Question

我将收到一堆目录和文件的完整路径，需要建立一个目录结构。此结构将使用定义为：

的表中的邻接列表存储在SQL Server中

CREATE TABLE [dbo].[DirTreeEntry] 
(
    [id] [int] IDENTITY(1,1) NOT NULL,
    [full_path] [nvarchar](2048) NOT NULL,
    [name] [nvarchar](255) NOT NULL,
    [is_file] [bit] NOT NULL,
    [is_root] [bit] NOT NULL,
    [parent_id] [int] NULL,
    [source_id] [int] NULL
)

以id为主键，name在这种情况下是直接名称，而不是完整路径，source_id引用源表条目（如果它明确存在）。 / p>

源数据将在此表中：

CREATE TABLE [dbo].[dir_path] 
(
     [dir_path_id] [int] IDENTITY(1,1) NOT NULL,
     [directory_path] [nvarchar](2048) NOT NULL,
     [name] [nvarchar](255) NOT NULL,
     [file_flag] [bit] NOT NULL,
     [filesize] [bigint] NULL,
     [create_date] [datetime] NOT NULL
)

完整路径实际上是directory_path和name的组合。

鉴于[dbo].[dir_path]中的以下条目：

1, '/root/subdir1/subdir2', 'subdir3', 0, NULL, '9/9/2014'
2, '/root/subdir1/subdir2/subdir3', 'somefile.txt', 1, 25, '9/9/2014'
3, '/etc/rc.d', 'rc.local', 1, 10, '9/9/2014'

我需要在[dbo].[DirTreeEntry]：

中结束这一点

1,'/','/',0,1, NULL,NULL
2,'/root','root',0,0,1,NULL
3,'/root/subdir1','subdir1',0,0,2,NULL
4,'/root/subdir1/subdir2','subdir2',0,0,3,NULL
5,'/root/subdir1/subdir2/subdir3','subdir3',0,0,4,1
6,'/root/subdir1/subdir2/subdir3/somefile.txt','somefile.txt',1,0,5,2
7,'/etc','etc',0,0,1,NULL
8,'/etc/rc.d','rc.d',0,0,7,NULL
9,'/etc/rc.d/rc.local','rc.local',1,0,8,3

{C}中的以下代码完全正是我在寻找C＃中的层次结构：

public class Node
{
   private readonly IDictionary<string, Node> _nodes = 
      new Dictionary<string, Node>();

   public string Path { get; set; }
}

public void AddPath(string path)
{
   char[] charSeparators = new char[] {'\\'};

   // Parse into a sequence of parts.
   string[] parts = path.Split(charSeparators, 
       StringSplitOptions.RemoveEmptyEntries);

   // The current node.  Start with this.
   Node current = this;

   // Iterate through the parts.
   foreach (string part in parts)
   {
       // The child node.
       Node child;

       // Does the part exist in the current node?  If
       // not, then add.
       if (!current._nodes.TryGetValue(part, out child))
       {
           // Add the child.
           child = new Node {
               Path = part
           };

           // Add to the dictionary.
           current._nodes[part] = child;
       }

       // Set the current to the child.
       current = child;
   }
}

但是，我可能会获得包含100,000多个条目的源数据，而且我不想在C＃端的内存中构建该结构，然后必须将所有这些数据发送到SQL。我已经有了一种快速的方法将源数据导入数据库，我现在需要一个存储过程来根据源数据构建[dbo].[DirTreeEntry]表。

任何指导都将不胜感激！

Answer 1

这应该可以解决问题。它可以运行一次以上，因为它不会添加任何已经存在的内容，但它不会处理并发更新。

DECLARE @ParentID as int
DECLARE @Path as nvarchar(2048)
DECLARE @Name as nvarchar(255)
DECLARE @LastPart as nvarchar(255)
DECLARE @PartialPath as nvarchar(2048)

-- Loop through the input table
DECLARE paths CURSOR FOR SELECT p.directory_path, p.name FROM dir_path p;
OPEN paths;

