U++ forum: Welcome to the forum

Search on this site

Search in forums

Home » U++ Library support » U++ Core » Strings with national specific characters are wrongly sorted - Sort

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Strings with national specific characters are wrongly sorted - Sort [message #57468]

Wed, 25 August 2021 13:01

Klugier
Messages: 1099
Registered: September 2012
Location: Poland, Kraków

Senior Contributor

Hello,

Today I found that sort returns wrong results in term of special characters:

#include <Core/Core.h>

using namespace Upp;

CONSOLE_APP_MAIN
{
	Vector<WString> vec = { "Zbig", "Ąć", "Ęc", "Ala", "Edward" };
	Sort(vec);
	
	for (const auto s : vec)
	{
		Cout() << s << "\n";
	}
}

The results are:

Ala
Edward
Zbig
Ąć
Ęc

and should be:

Ala
Ąć
Edward
Ęc
Zbig

This is probably corner case, because this world doesn't exist in Polish, but anyway the error is there. I believe it is more serve when these character are in the middle of the string and we have a lot of such words.

Here is the article about Polish alphabet and the order of letters.

Klugier

U++ - one framework to rule them all.

[Updated on: Wed, 25 August 2021 17:49]

Report message to a moderator

Re: Strings with national specific characters are wrongly sorted - Sort [message #57470 is a reply to message #57468]

Thu, 26 August 2021 13:09

busiek
Messages: 70
Registered: February 2011
Location: Poland

Member

For Polish alphabet I use such function to map wchar into something comparable according to our needs:

auto CharRepr = [](wchar c)
{
    wchar d = ToLower(c);
    d = d == u'ł' ? 'l' : d;
    return MakeTuple(IsLetter(c), ToAscii(d), c);
};

Then, to compare, I map each WString into the appropriate vector of tuples.

Cheers,
Kuba

Report message to a moderator

Re: Strings with national specific characters are wrongly sorted - Sort [message #57473 is a reply to message #57468]

Fri, 27 August 2021 09:51

mirek
Messages: 14255
Registered: November 2005

Ultimate Member

Klugier wrote on Wed, 25 August 2021 13:01

Hello,

Today I found that sort returns wrong results in term of special characters:
#include <Core/Core.h>

using namespace Upp;

CONSOLE_APP_MAIN
{
	Vector<WString> vec = { "Zbig", "Ąć", "Ęc", "Ala", "Edward" };
	Sort(vec);
	
	for (const auto s : vec)
	{
		Cout() << s << "\n";
	}
}
The results are:
Ala
Edward
Zbig
Ąć
Ęc
and should be:
Ala
Ąć
Edward
Ęc
Zbig
This is probably corner case, because this world doesn't exist in Polish, but anyway the error is there. I believe it is more serve when these character are in the middle of the string and we have a lot of such words.

Here is the article about Polish alphabet and the order of letters.

Klugier

This is not error, base [W]String comparison simply compares character values.

You need to use NLS specific sorting in this situation - LanguageInfo::Compare. That said, it really is specifically defined just for CZ and even there it would need improvement, OTOH the generic routine should at least work better that the result you get.

BTW, language specific sorting is extremely difficult topic if it should be done right in many languages...

Mirek

[Updated on: Fri, 27 August 2021 10:18]

Report message to a moderator

Re: Strings with national specific characters are wrongly sorted - Sort [message #57474 is a reply to message #57473]

Fri, 27 August 2021 09:57

mirek
Messages: 14255
Registered: November 2005

Ultimate Member

BTW, as I am revisiting this issue after a long long time, it looks like we should connect here to host platform APIs in the future.

Report message to a moderator

Previous Topic:	Probable nasty bug with StringBuffer
Next Topic:	POSIX home directory symbol '~' is causing trouble

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Fri Apr 25 12:16:51 CEST 2025

Total time taken to generate the page: 0.01246 seconds