Execute a large SQL script (with GO commands)

Viewed 125717

I need to execute a large set of SQL statements (creating a bunch of tables, views and stored procedures) from within a C# program.

These statements need to be separated by GO statements, but SqlCommand.ExecuteNonQuery() does not like GO statements. My solution, which I suppose I'll post for reference, was to split the SQL string on GO lines, and execute each batch separately.

Is there an easier/better way?

20 Answers

Use SQL Server Management Objects (SMO) which understands GO separators. See my blog post here: http://weblogs.asp.net/jongalloway/Handling-_2200_GO_2200_-Separators-in-SQL-Scripts-2D00-the-easy-way

Sample code:

public static void Main()    
{        
  string scriptDirectory = "c:\\temp\\sqltest\\";
  string sqlConnectionString = "Integrated Security=SSPI;" +
  "Persist Security Info=True;Initial Catalog=Northwind;Data Source=(local)";
  DirectoryInfo di = new DirectoryInfo(scriptDirectory);
  FileInfo[] rgFiles = di.GetFiles("*.sql");
  foreach (FileInfo fi in rgFiles)
  {
        FileInfo fileInfo = new FileInfo(fi.FullName);
        string script = fileInfo.OpenText().ReadToEnd();
        using (SqlConnection connection = new SqlConnection(sqlConnectionString))
        {
            Server server = new Server(new ServerConnection(connection));
            server.ConnectionContext.ExecuteNonQuery(script);
        }
   }
}

If that won't work for you, see Phil Haack's library which handles that: http://haacked.com/archive/2007/11/04/a-library-for-executing-sql-scripts-with-go-separators-and.aspx

This is what I knocked together to solve my immediate problem.

private void ExecuteBatchNonQuery(string sql, SqlConnection conn) {
    string sqlBatch = string.Empty;
    SqlCommand cmd = new SqlCommand(string.Empty, conn);
    conn.Open();
    sql += "\nGO";   // make sure last batch is executed.
    try {
        foreach (string line in sql.Split(new string[2] { "\n", "\r" }, StringSplitOptions.RemoveEmptyEntries)) {
            if (line.ToUpperInvariant().Trim() == "GO") {
                cmd.CommandText = sqlBatch;
                cmd.ExecuteNonQuery();
                sqlBatch = string.Empty;
            } else {
                sqlBatch += line + "\n";
            }
        }            
    } finally {
        conn.Close();
    }
}

It requires GO commands to be on their own line, and will not detect block-comments, so this sort of thing will get split, and cause an error:

ExecuteBatchNonQuery(@"
    /*
    GO
    */", conn);

You can use SQL Management Objects to perform this. These are the same objects that Management Studio uses to execute queries. I believe Server.ConnectionContext.ExecuteNonQuery() will perform what you need.

I look at this a few times at the end decided with EF implementation A bit modified for SqlConnection

public static void ExecuteSqlScript(this SqlConnection sqlConnection, string sqlBatch)
        {
            // Handle backslash utility statement (see http://technet.microsoft.com/en-us/library/dd207007.aspx)
            sqlBatch = Regex.Replace(sqlBatch, @"\\(\r\n|\r|\n)", string.Empty);

            // Handle batch splitting utility statement (see http://technet.microsoft.com/en-us/library/ms188037.aspx)
            var batches = Regex.Split(
                sqlBatch,
                string.Format(CultureInfo.InvariantCulture, @"^\s*({0}[ \t]+[0-9]+|{0})(?:\s+|$)", BatchTerminator),
                RegexOptions.IgnoreCase | RegexOptions.Multiline);

            for (int i = 0; i < batches.Length; ++i)
            {
                // Skip batches that merely contain the batch terminator
                if (batches[i].StartsWith(BatchTerminator, StringComparison.OrdinalIgnoreCase) ||
                    (i == batches.Length - 1 && string.IsNullOrWhiteSpace(batches[i])))
                {
                    continue;
                }

                // Include batch terminator if the next element is a batch terminator
                if (batches.Length > i + 1 &&
                    batches[i + 1].StartsWith(BatchTerminator, StringComparison.OrdinalIgnoreCase))
                {
                    int repeatCount = 1;

