-1

I referenced a few other blogs and I can't isolate the issue to this logic. Code is below. Throws the error message attached.

            -- 1.   Import Multiple Delimited Text Files into a SQL Database

            -- 1.1  Define the path to the input and define the terminators


            USE [Openair_Integration]
            DECLARE @Path NVARCHAR(255) = 'C:\\Users\nicolas.gutierrez.su\Downloads\\'
            DECLARE @RowTerminator NVARCHAR(5) = CHAR(13) + CHAR(10)
            DECLARE @ColumnTerminator NVARCHAR(5) = CHAR(9)

            -- 1.2  Define the list of input and output in a temporary table


            IF OBJECT_ID('[dbo].[Files_Temporary]', 'U') IS NOT NULL
            DROP TABLE [dbo].[Files_Temporary];
            CREATE TABLE [dbo].[Files_Temporary]
            (
            [ID] INT
            , [FileName] NVARCHAR(255)
            , [TableName] NVARCHAR(255)
            );

            INSERT INTO [dbo].[Files_Temporary] SELECT 1,   'booking.csv',  'dbo.booking'

            -- 1.3  Loop over the list of input and output and import each file to the correct table


            DECLARE @Counter INT = 1

            WHILE @Counter <= (SELECT COUNT(*) FROM [dbo].[Files_Temporary])
            BEGIN
            PRINT 'Counter is ''' + CONVERT(NVARCHAR(5), @Counter) + '''.'

            DECLARE @FileName NVARCHAR(255)
            DECLARE @TableName NVARCHAR(255)
            DECLARE @Header NVARCHAR(MAX)
            DECLARE @SQL_Header NVARCHAR(MAX)
            DECLARE @CreateHeader NVARCHAR(MAX) = ''
            DECLARE @SQL_CreateHeader NVARCHAR(MAX)

            SELECT @FileName = [FileName], @TableName = [TableName] FROM [dbo].[Files_Temporary] WHERE [ID] = @Counter

            IF OBJECT_ID('[dbo].[' + @TableName + ']', 'U') IS NULL
            BEGIN
        -----------------------------------------------------------------------------------------------------------------------------------------------------------
                PRINT 'Creating new table with name ''' + @TableName + '''.'

                IF OBJECT_ID('[dbo].[Header_Temporary]', 'U') IS NOT NULL
                DROP TABLE [dbo].[Header_Temporary];
                CREATE TABLE [dbo].[Header_Temporary]
                (
                    [Header] NVARCHAR(MAX)
                );

                SET @SQL_Header = '
                    BULK INSERT [dbo].[Header_Temporary]
                    FROM ''' + @Path + @FileName + '''
                    WITH
                    (
                        FIRSTROW = 1,
                        LASTROW = 1,
                        MAXERRORS = 0,
                        FIELDTERMINATOR = ''' + @RowTerminator + ''',
                        ROWTERMINATOR = ''' + @RowTerminator + '''
                    )'
                EXEC(@SQL_Header)

                SET @Header = (SELECT TOP 1 [Header] FROM [dbo].[Header_Temporary])
                PRINT 'Extracted header ''' + @Header + ''' for table ''' + @TableName + '''.'
        ----------------------------------------------------------------------------------------------------------------------------------------------------------------
                WHILE CHARINDEX(@ColumnTerminator, @Header) > 0
                BEGIN          
                    SET @CreateHeader = @CreateHeader + '[' + LTRIM(RTRIM(SUBSTRING(@Header, 1, CHARINDEX(@ColumnTerminator, @Header) - 1))) + '] NVARCHAR(255), '
                    SET @Header = SUBSTRING(@Header, CHARINDEX(@ColumnTerminator, @Header) + 1, LEN(@Header)) 
                END
                SET @CreateHeader = @CreateHeader + '[' + @Header + '] NVARCHAR(255)'

                SET @SQL_CreateHeader = 'CREATE TABLE [' + @TableName + '] (' + @CreateHeader + ')'
                EXEC(@SQL_CreateHeader)
            END

