C# 線上壓縮解壓縮實戰：SharpZipLib 串流處理大檔案不落地

場景二：物件儲存 → 即時解壓 → HTTP 下載（串流回應）

[HttpGet("download-decompress/{objectKey}")]
public async Task DecompressAndDownload(string objectKey, CancellationToken ct)
{
    // 1. 取得上游壓縮串流
    await using Stream srcStream = await _storage.OpenReadAsync(objectKey, ct);

    // 2. 設定回應標頭
    string fileName = Path.GetFileNameWithoutExtension(objectKey); // 移除 .gz
    Response.Headers.ContentDisposition = $"attachment; filename=\"{fileName}\"";
    Response.Headers.ContentType = "application/octet-stream";

    // 3. 包裝解壓縮串流 → 直接複製到 HTTP 回應 Body
    await using var gzIn = new GZipInputStream(srcStream);
    await gzIn.CopyToAsync(Response.Body, 81920, ct);
}

此模式適用於任何 Stream 來源：FileStream、MemoryStream、S3Object.ResponseStream、BlobClient.OpenRead()。

場景三：多檔案打包成 Zip 串流下載（不建立暫存 Zip）

[HttpGet("download-zip")]
public async Task DownloadMultipleAsZip([FromQuery] string[] keys, CancellationToken ct)
{
    Response.Headers.ContentDisposition = "attachment; filename=\"batch.zip\"";
    Response.Headers.ContentType = "application/zip";

    await using var zipOut = new ZipOutputStream(Response.Body) { IsStreamOwner = false };
    zipOut.SetLevel(3); // 快速壓縮，降低 CPU

    foreach (string key in keys)
    {
        // 逐個讀取物件並寫入 Zip Entry
        await using Stream objStream = await _storage.OpenReadAsync(key, ct);
        var entry = new ZipEntry(Path.GetFileName(key))
        {
            DateTime = DateTime.UtcNow,
            Size = objStream.CanSeek ? objStream.Length : -1
        };
        zipOut.PutNextEntry(entry);
        await objStream.CopyToAsync(zipOut, 81920, ct);
        zipOut.CloseEntry();
    }
    await zipOut.FinishAsync(ct); // 寫入中央目錄結構
}

優勢：記憶體佔用恆定 ≈ 緩衝區大小 × 併發數，無論打包多少 GB 檔案都不會 OOM。

進階：非同步列舉器逐筆產出 Zip Entry（.NET 6+ IAsyncEnumerable）

public async IAsyncEnumerable CreateZipEntriesAsync(
    IEnumerable keys, 
    [EnumeratorCancellation] CancellationToken ct)
{
    foreach (string key in keys)
    {
        ct.ThrowIfCancellationRequested();
        await using Stream s = await _storage.OpenReadAsync(key, ct);
        yield return new ZipEntry(Path.GetFileName(key))
        {
            DateTime = DateTime.UtcNow,
            Size = s.CanSeek ? s.Length : -1
        };
        // 注意：這裡只產生 Entry metadata，實際資料複製在呼叫端
    }
}

// 呼叫端
await using var zipOut = new ZipOutputStream(Response.Body) { IsStreamOwner = false };
await foreach (var entry in CreateZipEntriesAsync(keys, ct))
{
    zipOut.PutNextEntry(entry);
    await using Stream s = await _storage.OpenReadAsync(entry.Name, ct);
    await s.CopyToAsync(zipOut, 81920, ct);
    zipOut.CloseEntry();
}
await zipOut.FinishAsync(ct);

效能與資源管理檢核表

替代方案快速比較

掌握「串流包裝串流、CopyToAsync 穿透、IsStreamOwner 控制權」三大核心模式，即能在任何 .NET 服務中構建高效能、零暫存檔、可水平擴展的線上壓縮解壓縮管線，滿足現代雲原生架構的嚴苛資源限制。