BufDataset and LoadFromStream problem
lazarus / 31-LCL

----------------------------------------------------------------------
iru · 2010-05-25 21:04

Gentlefolk,

Lazarus 0.9.29 snapshot (using to address other issues with dbgrid),
windows xp, SP-3.

I have a peice of delphi code which I have migrated to lazarus. The
Lazarus code is to load the 'fields' in a .CSV file to a BufDataset,
the code crashes on 'LoadfromStream' with the following error message:

"the data stream format is not recognised".

The executed code is

   F := TFileStream.Create(FileName, fmOpenRead);
   S := TMemoryStream.Create;

   try
      S.SetSize(F.Size + 2);
      P := S.Memory;
      Move(CrLf[1], P^, 2);
      Inc(P, 2);
      F.ReadBuffer(P^, F.Size);
      MT.LoadFromStream(S, dfAny); 

The file contains one line of text '123,"test description","F","T"'.
After the ReadBuffer P contains  '123,"test description","F","T"r'.

Looking at the code in Bufdataset.  and
Bufdataset.TFpcBinaryDatapacketReader.RecognizeStream shows the code
doing some strange things where it compares

#13#10'123,"test d' against FpcBinaryIdent = 'BinBufDataset';

this fails and it is all down hill from there.......

Any ideas, solutions, explanation, examples of using a LoadFromStream
with BufDataset appreciated.

thanks, Ian.

----------------------------------------------------------------------
Loesje · 2010-05-27 09:13

LoadFromStream uses by default a binary format. You can add more
formats by adding more formatter-units.

There is currently no formatter for csv files, but you can use the
xml-format that Delphi's TClientDataset also uses by default. Add the
'xmldatapacketreader' to your program and it will recognize Delphi's
xml format.

----------------------------------------------------------------------
stonefull · 2010-05-27 12:32

I'd use TSdfDataSet.

----------------------------------------------------------------------
iru · 2010-05-27 19:09

Gentlefolk,

Thanks for the responses.

May I ask why use 'TSdfDataSet' ?????

Ian

----------------------------------------------------------------------
stonefull · 2010-05-28 04:50

TSdfDataset can define commas as separator, like csv format. It's very
simple.

----------------------------------------------------------------------
iru · 2010-05-29 06:25

Thanks for the response.

All goes well up to a point. I can load the contents of a file (can
see it in an associated DBGrid).

However when I try  SdfDataSet.FieldByName('Name').AsString a 'field
not known' exception occurs. The field is there, it is defined at
design time.

Looking at the code in dtaset.inc it would appear that no fields are
defined.

I will attempt to dynamically define the fields and see what happens.

Ian

----------------------------------------------------------------------
stonefull · 2010-05-29 07:05

TSdfDataset define field names as Field1, Field2, Field3, etc... You
can't modify them. Try

SdfDataSet.FieldByName('Field1').AsString

I wrote a post (in Portuguese) about this:

http://professorcarlos.blogspot.com/2010/03/lazarus-tratamento-de-
arquivos-texto.html

----------------------------------------------------------------------
tatamata · 2010-05-29 12:10

See http://wiki.lazarus.freepascal.org/How_to_write_in-memory_database
_applications_in_Lazarus/FPC#Saving_MemDatasets_To_Persistent_Files

----------------------------------------------------------------------
iru · 2010-05-29 18:00

Gentlefolk,

Thank you, abbrigado (my one word of Portugese learned on a breif
vizit to Mozambique (spelling ???)).

Things are now working OK.

Ian

----------------------------------------------------------------------
typo · 2010-05-29 18:04

It's "obrigado".