Article ID: 194975
Article Last Modified on 3/14/2005
APPLIES TO
- Microsoft ActiveX Data Objects 1.5
- Microsoft ActiveX Data Objects 2.0
- Microsoft ActiveX Data Objects 2.1 Service Pack 2
- Microsoft ActiveX Data Objects 2.5
- Microsoft ActiveX Data Objects 2.6
- Microsoft ActiveX Data Objects 2.7
- Microsoft Visual Basic 6.0 Professional Edition
- Microsoft Visual Basic 6.0 Enterprise Edition
This article was previously published under Q194975
For a Microsoft Visual Basic .NET version of this article, see 317034.
For a Microsoft Visual Basic .NET version of this article, see 316887.
SUMMARY
This article describes how to read and write Binary Large Objects (BLOBs) using GetChunk and AppendChunk methods against fields in ADO. It also includes sample code using the NWIND sample database.
MORE INFORMATION
The GetChunk and AppendChunk methods work with the LongVarChar, LongVarWChar, and LongVarBinary column types, also known as TEXT, NTEXT, and IMAGE columns, in Microsoft SQL Server, and as MEMO and OLE fields in Microsoft Jet databases. You can identify these columns in ADO by testing the Type property of a Field for the values adLongVarChar, adLongVarWChar, and adLongVarBinary. You can also test the Attributes property of a Field for the adFldLong flag:
If fld.Attributes And adFldLong Then ' You can use GetChunk/AppendChunk
Long columns are commonly referred to as BLOBs (Binary Large OBjects) even though they may contain text data. The sample code below provides two routines, BlobToFile and FileToBlob.
BlobToFile
BlobToFile determines the data type of the field and which of three methods to use to write the BLOB data to a disk file. If the BLOB data is small enough, it will reference the field value in its entirety without calling GetChunk. If the BLOB size is unknown, it will call WriteFromUnsizedBinary or WriteFromUnsizedText to write the data. This is less efficient in terms of making extra copies of the data in local memory than the WriteFromBinary and WriteFromText routines that are used when the size of the BLOB data is known:
BlobToFile Calls one of the below routines to use GetChunk WriteFromBinary Writes a LongVarBinary of known size to disk WriteFromUnsizedBinary Writes a LongVarBinary on unknown size WriteFromText Writes a LongVarChar of known size WriteFromUnsizedText Writes a LongVarChar of unknown size
FileToBlob
FileToBlob determines whether to use AppendChunk or directly assign the data to the BLOB field based on the size of the file. Because the size of the file can always be determined, there are no "Unsized" routines as there are in the BlobToFile sample code:
FileToBlob Calls one of the below routines to use AppendChunk ReadToBinary Reads a file into a LongVarBinary column ReadToText Reads a file into a LongVarChar column
Example
The sample code for BlobToFile and FileToBlob is stored in a Module, while the test code is behind the default form. The test code uses each of the three methods to save to disk the Photo (IMAGE/OLE/LongVarBinary) and Notes (TEXT/MEMO/LongVarChar/LongVarWChar) fields for Andrew Fuller from the Employees table of the NWIND database. It then reads the files back in and creates six new records, reading each of the three sets of files via the two different read methods. NOTE: Using ADO 2.1 and later, you might see the following error on the line of code:
Data = fld.GetChunk(BLOCK_SIZE) : Run-time error '94': Invalid use of Null
ADO 2.1 and later might report the ActualSize property of a Text type BLOB field as twice the number of characters. This is correct if the BLOB field contains Unicode text, because Unicode uses 2 bytes per character. This is incorrect if the BLOB field contains ANSI text, which uses 1-byte per character. If the ActualSize is twice the length of an ANSI field, then GetChunk eventually attempts to get past the end of the field.
You can see this behavior in the following scenarios, and might see it in other scenarios:
Scenario 1:
With a SQL Server 7 NTEXT field (ANSI text), using either the OLE DB Provider for SQL Server or the ODBC Provider with the SQL Server ODBC driver.
Note that using SQL Server TEXT field (Unicode) works with both providers.
Scenario 2:
With Access 97 MEMO fields and with Access 2000 MEMO fields with or without Unicode compression, using either the OLE DB Provider for Jet 4.0 or the ODBC Provider with the Jet ODBC 4.0 driver, ODBCJT32.DLL.