FETCH NEXT FROM paths INTO @Path, @LastPart;
WHILE @@FETCH_STATUS = 0
BEGIN
    -- Reset loop variables
    SET @ParentID = NULL
    SET @PartialPath = '';

    -- Split the full path into parts and loop through those
    DECLARE parts CURSOR FOR SELECT Value FROM dbo.splitstring(@Path+'/'+@LastPart, '/');
    OPEN parts;

    FETCH NEXT FROM parts INTO @Name;
    WHILE @@FETCH_STATUS = 0
    BEGIN
        -- Build the path for this iteration
        SET @PartialPath = (CASE WHEN @PartialPath = '/' THEN '/'+@Name ELSE @PartialPath + '/' + @Name END);

        IF @Name = '' BEGIN 
        SET 
            @Name = '/' 
        END

        -- Insert the new path when it doesn't exist yet
        INSERT INTO DirTreeEntry (name, full_path, parent_id, is_file, is_root)
        SELECT @Name, @PartialPath, @ParentID, 0, 0
        WHERE NOT EXISTS (SELECT ID FROM DirTreeEntry WHERE full_path = @PartialPath)

        -- Store the id for the next level
        SELECT @ParentID = ID FROM DirTreeEntry WHERE full_path = @PartialPath;

        FETCH NEXT FROM parts INTO @Name;
    END
    CLOSE parts;
    DEALLOCATE parts;

    FETCH NEXT FROM paths INTO @Path,@LastPart;
END

CLOSE paths;
DEALLOCATE paths;

-- Update missing values in target table
UPDATE DirTreeEntry SET source_id = S.dir_path_id, is_file = S.file_flag
FROM DirTreeEntry T 
INNER JOIN dir_path S ON S.directory_path+'/'+S.name = T.full_path

UPDATE DirTreeEntry SET is_root = 1 WHERE full_path = '/'

编辑：它可以更快地完成。使用此SplitString function的改编：

CREATE FUNCTION [dbo].[SplitPath]
    (
        @List NVARCHAR(MAX)
    )
    RETURNS TABLE
    AS
        RETURN ( SELECT [Path], [Name] FROM 
          ( 
            SELECT 
              [Path] = LTRIM(RTRIM(SUBSTRING(@List, 0,
              (CASE WHEN CHARINDEX('/', @List + '/', [Number]) > 1 THEN CHARINDEX('/', @List + '/', [Number]) ELSE 2 END)))),
              [Name] = LTRIM(RTRIM(SUBSTRING(@List, [Number],
              (CASE WHEN CHARINDEX('/', @List + '/', [Number]) - [Number] > 0 THEN CHARINDEX('/', @List + '/', [Number])  - [Number] ELSE 1 END))))
            FROM (SELECT Number = ROW_NUMBER() OVER (ORDER BY name)
              FROM sys.all_objects) AS x
              WHERE Number <= LEN(@List)
              AND SUBSTRING('/' + @List, [Number], LEN('/')) = '/'
          ) AS y
        );

这会吐出部分路径和名称，并修复'/'特殊情况。此输出可以与dir_path表交叉应用，在DirTreeEntry中创建基本条目。然后可以使用缺少的信息来丰富DirTreeEntry。

INSERT DirTreeEntry (full_path, name, is_file, is_root)
SELECT DISTINCT d.Path, d.Name, 0, 0 FROM dir_path p 
CROSS APPLY dbo.SplitPath(directory_path+'/'+name) as d
ORDER BY d.Path

UPDATE DirTreeEntry SET source_id = S.dir_path_id, is_file = S.file_flag
FROM DirTreeEntry T 
INNER JOIN dir_path S ON S.directory_path+'/'+S.name = T.full_path

UPDATE DirTreeEntry SET is_root = 1 WHERE full_path = '/'

UPDATE A SET parent_id = B.id
FROM DirTreeEntry A INNER JOIN DirTreeEntry B ON CASE WHEN B.full_path = '/' THEN B.full_path + A.name ELSE B.full_path + '/' + A.name END = A.full_path

可能还有进一步优化的空间，但这应该明显加快。

将C＃代码转换为SQL存储过程

1 个答案: