Migrando do AnswerHub para o WordPress:um conto de 10 tecnologias

1. Perl
2. SQL Server
3. PowerShell
4. Transmitir (FTP)
5. HTML
6. CSS
7. C#
8. MarkdownSharp
9. phpMyAdmin
10. MySQL

CREATE TABLE `network11_managed_files` (
  `c_id` bigint(20) NOT NULL,
  ...
);
 
INSERT INTO `network11_managed_files` (`c_id`, ...) VALUES (1, ...);

CREATE TABLE dbo.files
(
  c_id bigint NOT NULL,
  ...
);
 
INSERT dbo.files (c_id, ...) VALUES (1, ...);

SELECT 'Invoke-WebRequest -Uri '
  + '"$($url)' + RTRIM(c_id) + '-' + c_name + '"'
  + ' -OutFile "E:\s\temp\' + RTRIM(c_id) + '-' + c_name + '";'
  FROM dbo.files
  WHERE LOWER(c_mime_type) LIKE 'application/%';

$AllProtocols = [System.Net.SecurityProtocolType]'Ssl3,Tls,Tls11,Tls12';
[System.Net.ServicePointManager]::SecurityProtocol = $AllProtocols;
$u = "https://answers.sqlperformance.com/s/temp/";
 
Invoke-WebRequest -Uri "$($u)/1-proc.pesession"   -OutFile "E:\s\temp\1-proc.pesession";
Invoke-WebRequest -Uri "$($u)/14-test.pesession"  -OutFile "E:\s\temp\14-test.pesession";
Invoke-WebRequest -Uri "$($u)/15-a.QueryAnalysis" -OutFile "E:\s\temp\15-a.QueryAnalysis";
...

SELECT c_type, c_id, c_parent, oParent = c_originalParent, c_creation_date, c_title
  FROM dbo.nodes 
  WHERE c_originalParent = 285;
 
/*
c_type      c_id    c_parent  oParent  c_creation_date   accepted  c_title
----------  ------  --------  -------  ----------------  --------  -------------------------
question    285     NULL      285      2013-02-13 16:30            why is the MERGE JOIN ...
answer      287     285       285      2013-02-14 01:15  1         NULL
comment     289     285       285      2013-02-14 13:35            NULL
answer      293     285       285      2013-02-14 18:22            NULL
comment     294     287       285      2013-02-14 18:29            NULL
comment     298     285       285      2013-02-14 20:40            NULL
comment     299     298       285      2013-02-14 18:29            NULL
*/

/*
c_type        c_id    c_parent  oParent  c_creation_date   reason
----------    ------  --------  -------  ----------------  -------------------------
question      285     NULL      285      2013-02-13 16:30  question is ALWAYS first
++comment     289     285       285      2013-02-14 13:35  comments on the question before answers
answer        287     285       285      2013-02-14 01:15  first answer (accepted = 1)
++comment     294     287       285      2013-02-14 18:29  first comment on first answer
++comment     298     287       285      2013-02-14 20:40  second comment on first answer
++++comment   299     298       285      2013-02-14 18:29  reply to second comment on first answer
answer        293     285       285      2013-02-14 18:22  second answer
*/

DECLARE @foo TABLE
(
  c_type varchar(255), 
  c_id int, 
  c_parent int, 
  oParent int,
  accepted bit
);
 
INSERT @foo(c_type, c_id, c_parent, oParent, accepted) VALUES
('question', 285, NULL, 285, 0),
('answer',   287, 285 , 285, 1),
('comment',  289, 285 , 285, 0),
('comment',  294, 287 , 285, 0),
('comment',  298, 287 , 285, 0),
('comment',  299, 298 , 285, 0),
('answer',   293, 285 , 285, 0);
 
;WITH cte AS 
(
  SELECT 
    lvl = 0,
    f.c_type,
    f.c_id, f.c_parent, f.oParent,
    Sort = CONVERT(varchar(255),RIGHT('00000' + CONVERT(varchar(5),f.c_id),5))
  FROM @foo AS f WHERE f.c_parent IS NULL
  UNION ALL
  SELECT 
    lvl = c.lvl + 1,
    c_type = CONVERT(varchar(255), CASE
        WHEN f.accepted = 1 THEN 'accepted answer'
        WHEN f.c_type = 'comment' THEN c.c_type + ' ' + f.c_type
        ELSE f.c_type
      END),
    f.c_id, f.c_parent, f.oParent,
    Sort = CONVERT(varchar(255),c.Sort + RIGHT('00000' + CONVERT(varchar(5),f.c_id),5))
  FROM @foo AS f INNER JOIN cte AS c ON c.c_id = f.c_parent
)
SELECT lvl = CASE lvl WHEN 0 THEN 1 ELSE lvl END, c_type, c_id, c_parent, oParent, Sort
FROM cte
ORDER BY 
  oParent,
  CASE
    WHEN c_type LIKE 'question%'        THEN 1 -- it's a question *or* a comment on the question
    WHEN c_type LIKE 'accepted answer%' THEN 2 -- accepted answer *or* comment on accepted answer
    ELSE 3 END,
  Sort;