Note that, with Access 97 MEMO fields, both the OLE DB Provider for Jet 3.51 and the ODBC Provider with the Jet ODBC 3.51 driver work correctly.
There are several possible workarounds for Run-time error '94':
- Use rs.Fieldname.ActualSize \ 2 instead of rs.Fieldname.ActualSize. This resolves each of the specific scenarios listed above.
- Use method 2 or 3 below, neither of which rely on the ActualSize property.
Preparing the Data
- In Microsoft Access or other tool, open NWIND.MDB.
- Open the Employees table (or form) and locate "Andrew Fuller."
- Paste the contents of a large text file (between 30000 and 60000 bytes) into the Notes field.
- Save the changes and exit Access.
- Add an ODBC datasource that points to the NWIND.MDB file.
Sample Code
- Create a new Visual Basic project and from the Project menu, select References, and select Microsoft ActiveX Data Objects Library or Microsoft ActiveX Data Objects Library.
- Add two CommandButtons (cmdSave and cmdLoad) to the default form(Form1).
Add the following code. You will have to change the connect string supplied on the "cn.Open" line:
Option Explicit Private Sub CmdSave_Click() Dim cn As ADODB.Connection, rs As ADODB.Recordset, SQL As String Set cn = New ADODB.Connection Set rs = New ADODB.Recordset cn.CursorLocation = adUseServer cn.Open "dsn=nwind_jet" ' *** change this *** SQL = "SELECT * FROM Employees WHERE LastName='Fuller'" rs.Open SQL, cn, adOpenStatic, adLockReadOnly ' ' Save using GetChunk and known size. ' FieldSize (ActualSize) > Threshold arg (16384) ' BlobToFile rs!Photo, "c:\photo1.dat", rs!Photo.ActualSize, 16384 BlobToFile rs!Notes, "c:\notes1.txt", rs!Notes.ActualSize, 16384 ' Uncomment the next line of code, and comment the line above, ' to workaround Runtime error '94': Invalid use of Null ' BlobToFile rs!Notes, "c:\notes1.txt", rs!Notes.ActualSize \ 2, 16384 ' ' Save using GetChunk and unknown size. ' FieldSize not specified. ' BlobToFile rs!Photo, "c:\photo2.dat" BlobToFile rs!Notes, "c:\notes2.txt" ' ' Save without using GetChunk ' FieldSize (ActualSize) < Threshold arg (defaults to 1Mb) ' BlobToFile rs!Photo, "c:\photo3.dat", rs!Photo.ActualSize BlobToFile rs!Notes, "c:\notes3.txt", rs!Notes.ActualSize ' Uncomment the next line of code, and comment the line above, ' to workaround Runtime error '94': Invalid use of Null ' BlobToFile rs!Notes, "c:\notes3.txt", rs!Notes.ActualSize \ 2 rs.Close cn.Close End Sub Private Sub CmdLoad_Click() Dim cn As ADODB.Connection, rs As ADODB.Recordset, SQL As String Set cn = New ADODB.Connection Set rs = New ADODB.Recordset cn.CursorLocation = adUseServer cn.Open "dsn=ole_db_nwind_jet" SQL = "SELECT * FROM Employees" rs.Open SQL, cn, adOpenKeyset, adLockOptimistic ' ' Load using AppendChunk ' rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller11" FileToBlob "c:\photo1.dat", rs!Photo, 16384 FileToBlob "c:\notes1.txt", rs!Notes, 16384 rs.Update rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller21" FileToBlob "c:\photo2.dat", rs!Photo, 16384 FileToBlob "c:\notes2.txt", rs!Notes, 16384 rs.Update rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller31" FileToBlob "c:\photo3.dat", rs!Photo, 16384 FileToBlob "c:\notes3.txt", rs!Notes, 16384 rs.Update ' ' Load without using AppendChunk ' rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller12" FileToBlob "c:\photo1.dat", rs!Photo FileToBlob "c:\notes1.txt", rs!Notes rs.Update rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller22" FileToBlob "c:\photo2.dat", rs!Photo FileToBlob "c:\notes2.txt", rs!Notes rs.Update rs.AddNew rs!FirstName = "Test" rs!LastName = "Fuller32" FileToBlob "c:\photo3.dat", rs!Photo FileToBlob "c:\notes3.txt", rs!Notes rs.Update rs.Close cn.Close End Sub