            -------------------------------------------------------------------------------------------------------------------------------------------------------------
            PRINT 'Inserting data from ''' + @FileName + ''' to ''' + @TableName + '''.'
            DECLARE @SQL NVARCHAR(MAX)
            SET @SQL = '
                BULK INSERT [dbo].[' + @TableName + ']
                FROM ''' + @Path + @FileName + '''
                WITH
                (
                    FIRSTROW = 2,
                    MAXERRORS = 0,
                    FIELDTERMINATOR = ''' + @ColumnTerminator + ''',
                    ROWTERMINATOR = ''' + @RowTerminator + '''
                )'
            EXEC(@SQL)

            SET @Counter = @Counter + 1
            END;

            -- 1.4  Cleanup temporary tables


            IF OBJECT_ID('[dbo].[Files_Temporary]', 'U') IS NOT NULL
            DROP TABLE [dbo].[Files_Temporary];

            IF OBJECT_ID('[dbo].[Header_Temporary]', 'U') IS NOT NULL
            DROP TABLE [dbo].[Header_Temporary];

identified to long

1 Answers1

5

This line

DECLARE @ColumnTerminator NVARCHAR(5) = CHAR(9)

makes the assumption that columns are tab-delimited (9 = ASCII code for a tab character). Later on, the header row is split using this loop:

WHILE CHARINDEX(@ColumnTerminator, @Header) > 0
BEGIN          
    SET @CreateHeader = @CreateHeader + '[' + LTRIM(RTRIM(SUBSTRING(@Header, 1, CHARINDEX(@ColumnTerminator, @Header) - 1))) + '] NVARCHAR(255), '
    SET @Header = SUBSTRING(@Header, CHARINDEX(@ColumnTerminator, @Header) + 1, LEN(@Header)) 
END

Since the actual column delimiter is a comma, this loop is a no-op, so things fall through to the next line:

SET @CreateHeader = @CreateHeader + '[' + @Header + '] NVARCHAR(255)'

Here you are effectively attempting to define a single column named after the entire header row! Since the total number of characters in the header row is over the maximum length of a T-SQL identifier (128), the resulting CREATE TABLE statement raises an error.

To resolve the error for the file you have, the @ColumnTerminator assignment at the top just needs to be fixed to specify your actual column delimiter, i.e. a comma:

DECLARE @ColumnTerminator NVARCHAR(5) = N','

However, there are some other minor issues with your code, together with a small elephant in the room:

  • You declare using NVARCHAR not VARCHAR throughout, but use non-Unicode literals which makes me doubt you are really looking to support Unicode (Unicode literals in T-SQL are where you add the N prefix).

  • Nitpicky, but the traditional Windows max path length is 260 not 255; the inconsistent double slashing is also a bit odd.

  • You create permanent tables named XXXX_Temporary, which you then drop at the end. Why not create actual temporary tables in that case? (CREATE TABLE #TableName)

  • If you are on a more recent version of SQL Server, you can modernise the IF OBJECT_ID('TableName') IS NOT NULL DROP TABLE TableName statements to DROP TABLE IF EXISTS TableName

  • Similarly, the STRING_SPLIT function in a modern version provides a better way of simple splitting than your WHILE loop, which makes repeated small memory allocations.

  • At the moment you are blindly hardcoding square brackets. When you are hardcoding the name as well, and the name clearly does not need escaping, that's a bit pointless ('[dbo].[Something]'). Conversely, to properly escape unknown identifiers, you should use the QUOTENAME function instead, which will deal with any embedded square brackets properly.

  • Until SQL Server 2017 explicitly added a proper 'CSV' option, be warned that BULK INSERT does not handle 'real' CSV files, and instead parses data very simplistically. So, if you set FIELDTERMINATOR to a comma, then the following would not import properly:

    "Flat 1, Bishopstoke Views","High Street",Newtown

  • If you can't rely on import files having set formatting settings (or put another way: if you need to be flexible over what your BULK INSERT settings are), then sticking to a pure T-SQL approach isn't the best one - you'd be better off doing the file and string parsing in another language more suited to the task. It's not that procedural T-SQL doesn't 'work', it's just that it's no where near state of the art for imperative scripting - you'd be better off writing something in Python or whatever, and only use T-SQL on the edges. In fact, I'd even prefer using VBA over pure T-SQL for this sort of job.

Chris Rolliston
  • 345
  • 1
  • 2
  • 5