                    // Handle count parameter on the batch splitting utility statement
                    if (!string.Equals(batches[i + 1], BatchTerminator, StringComparison.OrdinalIgnoreCase))
                    {
                        repeatCount = int.Parse(Regex.Match(batches[i + 1], @"([0-9]+)").Value, CultureInfo.InvariantCulture);
                    }

                    for (int j = 0; j < repeatCount; ++j)
                    {
                       var command = sqlConnection.CreateCommand();
                       command.CommandText = batches[i];
                       command.ExecuteNonQuery();
                    }
                }
                else
                {
                    var command = sqlConnection.CreateCommand();
                    command.CommandText = batches[i];
                    command.ExecuteNonQuery();
                }
            }
        }

The "GO" batch separator keyword is actually used by SQL Management Studio itself, so that it knows where to terminate the batches it is sending to the server, and it is not passed to SQL server. You can even change the keyword in Management Studio, should you so desire.

If you don't want to install SMO objects you can use gplex tool (see this answer)

I also faced the same problem, and I could not find any other way but splitting the single SQL operation in separate files, then executing all of them in sequence.

Obviously the problem is not with lists of DML commands, they can be executed without GO in between; different story with DDL (create, alter, drop...)

If you don't want to go the SMO route you can search and replace "GO" for ";" and the query as you would. Note that soly the the last result set will be returned.

To avoid third parties, regexes, memory overheads and fast work with large scripts I created my own stream-based parser. It

  • checks syntax before
  • can recognize comments with -- or /**/

    -- some commented text
     /*
    drop table Users;
    GO
       */
    
  • can recognize string literals with ' or "

    set @s =
        'create table foo(...);
        GO
        create index ...';
    
  • preserves LF and CR formatting
  • preserves comments block in object bodies (stored procedures, views etc.)
  • and other constructions such as

          gO -- commented text
    

How to use

    try
    {
        using (SqlConnection connection = new SqlConnection("Integrated Security=SSPI;Persist Security Info=True;Initial Catalog=DATABASE-NAME;Data Source=SERVER-NAME"))
        {
            connection.Open();

            int rowsAffected = SqlStatementReader.ExecuteSqlFile(
                "C:\\target-sql-script.sql",
                connection,
                // Don't forget to use the correct file encoding!!!
                Encoding.Default,
                // Indefinitely (sec)
                0
            );
        }
    }
    // implement your handlers
    catch (SqlStatementReader.SqlBadSyntaxException) { }
    catch (SqlException) { }
    catch (Exception) { }

Stream-based SQL script reader

class SqlStatementReader
{
    public class SqlBadSyntaxException : Exception
    {
        public SqlBadSyntaxException(string description) : base(description) { }
        public SqlBadSyntaxException(string description, int line) : base(OnBase(description, line, null)) { }
        public SqlBadSyntaxException(string description, int line, string filePath) : base(OnBase(description, line, filePath)) { }
        private static string OnBase(string description, int line, string filePath)
        {
            if (filePath == null)
                return string.Format("Line: {0}. {1}", line, description);
            else
                return string.Format("File: {0}\r\nLine: {1}. {2}", filePath, line, description);
        }
    }

    enum SqlScriptChunkTypes
    {
        InstructionOrUnquotedIdentifier = 0,
        BracketIdentifier = 1,
        QuotIdentifierOrLiteral = 2,
        DblQuotIdentifierOrLiteral = 3,
        CommentLine = 4,
        CommentMultiline = 5,
    }

    StreamReader _sr = null;
    string _filePath = null;
    int _lineStart = 1;
    int _lineEnd = 1;
    bool _isNextChar = false;
    char _nextChar = '\0';

    public SqlStatementReader(StreamReader sr)
    {
        if (sr == null)
            throw new ArgumentNullException("StreamReader can't be null.");

        if (sr.BaseStream is FileStream)
            _filePath = ((FileStream)sr.BaseStream).Name;

        _sr = sr;
    }

    public SqlStatementReader(StreamReader sr, string filePath)
    {
        if (sr == null)
            throw new ArgumentNullException("StreamReader can't be null.");