/*
lvl  c_type                            c_id        c_parent    oParent     Sort
---- --------------------------------- ----------- ----------- ----------- --------------------
1    question                          285         NULL        285         00285               
1    question comment                  289         285         285         0028500289          
1    accepted answer                   287         285         285         0028500287          
2    accepted answer comment           294         287         285         002850028700294     
2    accepted answer comment           298         287         285         002850028700298     
3    accepted answer comment comment   299         298         285         00285002870029800299
1    answer                            293         285         285         0028500293     
*/

<div class="question">
  <span class="authorq" title=" Author : author name ">
    <i class="fas fa-user"></i>Author name</span> 
  <span class="createdq" title=" February 13th, 2013 ">
    <i class="fas fa-calendar-alt"></i>2013-02-13 16:30:36</span>
 
  <div class=mainbodyq>I don't understand why the merge operator is passing over 4million 
  rows to the hash match operator when there is only 41K and 19K from other operators.
 
	<div class=attach><i class="fas fa-file"></i>
	  <a target="_blank" href="/s/temp/254-tmp4DA0.queryanalysis" rel="noopener noreferrer">
      /s/temp/254-tmp4DA0.queryanalysis</a>
	</div>
  </div>
 
  <div class="comment indent1 ">
    <div class=linecomment>
	  <span class="authorc" title=" Author : author name ">
	    <i class="fas fa-user"></i>author name</span>
	  <span class="createdc" title=" February 14th, 2013 ">
	    <i class="fas fa-calendar-alt"></i>2013-02-14 13:35:39</span>
	</div>
    <div class=mainbodyc>
	  I am still trying to understand the significant amount of rows from the MERGE operator. 
	  Unless it's a result of a Cartesian product from the two inputs then finally the WHERE 
	  predicate is applied to filter out the unmatched rows leaving the 4 million row count.
    </div>
  </div>
  <div class="answer indent1 [accepted]">
    <div class=lineanswer>
	  <span class="authora" title=" Author : author name ">
	    <i class="fas fa-user"></i>author name</span>
	  <span class="createda" title=" February 14th, 2013 ">
	    <i class="fas fa-calendar-alt"></i>2013-02-14 01:15:42</span>
	</div>
    <div class=mainbodya>
	    The reason for the large number of rows can be seen in the Plan Explorer tool tip for 
		the Merge Join operator:
 
	    <img src="/s/temp/259-sp.png" alt="Merge Join tool tip" />
	  	...
	</div>
  </div>
</div>

CREATE /* OR ALTER */ PROCEDURE dbo.BuildMySQLInserts
  @LowerBound int = 1, 
  @UpperBound int = 750
AS
BEGIN
  SET NOCOUNT ON;
 
