStream(filePath, FileMode.Create, FileAccess.Write, FileShare.None,81920, true);// 分块写入策略var writeTask = Task.Run(() => workbook.Write(fs, false));// 进度刷新控制while (!writeTask.IsCompleted){if (fs.Position % (10 * bufferSize) == 0){await fs.FlushAsync(); // 异步刷新缓冲区}await Task.Delay(50); // 减少CPU占用}await writeTask;return true;}catch (Exception ex){return false;}}
}

using Microsoft.Data.SqlClient;
using (DbDataReader reader = await cmd.ExecuteReaderAsync())
{return  await ExcelHelper.OutPut(reader, extra);
}
//查询数据后直接塞进去就行

DBReader=>DataTable=>json=>datable =>excel 通过API 传输过于占用资源，
优化：
DBReader=>excel ,直接输出到服务器共享文件夹。
流输出效率更改、省略传输转换、异步内存清除、百万条分Sheet兼容2007。

三、分页查询 DbReader=>Excel (MiniExcel)

using System.Data;
using System.Data.SqlClient;
using System.Configuration;
using MiniExcelLibs;
namespace ExcelExport
{public class ExcelHelper{public void GetData(){SqlConnection sqlCon;try{var sheets = new Dictionary<string, object>();string excelFilePath = @"D:\Excel\DataFile.xlsx";string connectionString = ConnectionString;SqlCommand sqlcmd;sqlCon = new SqlConnection(connectionString);sqlCon.Open();sqlcmd = new SqlCommand(sql, sqlCon);//先取页数 =符合条件总条数/每页条数 maxLoopCounter = Convert.ToInt32(sqlcmd.ExecuteScalar());for (int i = 1; i <= maxLoopCounter; i++) //循环查询每页数据{string sheetName = "Sheet" + i.ToString();//  string sheetName = "Sheet5" ;sqlCon = new SqlConnection(connectionString);sqlCon.Open();string sql = "";sqlcmd = new SqlCommand(sql, sqlCon);sheets.Add(sheetName, sqlcmd.ExecuteReader());//按页写入sheet}MiniExcel.SaveAs(excelFilePath, sheets);}catch (Exception exception){}}}
}

SELECT *
FROM BigDataTable
ORDER BY CreateDate DESC
OFFSET @PageIndex * @PageSize ROWS
FETCH NEXT @PageSize ROWS ONLY; 

可惜的是FETCH 只支持SQL2012+,低版本就只有转存实体表的方式：

declare @ntile_value int =1000000 //一百万一页
SELECT id, NTILE(@ntile_value) OVER (ORDER BY id) AS page  INTO temp_BigTable FROM BigTable;
--使用 NTILE 添加页码字段->写入新表
--个人觉得建表、清表、表占用判断，也挺麻烦

总结：

服务器流导出更快，使用NPOI、和MiniExcel等都是异曲同工。数据量百万级，我觉得应用层分页更好，数据更大那数据库分页后传输更好。Interop作为微软官方com组件确实更方便，但跨平台就不行了。总的来说NPOI确实各方面更适用一些。