SetString / AppendChunk behaviour
lazarus / 16-Databases

----------------------------------------------------------------------
RayoGlauco · 2022-08-26 08:47

Hello,

Some time ago, I had written some code to write data to MS Access blob
fields. It was working fine, but recently I have recompiled the
program (Lazarus 2.2.2 on Windows 10) and now something is wrong.

I guess something in the internal handling of the string has changed,
and now the bytes are not transferred correctly (it's very likely that
the string contains the character #0 multiple times).

Does someone have a clue how to fix it?

Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i var 
fieldname: WideString;  buffer: packed array [0..MaxBuffer - 1] of
byte;  s: rawbytestring;  nbytes:integer;  st: TMemoryStream  ADOconn,
ADOrset: Variant;begin  ...  ADOconn :=
CreateOleObject('ADODB.Connection');  ...  ADOrset :=
CreateOleObject('ADODB.Recordset');  ...  st.ReadBuffer(buffer,
nbytes);  SetString(s, @buffer[0], nbytes);   
ADOrset.Fields[fieldname].AppendChunk(s);     ...end;

----------------------------------------------------------------------
Thaddy · 2022-08-26 10:58

The buffer is probably way to smal  buffer: packed array [0..MaxBuffer
- 1] of byte;
l, it should be   buffer: packed array [0..MaxBuffer*4 - 1] of byte;
because unicodestring or widestring can take upto 4 bytes per
unicodechar.
You can of course do that multiplication by 4 already in the
declaration of MaxBuffer! But I suspect you haven't done that.
So:
Check if the buffer is large enough
maybe change widestring to UTF8String or UnicodeString, which work
both like Pascal string types. Widestring doesn't.

----------------------------------------------------------------------
RayoGlauco · 2022-08-26 15:45

Thank you Taddy for your reply, but the code worked in previous
versions of Lazarus.

I have verified that the string passed to AppendChunck is always
converted to Widestring, in the 2.2.2 version of Lazarus.

And that's the problem: the content of the string is altered by
Lazarus just before being passed to AppendChunck.

I want to use the latest version of Lazarus, so I'll find another way
to get the job done.

----------------------------------------------------------------------
RayoGlauco · 2022-08-26 16:39

I have found a solution that works: put the data directly into a
WideString.

The only thing is that the number of bytes that can be written will
always be a multiple of 2.

Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i var  ... 
cptr: ^char;  s: WideString;begin  ...  SetLength(s, nbytes div 2 +
nbytes mod 2);  cptr:=@s[1];  for i:=0 to nbytes-1 do    cptr[i]:=
char(buffer[i]);  ADOrset.Fields[fieldname].AppendChunk(s);  ...end;

----------------------------------------------------------------------
KodeZwerg · 2022-08-26 16:49

My advise would be.... when you using "WideString" you also should use
"WideChar" and multiply with SizeOf(WideChar) to have correct length.

----------------------------------------------------------------------
RayoGlauco · 2022-08-26 17:13

Thanks KodeZwerg, but my problem is that I want to write a byte stream
to a MSaccess blob field.

It needs to be done by using an AnsiString, but Lazarus kindly
converts my AnsiString to a WideString, altering the original content.

That's why I copy the bytes directly into a WideString.

This is my last solution:

Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i var  ... 
s: WideString;begin  ...  // nbytes must be even  if nbytes mod 2 = 1
then  begin    buffer[nbytes]:= 0;  // set extra byte to 0   
inc(nbytes);         // make nbytes even  end;  SetLength(s, nbytes
div 2);  move(buffer[0], s[1], nbytes); 
ADOrset.Fields[fieldname].AppendChunk(s);  ...end;