        _sr = sr;
        _filePath = filePath;
    }

    public int LineStart { get { return _lineStart; } }
    public int LineEnd { get { return _lineEnd == 1 ? _lineEnd : _lineEnd - 1; } }

    public void LightSyntaxCheck()
    {
        while (ReadStatementInternal(true) != null) ;
    }

    public string ReadStatement()
    {
        for (string s = ReadStatementInternal(false); s != null; s = ReadStatementInternal(false))
        {
            // skip empty
            for (int i = 0; i < s.Length; i++)
            {
                switch (s[i])
                {
                    case ' ': continue;
                    case '\t': continue;
                    case '\r': continue;
                    case '\n': continue;
                    default:
                        return s;
                }
            }
        }
        return null;
    }

    string ReadStatementInternal(bool syntaxCheck)
    {
        if (_isNextChar == false && _sr.EndOfStream)
            return null;

        StringBuilder allLines = new StringBuilder();
        StringBuilder line = new StringBuilder();
        SqlScriptChunkTypes nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
        SqlScriptChunkTypes currentChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
        char ch = '\0';
        int lineCounter = 0;
        int nextLine = 0;
        int currentLine = 0;
        bool nextCharHandled = false;
        bool foundGO;
        int go = 1;

        while (ReadChar(out ch))
        {
            if (nextCharHandled == false)
            {
                currentChunk = nextChunk;
                currentLine = nextLine;

                switch (currentChunk)
                {
                    case SqlScriptChunkTypes.InstructionOrUnquotedIdentifier:

                        if (ch == '[')
                        {
                            currentChunk = nextChunk = SqlScriptChunkTypes.BracketIdentifier;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == '"')
                        {
                            currentChunk = nextChunk = SqlScriptChunkTypes.DblQuotIdentifierOrLiteral;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == '\'')
                        {
                            currentChunk = nextChunk = SqlScriptChunkTypes.QuotIdentifierOrLiteral;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == '-' && (_isNextChar && _nextChar == '-'))
                        {
                            nextCharHandled = true;
                            currentChunk = nextChunk = SqlScriptChunkTypes.CommentLine;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == '/' && (_isNextChar && _nextChar == '*'))
                        {
                            nextCharHandled = true;
                            currentChunk = nextChunk = SqlScriptChunkTypes.CommentMultiline;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == ']')
                        {
                            throw new SqlBadSyntaxException("Incorrect syntax near ']'.", _lineEnd + lineCounter, _filePath);
                        }
                        else if (ch == '*' && (_isNextChar && _nextChar == '/'))
                        {
                            throw new SqlBadSyntaxException("Incorrect syntax near '*'.", _lineEnd + lineCounter, _filePath);
                        }
                        break;

                    case SqlScriptChunkTypes.CommentLine:

                        if (ch == '\r' && (_isNextChar && _nextChar == '\n'))
                        {
                            nextCharHandled = true;
                            currentChunk = nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                            currentLine = nextLine = lineCounter;
                        }
                        else if (ch == '\n' || ch == '\r')
                        {
                            currentChunk = nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                            currentLine = nextLine = lineCounter;
                        }
                        break;

                    case SqlScriptChunkTypes.CommentMultiline:

                        if (ch == '*' && (_isNextChar && _nextChar == '/'))
                        {
                            nextCharHandled = true;
                            nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                            nextLine = lineCounter;
                        }
                        else if (ch == '/' && (_isNextChar && _nextChar == '*'))
                        {
                            throw new SqlBadSyntaxException("Missing end comment mark '*/'.", _lineEnd + currentLine, _filePath);
                        }
                        break;

                    case SqlScriptChunkTypes.BracketIdentifier:

                        if (ch == ']')
                        {
                            nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                            nextLine = lineCounter;
                        }
                        break;

                    case SqlScriptChunkTypes.DblQuotIdentifierOrLiteral:

                        if (ch == '"')
                        {
                            if (_isNextChar && _nextChar == '"')
                            {
                                nextCharHandled = true;
                            }
                            else
                            {
                                nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                                nextLine = lineCounter;
                            }
                        }
                        break;

                    case SqlScriptChunkTypes.QuotIdentifierOrLiteral:

                        if (ch == '\'')
                        {
                            if (_isNextChar && _nextChar == '\'')
                            {
                                nextCharHandled = true;
                            }
                            else
                            {
                                nextChunk = SqlScriptChunkTypes.InstructionOrUnquotedIdentifier;
                                nextLine = lineCounter;
                            }
                        }
                        break;
                }
            }
            else
                nextCharHandled = false;

            foundGO = false;
            if (currentChunk == SqlScriptChunkTypes.InstructionOrUnquotedIdentifier || go >= 5 || (go == 4 && currentChunk == SqlScriptChunkTypes.CommentLine))
            {
                // go = 0 - break, 1 - begin of the string, 2 - spaces after begin of the string, 3 - G or g, 4 - O or o, 5 - spaces after GO, 6 - line comment after valid GO
                switch (go)
                {
                    case 0:
                        if (ch == '\r' || ch == '\n')
                            go = 1;
                        break;
                    case 1:
                        if (ch == ' ' || ch == '\t')
                            go = 2;
                        else if (ch == 'G' || ch == 'g')
                            go = 3;
                        else if (ch != '\n' && ch != '\r')
                            go = 0;
                        break;
                    case 2:
                        if (ch == 'G' || ch == 'g')
                            go = 3;
                        else if (ch == '\n' || ch == '\r')
                            go = 1;
                        else if (ch != ' ' && ch != '\t')
                            go = 0;
                        break;
                    case 3:
                        if (ch == 'O' || ch == 'o')
                            go = 4;
                        else if (ch == '\n' || ch == '\r')
                            go = 1;
                        else
                            go = 0;
                        break;
                    case 4:
                        if (ch == '\r' && (_isNextChar && _nextChar == '\n'))
                            go = 5;
                        else if (ch == '\n' || ch == '\r')
                            foundGO = true;
                        else if (ch == ' ' || ch == '\t')
                            go = 5;
                        else if (ch == '-' && (_isNextChar && _nextChar == '-'))
                            go = 6;
                        else
                            go = 0;
                        break;
                    case 5:
                        if (ch == '\r' && (_isNextChar && _nextChar == '\n'))
                            go = 5;
                        else if (ch == '\n' || ch == '\r')
                            foundGO = true;
                        else if (ch == '-' && (_isNextChar && _nextChar == '-'))
                            go = 6;
                        else if (ch != ' ' && ch != '\t')
                            throw new SqlBadSyntaxException("Incorrect syntax was encountered while parsing go.", _lineEnd + lineCounter, _filePath);
                        break;
                    case 6:
                        if (ch == '\r' && (_isNextChar && _nextChar == '\n'))
                            go = 6;
                        else if (ch == '\n' || ch == '\r')
                            foundGO = true;
                        break;
                    default:
                        go = 0;
                        break;
                }
            }
            else
                go = 0;

            if (foundGO)
            {
                if (ch == '\r' || ch == '\n')
                {
                    ++lineCounter;
                }
                // clear GO
                string s = line.Append(ch).ToString();
                for (int i = 0; i < s.Length; i++)
                {
                    switch (s[i])
                    {
                        case ' ': continue;
                        case '\t': continue;
                        case '\r': continue;
                        case '\n': continue;
                        default:
                            _lineStart = _lineEnd;
                            _lineEnd += lineCounter;
                            return allLines.Append(s.Substring(0, i)).ToString();
                    }
                }
                return string.Empty;
            }

            // accumulate by string
            if (ch == '\r' && (_isNextChar == false || _nextChar != '\n'))
            {
                ++lineCounter;
                if (syntaxCheck == false)
                    allLines.Append(line.Append('\r').ToString());
                line.Clear();
            }
            else if (ch == '\n')
            {
                ++lineCounter;
                if (syntaxCheck == false)
                    allLines.Append(line.Append('\n').ToString());
                line.Clear();
            }
            else
            {
                if (syntaxCheck == false)
                    line.Append(ch);
            }
        }