Add a new module to the project (Module1) with the following code:
Option Explicit Const BLOCK_SIZE = 16384 Sub BlobToFile(fld As ADODB.Field, ByVal FName As String, _ Optional FieldSize As Long = -1, _ Optional Threshold As Long = 1048576) ' ' Assumes file does not exist ' Data cannot exceed approx. 2Gb in size ' Dim F As Long, bData() As Byte, sData As String F = FreeFile Open FName For Binary As #F Select Case fld.Type Case adLongVarBinary If FieldSize = -1 Then ' blob field is of unknown size WriteFromUnsizedBinary F, fld Else ' blob field is of known size If FieldSize > Threshold Then ' very large actual data WriteFromBinary F, fld, FieldSize Else ' smallish actual data bData = fld.Value Put #F, , bData ' PUT tacks on overhead if use fld.Value End If End If Case adLongVarChar, adLongVarWChar If FieldSize = -1 Then WriteFromUnsizedText F, fld Else If FieldSize > Threshold Then WriteFromText F, fld, FieldSize Else sData = fld.Value Put #F, , sData ' PUT tacks on overhead if use fld.Value End If End If End Select Close #F End Sub Sub WriteFromBinary(ByVal F As Long, fld As ADODB.Field, _ ByVal FieldSize As Long) Dim Data() As Byte, BytesRead As Long Do While FieldSize <> BytesRead If FieldSize - BytesRead < BLOCK_SIZE Then Data = fld.GetChunk(FieldSize - BLOCK_SIZE) BytesRead = FieldSize Else Data = fld.GetChunk(BLOCK_SIZE) BytesRead = BytesRead + BLOCK_SIZE End If Put #F, , Data Loop End Sub Sub WriteFromUnsizedBinary(ByVal F As Long, fld As ADODB.Field) Dim Data() As Byte, Temp As Variant Do Temp = fld.GetChunk(BLOCK_SIZE) If IsNull(Temp) Then Exit Do Data = Temp Put #F, , Data Loop While LenB(Temp) = BLOCK_SIZE End Sub Sub WriteFromText(ByVal F As Long, fld As ADODB.Field, _ ByVal FieldSize As Long) Dim Data As String, CharsRead As Long Do While FieldSize <> CharsRead If FieldSize - CharsRead < BLOCK_SIZE Then Data = fld.GetChunk(FieldSize - BLOCK_SIZE) CharsRead = FieldSize Else Data = fld.GetChunk(BLOCK_SIZE) CharsRead = CharsRead + BLOCK_SIZE End If Put #F, , Data Loop End Sub Sub WriteFromUnsizedText(ByVal F As Long, fld As ADODB.Field) Dim Data As String, Temp As Variant Do Temp = fld.GetChunk(BLOCK_SIZE) If IsNull(Temp) Then Exit Do Data = Temp Put #F, , Data Loop While Len(Temp) = BLOCK_SIZE End Sub Sub FileToBlob(ByVal FName As String, fld As ADODB.Field, _ Optional Threshold As Long = 1048576) ' ' Assumes file exists ' Assumes calling routine does the UPDATE ' File cannot exceed approx. 2Gb in size ' Dim F As Long, Data() As Byte, FileSize As Long F = FreeFile Open FName For Binary As #F FileSize = LOF(F) Select Case fld.Type Case adLongVarBinary If FileSize > Threshold Then ReadToBinary F, fld, FileSize Else Data = InputB(FileSize, F) fld.Value = Data End If Case adLongVarChar, adLongVarWChar If FileSize > Threshold Then ReadToText F, fld, FileSize Else fld.Value = Input(FileSize, F) End If End Select Close #F End Sub Sub ReadToBinary(ByVal F As Long, fld As ADODB.Field, _ ByVal FileSize As Long) Dim Data() As Byte, BytesRead As Long Do While FileSize <> BytesRead If FileSize - BytesRead < BLOCK_SIZE Then Data = InputB(FileSize - BytesRead, F) BytesRead = FileSize Else Data = InputB(BLOCK_SIZE, F) BytesRead = BytesRead + BLOCK_SIZE End If fld.AppendChunk Data Loop End Sub Sub ReadToText(ByVal F As Long, fld As ADODB.Field, _ ByVal FileSize As Long) Dim Data As String, CharsRead As Long Do While FileSize <> CharsRead If FileSize - CharsRead < BLOCK_SIZE Then Data = Input(FileSize - CharsRead, F) CharsRead = FileSize Else Data = Input(BLOCK_SIZE, F) CharsRead = CharsRead + BLOCK_SIZE End If fld.AppendChunk Data Loop End Sub
- Run the project and click the cmdSave button.
