TStringList wrong behavior
lazarus / 8-Other

----------------------------------------------------------------------
typo · 2010-05-16 18:46

Don't you think that putting accented words before all the other words
is a wrong behavior? I think it is simply wrong... I don't know any
language that do this kind of sorting.

----------------------------------------------------------------------
Martin_fr · 2010-05-16 21:34

I've seen that in databases, and afaik even windows has a setting to
do so.

There is no "correct" sorting order by default. As you mention it's
language specific => it's called collation.

TStringlist probably uses binary collation, in which letters are
ordered by their numeric value; and not by any language specific rule.

Using something like TListBox, may or may not leave the sorting to the
Operating System, the Operating System can be set to a different
collation.

From http://en.wikipedia.org/wiki/Collation
"A numeric sort may order ñ incorrectly following z "
Actually it isn't incorrect. For a numeric (binary collation) this is
the correct place. (it is only incorrect, if your intend was to sort
by another collation, but then the error is that you used the wrong
collation.

And for some languages there is more than one collation available.
http://de.wikipedia.org/wiki/Alphabetische_Sortierung gives 4
different ways how things can be sorted in German.

TStringlist  has no setting to learn which collation you want to use
(the language of the text doesn't matter, the collation does).
So TStringlist sorts things correct, according to the collation it
uses.

----------------------------------------------------------------------
typo · 2010-05-16 22:04

QuoteUsing something like TListBox, may or may not leave the sorting
to the Operating System, the Operating System can be set to a
different collation.
But, if TListBox does the thing, why TStringList can't do?

----------------------------------------------------------------------
Martin_fr · 2010-05-16 22:17

As I said:

TListbox is probably sorted by your Operating System.

TStringList is OS independent, it doesn't call the OS.

For many tasks the binary sort order is more than enough. eg keeping
data sorted, so you can do binary search on it.

---
Now that doesn't mean that it wouldn't be a nice feature. It only
says, that the current behaviour isn't (imho) wrong as such.

having said that, you can always use
TstringList.CustomSort

I don't know if there is any ready made collation sorting available,
or if you have to write it from scratch....

----------------------------------------------------------------------
theo · 2010-05-17 03:25

Didn't you ask this before?
http://forum.lazarus.freepascal.org/index.php/topic,8757.0.html

----------------------------------------------------------------------
typo · 2010-05-17 04:09

Before I asked if this was intentional. Now I'm asking if this is
right.

----------------------------------------------------------------------
typo · 2010-05-17 05:06

I think you'll answer it is the only one, so it is right

----------------------------------------------------------------------
Martin_fr · 2010-05-17 05:27

There even exists a bug:
http://bugs.freepascal.org/view.php?id=15864

but the forum is a better place to discuss it.

----------------------------------------------------------------------
typo · 2010-05-17 05:34

AnsiCompare serves only to non-accented languages.

----------------------------------------------------------------------
typo · 2010-05-17 06:04

Perhaps something in Assembly.

----------------------------------------------------------------------
theo · 2010-05-17 07:23

Can you please show us which result you get and what you expect?

Please try this code:

Code: [Select]function UTF8Compare(List: TStringList; Index1, Index2:
Integer): Integer;
begin
  Result:=WideCompareStr(UTF8Decode(List[Index1]),UTF8Decode(List[Inde
x2]));
end;

procedure TForm1.Button1Click(Sender: TObject);
var sl:TStringList;
begin
 sl:=TStringList.Create;
 sl.Text:=Memo1.text;
 sl.CustomSort(@UTF8Compare);
 Memo2.Text:=sl.text;
 sl.free;
end;

Does the result difffer from this chart (Windows Vista for PT_BR) ?
http://www.collation-
charts.org/vista/vista.0416.CP1252.Portuguese_Brazil.html

----------------------------------------------------------------------
typo · 2010-05-17 07:50

Well, thanks, it is very fast and I will use it, but why it is not in
the TStringList.Sort method?

Quotefunction TStringList.DefaultCompareText(const s1, s2: string):
PtrInt;
begin
  if FCaseSensitive then
    Result := AnsiCompareStr(s1, s2)
  else
    Result := AnsiCompareText(s1, s2);
end;

----------------------------------------------------------------------
typo · 2010-05-17 07:57

QuoteProcedure TStringList.Sort;

begin
  CustomSort(@StringListAnsiCompare);
end;

----------------------------------------------------------------------
theo · 2010-05-17 08:30

Because Unicode support in the RTL is not ready.
See: http://wiki.freepascal.org/FPC_Unicode_support#Roadmap_of_RTL_Uni
code_support_with_UTF8String

----------------------------------------------------------------------
typo · 2010-05-17 08:32

OK.