可參考本文章:

https://dba.stackexchange.com/questions/245990/postgresql-extremely-slow-count-with-index-simple-query

vacuum analyze insider_event;
執行了 7 分鐘

ALTER TABLE insider_event SET (autovacuum_vacuum_scale_factor = 0, autovacuum_analyze_scale_factor = 0,  autovacuum_vacuum_threshold = 10000, autovacuum_analyze_threshold = 10000);

跑了 1.2 分鐘

修改 work_mem:
https://www.pgmustard.com/blog/work-mem

記憶體調整
https://www.enterprisedb.com/postgres-tutorials/how-tune-postgresql-memory

ubuntu安裝時依照版本不同有些會需要先配置mirror site
mirror site的功用是提供下載套件與系統更新的站台
而要連網確認mirror site會需要配置網路接口,DNS與開放防火牆
17.10以後版本可前往 /etc/netplan 底下的.yaml文件內查看設置
相關參數說明↓
addresses: 對外ip (ex: 192.168.1.10/24)
gateway4: 網路關口, IPv4從本機連結網際網路的端口, 通常設定路由器的IP地址
nameservers: 底下的 addresses為DNS位置, 可設置 8.8.8.8 (google的公共服務器)
設置完後執行 sudo netplan apply變更配置
17.10以前的自己去問GPT, 配置差不多只是位置不一樣

輸入'systemd-resolve --status'查看配置
輸入'ping google.com'測試網路是否正常

以上配置完畢 防火牆有開ubuntu以後可以配置mirror site↓
ubuntu官方承認的Site清單
*注意!! 更新封包皆會從設定的Site取得, 雖是官方認證但是仍有風險, 有高度資安需求建議自行架設mirror site*
mirror site的配置於 /etc/apt/sources.list
'deb http://{domain}' 的皆需修改
可使用 ctrl + \ 來一次修改
完成後執行 'sudp apt-get update' 測試mirror site是否連通

PromiseAll: async function (array) {

    let taskList = [];

    let propList = [];

    for (let obj of array) {

        for (let prop in obj) {

            taskList.push(obj[prop]);

            propList.push(prop);

        }

    }

    let resp = await Promise.all(taskList);

    let result = {};

    let counter = 0;

    for (let prop of propList) {

        result[prop] = resp[counter];

        counter += 1;

    }

    return result;

}

PromiseAll 是將API一次發送並接收回傳值的工具
使用方法範例:

const taskList = [{

    adPosition : BannerPositionDataService.GetList(),

    big : BannerDataService.GetList(100),

    smallTop : BannerDataService.GetOne(200),

    smallBottom : BannerDataService.GetOne(300),

    section2 : BannerDataService.GetList(400),

    section3 : BannerDataService.GetList(500),

    section4 : BannerDataService.GetList(600),

    recommends : ProductDataService.GetRndList()

}];



let resps = await UJ.PromiseAll(taskList);

要特別記得TaskList中的Method不需要做await,不然就沒有意義了

DeepBinding: function (vueData, data) {



    if (Array.isArray(data)) {

        if (!Array.isArray(vueData)) {

            vueData = [];

        } else {

            vueData.splice(0);

        }

        for (let prop in data) {

            vueData.push(data[prop]);

        }

    }

    else if (typeof (data) === 'object') {

        if (Object.keys(data).length === 0) {

            return;

        }

        for (let prop in data) {

            if (vueData[prop] === undefined || data[prop] === null ||

                (!Array.isArray(vueData[prop] && vueData !== null && typeof (data) !== 'object'))) {

                vueData[prop] = data[prop];

            } else {

                this.DeepBinding(vueData[prop], data[prop]);

            }

        }

    } else {

        vueData = data;

    }

}

在資料回傳後要Binding到Vue Data上面或是任意Object的Property上可以使用這個,
不使用DeepBinding是因為瀏覽器版本限制,這個方式不受瀏覽器版本限制,但是只有提供一階,需要多階請自行改寫.
使用方法範例:

UJ.DeepBinding(this, resp);

這邊的this代表的是Vue的Data
 

Entity Framework 提供了 ExecuteSqlRawAsync 和 FromSqlRaw  之後，可以和 Dapper 非常類似的用法。

我們在用 Dapper 時。最常用的就是 sql command 加上一個物件做為參數，就可以執行 CRUD 的動作。

其實用 Entity Framework 的 ExecuteSqlRawAsync 和 FromSqlRaw 也可以逹到幾乎一樣的效果。

ExecuteSqlRawAsync 和 FromSqlRaw 接受的參數是 object array (其實是 Microsoft.Data.SqlClient.SqlParameter 的 array)

