Performance of Adptive Card Template Rendering (C#)

Adaptive Card 模板渲染

October 8, 2022

测试场景

渲染一个卡片列表。三种方式

  1. 使用 json 模板渲染单个卡片,代码循环拼接列表
  2. 匿名 Object 循环拼接
  3. 纯模板渲染 List 生成

理论上, 模板替换 json 字符串是最快的,但是解析模板循环是最慢的。

模板渲染卡片

卡片模板

{
    "$schema": "http://adaptivecards.io/schemas/adaptive-card.json",
    "type": "AdaptiveCard",
    "body": [
        {
            "altText": "${alt}",
            "url": "${src}",
            "type": "Image"
        }
    ]
}

渲染逻辑

public static MessagingExtensionResponse CardLoop(IEnumerable<Img> images)
    {
        List<MessagingExtensionAttachment> attachments = images
            .Select(
                img =>
                    new MessagingExtensionAttachment()
                    {
                        ContentType = "application/vnd.microsoft.card.adaptive",
                        Content = JObject.Parse(CardTemplate.Expand(img)), // template
                        Preview = new Attachment() { }
                    }
            )
            .ToList();

        return new MessagingExtensionResponse
        {
            ComposeExtension = new MessagingExtensionResult("grid", "result", attachments),
        };
    }

纯粹对象拼接 

 public static MessagingExtensionResponse PureObject(IEnumerable<Img> images)
    {
        List<MessagingExtensionAttachment> attachments = images
            .Select(
                img =>
                    new MessagingExtensionAttachment()
                    {
                        ContentType = "application/vnd.microsoft.card.adaptive",
                        Content = JObject.FromObject(
                            new
                            {
                                type = "AdaptiveCard",
                                body = new[]
                                {
                                    new
                                    {
                                        altText = img.Alt,
                                        url = img.Src,
                                        type = "Image",
                                    }
                                }
                            }
                        ),
                        Preview = new ThumbnailCard()
                        {
                            Images = new List<CardImage>() { new CardImage(img.Src, img.Alt) }
                        }.ToAttachment()
                    }
            )
            .ToList();

        return new MessagingExtensionResponse
        {
            ComposeExtension = new MessagingExtensionResult("grid", "result", attachments),
        };
    }

模板渲染列表

列表模板

{
    "attachments": [
        {
            "$data": "${Imgs}",
            "content": {
                "type": "AdaptiveCard",
                "body": [
                    {
                        "altText": "${alt}",
                        "url": "${src}",
                        "type": "Image"
                    }
                ],
                "$schema": "http://adaptivecards.io/schemas/adaptive-card.json",
                "version": "1.5"
            },
            "contentType": "application/vnd.microsoft.card.adaptive",
            "preview": {
                "content": {
                    "images": [
                        {
                            "alt": "${alt}",
                            "url": "${src}"
                        }
                    ]
                },
                "contentType": "application/vnd.microsoft.card.thumbnail"
            }
        }
    ],
    "type": "result",
    "attachmentLayout": "grid"
}

渲染逻辑

 public static MessagingExtensionResponse FromTempateList(IEnumerable<Img> images)
    {
        var res = JObject.Parse(ListTemplate.Expand(new {Imgs:images}));
        return new MessagingExtensionResponse
        {
            ComposeExtension = res.ToObject<MessagingExtensionResult>(),
        };
    }

Benchmark result 


BenchmarkDotNet=v0.13.2, OS=Windows 10 (10.0.19044.2006/21H2/November2021Update)
Intel Xeon CPU E5-1650 v4 3.60GHz, 1 CPU, 12 logical and 6 physical cores
.NET SDK=6.0.401
  [Host]     : .NET 6.0.9 (6.0.922.41905), X64 RyuJIT AVX2
  DefaultJob : .NET 6.0.9 (6.0.922.41905), X64 RyuJIT AVX2
Method Mean Error StdDev
MixedObject 26,301.74 μs 496.889 μs 531.666 μs
PureObject 38.26 μs 0.759 μs 1.267 μs
PureTemplate 58,808.77 μs 1,139.154 μs 1,481.222 μs

结果上看，

  1. 纯Object的拼接新能上最佳(代码逻辑，效率最高)
  2. 直接模板渲染N张卡片的列表,比N慢一倍。（数组模板效率最低）