  ;WITH CTE AS 
  (
    SELECT lvl = 0,
            [type] = CONVERT(varchar(100),f.[type]),
            f.id,
            f.parent,
            f.master_parent,
            created = CONVERT(char(10), f.created, 120) + ' ' 
			        + CONVERT(char(8),  f.created, 108),
            f.state,
            Sort = CONVERT(varchar(100),RIGHT('0000000000' 
			     + CONVERT(varchar(10),f.id),10))
    FROM dbo.foo AS f
    WHERE f.type = 'question' 
      AND master_parent BETWEEN @LowerBound AND @UpperBound
    UNION ALL
    SELECT lvl = c.lvl + 1,
            CONVERT(varchar(100),CASE
                WHEN f.[state] = '[accepted]' THEN 'accepted answer'
                WHEN f.type = 'comment' THEN c.type + ' ' + f.type
                ELSE f.type
            END),
            f.id,
            f.parent,
            f.master_parent,
            created = CONVERT(char(10), f.created, 120) + ' ' 
			        + CONVERT(char(8), f.created, 108),
            f.state,
            Sort = CONVERT(varchar(100),c.sort + RIGHT('0000000000' 
			     + CONVERT(varchar(10),f.id),10))
    FROM dbo.foo AS f
    JOIN CTE AS c ON c.id = f.parent
)
SELECT 
  master_parent, 
  prefix = CASE WHEN lvl = 0 THEN 
    CONVERT(varchar(11), master_parent) + ', 3, ''' + created + ''', ''' 
	+ created + ''',''' END, 
  bodypre = '<div class="' + COALESCE(c_type, RTRIM(LEFT([type],8))) 
	  + CASE WHEN c_type <> 'question' THEN ' indent' + RTRIM(lvl) 
	  + COALESCE(' ' + [state], '') ELSE '' END + '">'
	  + CASE WHEN c_type <> 'question' THEN 
	    '<div class=line' + c_type + '>' ELSE '' END 
	  + '<span class="author' + LEFT(c_type, 1) + '" title=" Author : ' 
	  + REPLACE(REPLACE(Fullname,'''','\'''),'"','') 
	  + ' "><i class="fas fa-user"></i>' + REPLACE(Fullname,'''','\''') --"
	  + '</span> <span class="created' + LEFT(c_type,1) + '" title=" ' 
	  + DATENAME(MONTH, c_creation_date) + ' ' + RTRIM(DAY(c_creation_date)) 
	  + CASE 
        WHEN DAY(c_creation_date) IN (1,21,31) THEN 'st'
        WHEN DAY(c_creation_date) IN (2,22) THEN 'nd'
        WHEN DAY(c_creation_date) IN (3,23) THEN 'rd' ELSE 'th' END
        + ', ' + RTRIM(YEAR(c_creation_date)) 
      + ' "><i class="fas fa-calendar-alt"></i>' + created + '</span>'
      + CASE WHEN c_type <> 'question' THEN '</div>' ELSE '' END,
  body = '<div class=mainbody' + left(c_type,1) + '>' 
	  + REPLACE(REPLACE(c_body, char(39), '\' + char(39)), '’', '\' + char(39)),
  bodypost = COALESCE(urls, '') + '</div></div>',--' 
	  + CASE WHEN c_type = 'question' THEN '</div>' ELSE '' END, 
  suffix = ''',''' + REPLACE(n.c_title, '''', '\''') + ''','''',''publish'',
	  ''closed'',''closed'','''',''' + REPLACE(n.c_plug, '''', '\''') 
	  + ''','''','''',''' + created + ''',''' + created + ''','''',0,
	  ''https://answers.sqlperformance.com/?p=' + CONVERT(varchar(11), master_parent) 
	  + ''', 0, ''post'','''',0);',
  rn = RTRIM(ROW_NUMBER() OVER (PARTITION BY master_parent 
      ORDER BY master_parent,
      CASE
        WHEN [type] LIKE 'question%' THEN 1
        WHEN [type] LIKE 'accepted answer%' THEN 2
        ELSE 3
      END,
      Sort)), 
  c = RTRIM(COUNT(*) OVER (PARTITION BY master_parent))
FROM CTE
LEFT OUTER JOIN dbo.network11_nodes AS n
ON cte.id = n.c_id
LEFT OUTER JOIN dbo.Users AS u
ON n.c_author = u.UserID
LEFT OUTER JOIN 
(
  SELECT NodeID, urls = STRING_AGG('<div class=attach>
    <i class="fas fa-file' 
	+ CASE WHEN c_mime_type IN ('image/jpeg','image/png') 
      THEN '-image' ELSE '' END 
    + '"></i><a target="_blank" href=' + url + ' rel="noopener noreferrer">' + url + '</a></div>', '\n') 
  FROM dbo.Attachments 
  GROUP BY NodeID
) AS a
ON n.c_id = a.NodeID
ORDER BY master_parent,
  CASE
    WHEN [type] LIKE 'question%' THEN 1
    WHEN [type] LIKE 'accepted answer%' THEN 2
    ELSE 3
  END,
  Sort;
END
GO

SELECIONE algo de dbo.sometable;

• MarkdownSharp não permite coisas como target=_blank , então eu mesmo teria que injetá-los após o processamento;
• código (qualquer coisa prefixada com quatro espaços) herda os wrappers

  using System.Text;
  using System.Data;
  using System.Data.SqlClient;
  using MarkdownSharp;
  using System.IO;
   
  namespace AnswerHubMigrator
  {
    class Program
    {
      static void Main(string[] args)
      {
        StringBuilder output;
        string suffix = "";
        string thisfile = "";
   
        // pass two arguments on the command line, e.g. 1, 750
        int LowerBound = int.Parse(args[0]);
        int UpperBound = int.Parse(args[1]);
   