所以我們先做一個 Object to Microsoft.Data.SqlClient.SqlParameter Array 的擴充, 可參考: https://gist.github.com/aliozgur/75182b2e9b0a58b83443

不過很奇怪的是，原作者提供的擴充轉出來的會是 System.Data.SqlClient.SqlParameter  Array 無法直接使用於 ExecuteSqlRawAsync 和 FromSqlRaw，所以要稍微改一下，把 using System.Data.SqlClient; 改為 using Microsoft.Data.SqlClient; 即可:



另外，我們再自行 對 DbContext 做一個擴充如下:

        /// <summary>

        /// 會把物件 Parameter 的各 Property 帶入 SQL 中.

        /// </summary>

        /// <param name="dbct"></param>

        /// <param name="sql"></param>

        /// <param name="parameters"></param>

        /// <returns></returns>

        public static async Task<int> ExecuteParameterSqlAsync(this DbContext dbct, string sql, object parameters)

        {

            return await dbct.Database.ExecuteSqlRawAsync(sql, parameters.ToSqlParamsArray());

        }

最後的結果可以做到以下的效果: (res3 和 res4 是配合 FormattableString 的範例 )

 
範例程式碼:

        public async Task<object> EfTest()

        {

            var dbct = Ds.NewContext.GvContext;

            var insertSql = @"Insert into ProfileEvent(JobId,[Name],GvShoplineId,LiShoplineId,Phone,Email,LineId,GaClientId)

            Values(@JobId, @Name, null , null , null , @Email, @GaClientId , @GaClientId )";



            //執行 SQL 的原生寫法

            await dbct.Database.ExecuteSqlRawAsync(insertSql, new object[]

            {

                new SqlParameter("@JobId", 10),

                new SqlParameter("@Email", "XX'TT"),

                new SqlParameter("@Name", "AA'BB"),

                new SqlParameter("@GaClientId", "GaClientId"),

            });



            //執行 SQL 的擴充寫法

            await dbct.ExecuteParameterSqlAsync(insertSql, new

            {

                JobId = 10,

                Email = "XX'TT",

                Name = "AA'BB",

                GaClientId = "GaClientId"

            });



            //關於查詢

            string name = "%'B%";

            //原生寫法

            var res1 = await dbct.ProfileEvents.FromSqlRaw("Select top 100 * from ProfileEvent where Name like @name",

                new object[]

                    {

                        new SqlParameter("@name", "%'B%")

                    })

                .ToListAsync();



            //擴充, object to SqlParameter array

            var res2 = await dbct.ProfileEvents.FromSqlRaw("Select top 100 * from ProfileEvent where Name like @name",

                new { name }.ToSqlParamsArray())

                .ToListAsync();



            var res3 = await dbct.ProfileEvents.FromSql($"Select top 100 * from ProfileEvent where Name like {name}")

                .ToListAsync();



            var res4 = await dbct.ProfileEvents.FromSqlInterpolated($"Select top 100 * from ProfileEvent where Name like {name}")

                .ToListAsync();



            return new { res1, res2, res3, res4 };

        }

 

 
正常情形下，不應該出現需瀏覽目錄的錯誤訊息
如遇此錯誤，檢查IIS的IP位址及網域限制設定是否跑掉。
 
  
 

這個指令還是會把所有的資料 Select 出來，再更新

原指令:

UPDATE Job Set En_Status = 200 Where En_Status = 100 and LastTouchAt < '2023-05-06 12:34:56'
其中  '2023-05-06 12:34:56' 是 DateTime.Now.AddMinutes(-2) 的結果(Web Server 端的時間扣 2 分鐘)

但，若是改使用 UpdateFromQueryAsync 如下:
var c = await Ds.NewContext.GvContext.Jobs.Where(j => j.En_Status == Cst.Job.Status.Running && j.LastTouchAt < DateTime.Now.AddMinutes(-2))
                .UpdateFromQueryAsync(j => new Ds.Gv.Job { En_Status = Cst.Job.Status.ReStarting });