- In the C:\ directory, you should find the following files: notes1.txt
notes2.txt
notes3.txt
photo1.dat
photo2.dat
photo3.dat
The three "photo" files should be the same size as each other. The three "notes" files should be the same size as each other.
- Click the cmdLoad button.
- Open the database using Access and you should see six additional employees with photos and notes loaded back correctly.
Notes
Following are some suggestions for using BLOBs with ADO. These parallel many of the suggestions in the following Microsoft Knowledge Base article:
153238 How To Use GetChunk and AppendChunk Methods of RDO Object
- It is more efficient in terms of retrieval of BLOB data to simply store the data in files on the server with a pointer in the main record (or you can use some sort of structured directory/file naming system based on the primary key value). This has the advantage of (a) eliminating server overhead, (b) allowing the files to be stored on a second server, (c) allowing network security attributes to be set on individual files, and (d) allowing retrieval of files even when the server is down. This is especially true if the files are some sort of document type, such as bitmaps (.bmp), word processor files (.doc), or spreadsheets (.xls) where you can point the host application directly to the file on the server.
- When using certain providers, most notably ODBC to SQL Server and other databases, you may have to take special care in retrieving BLOB data, such as placing BLOB columns at the end of the field list and referencing all non-BLOB fields prior to access BLOB columns. This will depend on a number of factors, such as:
- Provider (typically ODBC)
- Back-end server
- Cursor Location (typically client)
- Cursor Type
- Whether you're selecting from a VIEW or getting records returned from a stored procedure.
Because this depends on a variety of factors, below is a guide if you are having problems with BLOB columns:
- Try a native OLE DB provider instead of an ODBC provider.
- Use Server-side cursors (such as adOpenKeyset).
- Select the Primary Key column(s) in addition to any other columns.
- Select the BLOB columns last. Select individual fields, not "*".
- Access all non-BLOB columns first (store them if necessary).
- Access BLOB columns in the order specified. You may only be able to reference it once before the cursor loses its value.
- When editing a BLOB column using the AppendChunk method, you may have to edit at least one non-BLOB column in your recordset as well. BLOBs are typically not updateable with Static or Forward-only cursors on ODBC datasources.
- If you use ODBC to Jet, you can't update a recordset returned by a stored procedures (QueryDef) at all because the driver forces them to be read-only.
- The Microsoft Oracle OLE DB provider does not currently support random Access to BLOB data with server-side cursors - the BLOB column must appear to the end of the SELECT clause.
With the ODBC cursor library, it is not possible to use the GetChunk or AppendChunk methods on a recordset returned from a stored procedure. This is because the BLOB data is not normally retrieved with the rest of the data in order to save bandwidth. When a stored procedure creates a recordset, the cursor driver cannot determine how to query for the BLOB data after the fact because it cannot determine the base tables or key fields to use. Server-side cursors alleviate this problem but limit you to a single statement per stored procedure (a SQL Server restriction).
The fact that users want to update their BLOB column demands that they expose their base tables and create the cursor by using a standard select statement from that base table. This would be true even if you were coding directly to ODBC (not an ADO thing).
REFERENCES
For additional information, please see the following articles in the Microsoft Knowledge Base:
185958 How To Use ADO GetChunk/AppendChunk with Oracle for BLOB Data
189415 FILE: AdoChunk.exe Using GetChunk and AppendChunk in Visual C++
Using Data Access Objects:
103257 ACC: Reading, Storing, &Writing Binary Large Objects (BLOBs)
Additional query words: concatenate
Keywords: kbhowto kbbug kbdatabase KB194975