        // auto-expand URLs, and only accept bold/italic markdown
        // when it completely surrounds an entire word
        var options = new MarkdownOptions
        {
          AutoHyperlink = true,
          StrictBoldItalic = true
        };
        MarkdownSharp.Markdown mark = new MarkdownSharp.Markdown(options);
   
        using (var conn = new SqlConnection("Server=.\\SQL2017;Integrated Security=true"))
        using (var cmd = new SqlCommand("MigrateDB.dbo.BuildMySQLInserts", conn))
        {
   
          cmd.CommandType = CommandType.StoredProcedure;
          cmd.Parameters.Add("@LowerBound", SqlDbType.Int).Value = LowerBound;
          cmd.Parameters.Add("@UpperBound", SqlDbType.Int).Value = UpperBound;
          conn.Open();
          using (var reader = cmd.ExecuteReader())
          {
            // use a StringBuilder to dump output to a file
            output = new StringBuilder();
            while (reader.Read())
            {
              // on first pass, make a new delete/insert
              // delete is to make the commands idempotent
              if (reader["rn"].Equals("1"))
              {
   
                // for each master parent, I would create a
                // new WordPress post, inheriting the parent ID
                output.Append("DELETE FROM `wp_posts` WHERE ID = ");
                output.Append(reader["master_parent"].ToString());
                output.Append("; INSERT INTO `wp_posts` (`ID`, `post_author`, ");
                output.Append("`post_date`, `post_date_gmt`, `post_content`, ");
                output.Append("`post_title`, `post_excerpt`, `post_status`, ");
                output.Append("`comment_status`, `ping_status`, `post_password`,");
                output.Append(" `post_name`, `to_ping`, `pinged`, `post_modified`,");
                output.Append(" `post_modified_gmt`, `post_content_filtered`, ");
                output.Append("`post_parent`, `guid`, `menu_order`, `post_type`, ");
                output.Append("`post_mime_type`, `comment_count`) VALUES (");
   
                // I'm sure some of the above columns are optional, but identifying
                // those would not be a valuable use of time IMHO
   
                output.Append(reader["prefix"]);
   
                // hold on to the additional values until last row
                suffix = reader["suffix"].ToString();
              }
   
              // manipulate the body content to be WordPress and INSERT statement-friendly
              string body = reader["body"].ToString().Replace(@"\n", "\n");
              body = mark.Transform(body).Replace("href=", "target=_blank href=");
              body = body.Replace("<p>", "").Replace("</p>", "");
              body = body.Replace("<pre><code>", "<pre lang=\"tsql\">");
              body = body.Replace("</code></"+"pre>", "</"+"pre>");
              body = body.Replace(@"'", "\'").Replace(@"’", "\'");
   
              body = reader["bodypre"].ToString() + body.Replace("\n", @"\n");
              body += reader["bodypost"].ToString();
              body = body.Replace("&lt;", "<").Replace("&gt;", ">");
              output.Append(body);
   
              // if we are on the last row, add additional values from the first row
              if (reader["c"].Equals(reader["rn"]))
              {
                output.Append(suffix);
              }
            }
   
            thisfile = UpperBound.ToString();
            using (StreamWriter w = new StreamWriter(@"C:\wp\" + thisfile + ".sql"))
            {
              w.WriteLine(output);
              w.Flush();
            }
          }
        }
      }
    }
  }

  
  Sim, isso é um monte de código feio, mas finalmente me levou ao conjunto de saída que não faria o phpMyAdmin vomitar, e que o WordPress apresentaria bem (o suficiente). Eu simplesmente chamei o programa C# várias vezes com os diferentes intervalos de parâmetros:
  

  AnswerHubMigrator    1  750
  AnswerHubMigrator  751 1500
  AnswerHubMigrator 1501 2250
  ...

  
  Então eu abri cada um dos arquivos, colei-os no phpMyAdmin e apertei GO:
  phpMyAdmin (clique para ampliar)
  Claro que eu tive que adicionar um pouco de CSS no WordPress para ajudar a diferenciar entre perguntas, comentários e respostas, e também recuar comentários para mostrar respostas a perguntas e respostas, aninhar comentários respondendo a comentários e assim por diante. Aqui está a aparência de um trecho quando você analisa as perguntas de um mês:
  Tijolo de pergunta (clique para ampliar)
  E, em seguida, uma postagem de exemplo, mostrando imagens incorporadas, vários anexos, comentários aninhados e uma resposta:
  Exemplo de pergunta e resposta (clique para ir até lá)
  Ainda estou tentando recuperar algumas postagens que foram enviadas ao site depois que o último backup foi feito, mas convido você a navegar. Por favor, avise-nos se você encontrar algo faltando ou fora do lugar, ou apenas para nos dizer que o conteúdo ainda é útil para você. Esperamos reintroduzir a funcionalidade de upload do plano a partir do Plan Explorer, mas isso exigirá algum trabalho de API no novo site de suporte, portanto, não tenho um ETA para você hoje.
  
  Answers.SQLPerformance.com