產生的 SQL 如下:
UPDATE A 
SET A.[En_Status] = @zzz_BatchUpdate_0
FROM [Job] AS A
INNER JOIN ( SELECT [j].[Id], [j].[CancelledAt], [j].[CancelledBy], [j].[En_Status], [j].[EndAt], [j].[Exception], [j].[Filename], [j].[InformationJson], [j].[InitAt], [j].[Is_CheckOnly], [j].[LastTouchAt], [j].[LastTouchMessage], [j].[LoopStartAt], [j].[Name], [j].[ScheduleId], [j].[TotalTouch], [j].[TouchCount]
FROM [Job] AS [j]
WHERE [j].[En_Status] = 100 AND [j].[LastTouchAt] < DATEADD(minute, CAST(-2.0E0 AS int), GETDATE())
           ) AS B ON A.[Id] = B.[Id]

有兩個要注意的地方:
1. 它會先 Select 全欄位，再做更新

2. 它的時間是 DB Server 的現在時間。不是 Web Server 端的時間。


順便記錄一下。若是要執行 Update xx Ser cc = cc + 1 Where ...

EF 可寫為:
var c = await Ds.NewContext.GvContext.Jobs.Where(j => j.En_Status == Cst.Job.Status.Running && j.LastTouchAt < DateTime.Now.AddMinutes(-2))
                .UpdateFromQueryAsync(j => new Ds.Gv.Job { TotalTouch = j.TotalTouch + 1 });

轉換的 SQL 為:
UPDATE A 
SET A.[TotalTouch] = B.[TotalTouch] + 1
FROM [Job] AS A
INNER JOIN ( SELECT [j].[Id], [j].[CancelledAt], [j].[CancelledBy], [j].[En_Status], [j].[EndAt], [j].[Exception], [j].[Filename], [j].[InformationJson], [j].[InitAt], [j].[Is_CheckOnly], [j].[LastTouchAt], [j].[LastTouchMessage], [j].[LoopStartAt], [j].[Name], [j].[ScheduleId], [j].[TotalTouch], [j].[TouchCount]
FROM [Job] AS [j]
WHERE [j].[En_Status] = 100 AND [j].[LastTouchAt] < DATEADD(minute, CAST(-2.0E0 AS int), GETDATE())
           ) AS B ON A.[Id] = B.[Id]

 
在註冊Swagger時如圖中
1. 將各自action名稱代用不同編碼，避免名稱相撞導致swagger不能正確辨識。
2. 將專案XML輸出改為專案名稱($(AssemblyName))供註冊時指定反射用
↓ (對專案點擊右鍵 => 屬性 => 生成(Build) => 輸出(Output) => XML 文件名稱(XML documentation file path))
 
 

Console主控台的輸出，只要是非英文，都會是亂碼呈現



情況：
 
解決方法：
輸入：[Console]::OutputEncoding = [System.Text.Encoding]::UTF8



 
 
參考：
一次解決CONSOLE主控台輸出中文亂碼的編碼問題
【茶包射手日記】PowerShell 串接 EXE 輸出中文變亂碼

產生 FileStreamResult 物件的 function 如下: (目前置於 SU 之中，以便轉移)

        static List<Brush> CaptchaBrushes = null;

        public static FileStreamResult CreateCaptcha(string captcha)

        {

            if (CaptchaBrushes == null)

            {

                CaptchaBrushes = new List<Brush>();

                CaptchaBrushes.Add(Brushes.White);

                CaptchaBrushes.Add(Brushes.Gold);

                CaptchaBrushes.Add(Brushes.LightSkyBlue);

                CaptchaBrushes.Add(Brushes.LimeGreen);

                CaptchaBrushes.Add(Brushes.AliceBlue);

                CaptchaBrushes.Add(Brushes.AntiqueWhite);

                CaptchaBrushes.Add(Brushes.BurlyWood);

                CaptchaBrushes.Add(Brushes.Silver);

            }



            int width = 90;

            int height = 45;



            //https://stackoverflow.com/questions/61365732/cannot-access-a-closed-stream-when-returning-filestreamresult-from-c-sharp-netc

            //Using statements close and unload the variable from memory set in the using statement which is why you are getting an error trying to access a closed memory stream.You don't need to use a using statement if you are just going to return the result at the end.



            //這個 memory stream 不用關閉或 dispose

            var ms = new MemoryStream();



            // 釋放所有在 GDI+ 所佔用的記憶體空間 ( 非常重要!! )

            using (Bitmap _bmp = new Bitmap(width, height))

            using (Graphics _graphics = Graphics.FromImage(_bmp))

            using (Font _font = new Font("Courier New", 24, FontStyle.Bold)) // _font 設定要出現在圖片上的文字字型、大小與樣式

            {

                // (封裝 GDI+ 繪圖介面) 所有繪圖作業都需透過 Graphics 物件進行操作

                _graphics.Clear(Color.Black);



