NSL vs Bytes Order.

[Enrico]

Hi.
I am trying to get away from the NSL nuisance nightmare (almost every single PC has some little difference, and needs to Pack the tables when copying the database from one PC to another).
2 Questions.
A) For accented languages (French, Spanish, etc) the sort order of the strings indexed would work just fine after this conversion? That is, will it keep the same sort order (for French and Spanish languages)?
B) Is is possible to change the Locale in code? If so, how, any hint/example? I know how to change options, but I have no idea of how to change Locale.
Thanks

> I am trying to get away from the NSL nuisance nightmare (almost every
> single PC has some little difference, and needs to Pack the tables
> when copying the database from one PC to another).

You won't be able to; even incremental upgrades to Windows might change
the NLS version and require running the table update.

> A) For accented languages (French, Spanish, etc) the sort order of the
> strings indexed would work just fine after this conversion? That is,
> will it keep the same sort order (for French and Spanish languages)?

I think you will find that your French etc users won't be very happy
with byte order.

> B) Is is possible to change the Locale in code? If so, how, any
> hint/example? I know how to change options, but I have no idea of how
> to change Locale.

Use the Source Code generator in EM (right-click database node).

--
Eivind Bakkestuen [NDD]

[Enrico]

Thanks Eivid.
Ok, we DO need NSL.
However, I had done a test with the code below (before your anser), and strangely it also changed the fied values (not only the "locale").
For example, lines with 2 ALB?NDIGAS were converted to 2 ALBУNDIGAS.

procedure fixlocale(nxDBase: Tnxdatabase;tablename:string);
 var
   idxField,i,idx,idx1: Integer;
   NewDict, Dict: TnxDataDictionary;
   TaskInfo: TnxAbstractTaskInfo;
   Completed: Boolean;
   TableMapper: TnxTableMapperDescriptor;
   TaskStatus: TnxTaskStatus;
   NeedRestruct: Boolean;
   aindexdescriptior:TnxIndexDescriptor;
   aIndexFieldDescriptor :  TnxKeyFieldDescriptor;
 begin
 try
   tablename:=StringReplace(tablename,'.nx1','',[rfIgnoreCase,rfReplaceAll]);
   Dict := TnxDataDictionary.Create;
   nxdbase.GetDataDictionary(Tablename, '', Dict);


   TableMapper := TnxTableMapperDescriptor.Create;;
   try
     NewDict := TnxDataDictionary.Create;
     NewDict.Assign(Dict);
     with newdict do begin

       with newdict.AddLocaleDescriptor do begin
        Locale := 2;
        UseStringSort:=true;
       end;

       for idx:=0 to NewDict.IndicesDescriptor.IndexCount-1 do  begin
         if NewDict.IndicesDescriptor.IndexDescriptor[idx].KeyDescriptor
           is TnxCompKeyDescriptor then begin
             for idx1:=0 to
               (NewDict.IndicesDescriptor.IndexDescriptor[idx].KeyDescriptor as
                TnxCompKeyDescriptor).KeyFieldCount-1 do begin
                   (NewDict.IndicesDescriptor.IndexDescriptor[idx].KeyDescriptor as
                      TnxCompKeyDescriptor).KeyFields[idx1].RemoveLocaleDescriptor;

                end;
         end;
       end;


     end;
     TableMapper.MapAllTablesAndFieldsByName(Dict, NewDict);
     // and finally restructure and wait until the process is done
     nxCheck(nxDBase.RestructureTableEx(Tablename, '', newDict, TableMapper, TaskInfo));
     try
       if Assigned(TaskInfo) then begin
         repeat
           TaskInfo.GetStatus(Completed, TaskStatus);
           if not Completed then begin
             Sleep(500)
           end;
         until Completed;
         nxCheck(TaskStatus.tsErrorCode);
       end;
     finally
       if Assigned(TaskInfo) then
         TaskInfo.Free;
     end;
   finally
     Dict.Free;
     NewDict.free;
     TableMapper.Free;
   end;
 except
   on e: Sysutils.Exception do
      showmessage(e.message);

 end;
 end;

> and strangely it also changed the fied values (not only the "locale").
> For example, lines with 2 ALB?NDIGAS were converted to 2
> ALBУNDIGAS.

I'm pretty sure that changing the locale does not actually change the
field value; the bytes stored in the string field would be exactly the
same as before. But changing the locale obviously influences how those
bytes are interpreted when displayed.

--
Eivind Bakkestuen [NDD]

Eivind Bakkestuen wrote:

> > and strangely it also changed the fied values (not only the "locale").
> > For example, lines with 2 ALB?NDIGAS were converted to 2
> > ALBУNDIGAS.
>
> I'm pretty sure that changing the locale does not actually change the
> field value; the bytes stored in the string field would be exactly the
> same as before. But changing the locale obviously influences how those
> bytes are interpreted when displayed.

Changing the locale (and with it, the codepage, in case of non-wide strings)
will perform a codepage translation from the previous codepage to the new
codepage. In case any characters can not be mapped from the source to target
codepage, it depends if the restructure is using dlaBestFit or not. If it's
used, the unmappable characters should be mapped to ?, otherwise the
restructure should fail.

Changing from/to byte order (no locale specified) is a special case, in which
the bytes are simply accepted or left as is, and it's your responsibility to
make sure the bytes that make up the strings are in the locale/codepage you
have specified.