        // this is the end of the stream, return it without GO, if GO exists
        switch (currentChunk)
        {
            case SqlScriptChunkTypes.InstructionOrUnquotedIdentifier:
            case SqlScriptChunkTypes.CommentLine:
                break;
            case SqlScriptChunkTypes.CommentMultiline:
                if (nextChunk != SqlScriptChunkTypes.InstructionOrUnquotedIdentifier)
                    throw new SqlBadSyntaxException("Missing end comment mark '*/'.", _lineEnd + currentLine, _filePath);
                break;
            case SqlScriptChunkTypes.BracketIdentifier:
                if (nextChunk != SqlScriptChunkTypes.InstructionOrUnquotedIdentifier)
                    throw new SqlBadSyntaxException("Unclosed quotation mark [.", _lineEnd + currentLine, _filePath);
                break;
            case SqlScriptChunkTypes.DblQuotIdentifierOrLiteral:
                if (nextChunk != SqlScriptChunkTypes.InstructionOrUnquotedIdentifier)
                    throw new SqlBadSyntaxException("Unclosed quotation mark \".", _lineEnd + currentLine, _filePath);
                break;
            case SqlScriptChunkTypes.QuotIdentifierOrLiteral:
                if (nextChunk != SqlScriptChunkTypes.InstructionOrUnquotedIdentifier)
                    throw new SqlBadSyntaxException("Unclosed quotation mark '.", _lineEnd + currentLine, _filePath);
                break;
        }

        if (go >= 4)
        {
            string s = line.ToString();
            for (int i = 0; i < s.Length; i++)
            {
                switch (s[i])
                {
                    case ' ': continue;
                    case '\t': continue;
                    case '\r': continue;
                    case '\n': continue;
                    default:
                        _lineStart = _lineEnd;
                        _lineEnd += lineCounter + 1;
                        return allLines.Append(s.Substring(0, i)).ToString();
                }
            }
        }

        _lineStart = _lineEnd;
        _lineEnd += lineCounter + 1;
        return allLines.Append(line.ToString()).ToString();
    }

    bool ReadChar(out char ch)
    {
        if (_isNextChar)
        {
            ch = _nextChar;
            if (_sr.EndOfStream)
                _isNextChar = false;
            else
                _nextChar = Convert.ToChar(_sr.Read());
            return true;
        }
        else if (_sr.EndOfStream == false)
        {
            ch = Convert.ToChar(_sr.Read());
            if (_sr.EndOfStream == false)
            {
                _isNextChar = true;
                _nextChar = Convert.ToChar(_sr.Read());
            }
            return true;
        }
        else
        {
            ch = '\0';
            return false;
        }
    }

    public static int ExecuteSqlFile(string filePath, SqlConnection connection, Encoding fileEncoding, int commandTimeout)
    {
        int rowsAffected = 0;
        using (FileStream fs = new FileStream(filePath, FileMode.Open, FileAccess.Read, FileShare.Read))
        {
            // Simple syntax check (you can comment out these two lines below)
            new SqlStatementReader(new StreamReader(fs, fileEncoding)).LightSyntaxCheck();
            fs.Seek(0L, SeekOrigin.Begin);

            // Read statements without GO
            SqlStatementReader rd = new SqlStatementReader(new StreamReader(fs, fileEncoding));
            string stmt;
            while ((stmt = rd.ReadStatement()) != null)
            {
                using (SqlCommand cmd = connection.CreateCommand())
                {
                    cmd.CommandText = stmt;
                    cmd.CommandTimeout = commandTimeout;
                    int i = cmd.ExecuteNonQuery();
                    if (i > 0)
                        rowsAffected += i;
                }
            }
        }
        return rowsAffected;
    }
}

You can just use ; at the end of each statement as it worked for me. Really don't know if there are any drawbacks to it.

Here is the most elegant solution I could find.

Let's say your SQL script containing GO statements is in a file script.sql.

You can do:

$query = ((Get-Content -Raw "script.sql") -replace '([\s\n]*)GO([\s\n]+)','$1$2')

This will remove all GO statements without removing other occurences of the string go anywhere else (according to my tests).

You can then do:

$SqlConnection = New-Object System.Data.SqlClient.SqlConnection                
$SqlCmd = New-Object System.Data.SqlClient.SqlCommand

$SqlConnection.ConnectionString = <your_connection_string>

$SqlConnection.Open()
   
$SqlCmd.Connection = $SqlConnection 
$SqlCmd.CommandText = $query
$SqlCmd.ExecuteNonQuery()

which should normally work without issues.

Related