                // 如果想啟用「反鋸齒」功能，可以將以下這行取消註解

                //_graphics.TextRenderingHint = TextRenderingHint.AntiAlias;



                // 將亂碼字串「繪製」到之前產生的 Graphics 「繪圖板」上

                var x = 10;

                for(var i = 0; i < captcha.Length; i++)

                {

                    _graphics.DrawString(captcha.Substring(i, 1), _font, CaptchaBrushes[Su.MathUtil.GetRandomInt(CaptchaBrushes.Count)], x, Su.MathUtil.GetRandomInt(15));

                    x += 10 + Su.MathUtil.GetRandomInt(10);

                }



                // 畫線

                

                _graphics.DrawLine(new Pen(CaptchaBrushes[Su.MathUtil.GetRandomInt(CaptchaBrushes.Count)], 1),

                    Su.MathUtil.GetRandomInt(0, Convert.ToInt32((width * 0.9 / 2))), 0, Su.MathUtil.GetRandomInt(Convert.ToInt32(width / 2), Convert.ToInt32(width * 1.9 / 2)), height);

                

                _graphics.DrawLine(new Pen(CaptchaBrushes[Su.MathUtil.GetRandomInt(CaptchaBrushes.Count)], 1),

                    Su.MathUtil.GetRandomInt(Convert.ToInt32(width / 2), Convert.ToInt32(width * 1.9 / 2)), 0, Su.MathUtil.GetRandomInt(0, Convert.ToInt32((width * 0.9 / 2))), height);

                

                _graphics.DrawLine(new Pen(CaptchaBrushes[Su.MathUtil.GetRandomInt(CaptchaBrushes.Count)], 1),

                    0,

                    Su.MathUtil.GetRandomInt(height / 2),

                    width,

                    height / 2 + Su.MathUtil.GetRandomInt(height / 2)

                    );



                _graphics.DrawLine(new Pen(CaptchaBrushes[Su.MathUtil.GetRandomInt(CaptchaBrushes.Count)], 1),

                 0,

                 height / 2 + Su.MathUtil.GetRandomInt(height / 2),

                 width,

                    Su.MathUtil.GetRandomInt(height / 2)

                 );



                _bmp.Save(ms, System.Drawing.Imaging.ImageFormat.Jpeg);

            }



            ms.Seek(0, SeekOrigin.Begin);



            // Controller 的型別為 FileResult

            return new FileStreamResult(ms, "image/jpeg")

            { FileDownloadName = $"{DateTime.Now.Ymdhmsf()}.jpg" };

        }

輸出用的 Controller 和 Action 如下:

namespace Web.Controllers

{

    public class CaptchaController : Controller

    {

        [Route("captcha")]

        public async Task<FileStreamResult> Index()

        {

            //產生 Captcha 並存入 Session 之中。目前是四位數字

            string captcha = (await Ah.ReGetAsync<object>("api/kol/create-captcha-code")).ToString();



            //產生圖檔並回傳 FileStreamResult

            return Su.Wu.CreateCaptcha(captcha);

        }

    }

}

在 Postgres 使用 TransactionScope 似乎很空易就會發生 Transaction Abort 的情況。所以必需自行實做 retry 的機制

            int tries = 0;

            while (tries < Su.PgSql.maxTransactionAborted)

            {

                try

                {

                    using (var scop = new System.Transactions.TransactionScope(TransactionScopeAsyncFlowOption.Enabled))

                    {

                        var context = ShopBandContext.Context;



                        updateCount = await context.UpdateAsync<MarketDiscount>(dto, Sc.ModifyInfo,

                            onlyColumns: "StartAt,EndAt,NewPrice,FrontEndMemo,BackEndMemo");



                        scop.Complete();

                        break; //這個 Break 很重要

                    }

                }

                catch (Exception ex)

                {

                    // 我發現因為是非同步模式，所以要檢查很多層的 InnerException

                    if (tries < Su.PgSql.maxTransactionAborted

                        && ((ex.InnerException != null && ex.InnerException is Npgsql.PostgresException && ((Npgsql.PostgresException)ex.InnerException).SqlState == "40001")

                            || (ex.InnerException.InnerException != null && ex.InnerException.InnerException is Npgsql.PostgresException && ((Npgsql.PostgresException)ex.InnerException.InnerException).SqlState == "40001")))

                    {

                        //隨機休息 10 ~ 20 ms, 等另一個 job 完工

                        System.Threading.Thread.Sleep(Su.MathUtil.Random.Next(10, 20));

                        tries++;

                    }

                    else

                    {

                        throw;

                    }

                }

            }