ID: Q100366
The information in this article applies to:
This article provides supplementary information to Section 18.2.2.1 "Comparing and Sorting Strings" in the Windows version 3.1 Software Development Kit "Programmer's Reference, Volume 1: Overview" manual and the Windows Help International Overview section. Specifically, this article provides information about the sort order used by the Windows lstrcmp and lstrcmpi functions, the location of accented characters in the sort, and how primary and secondary values, or weights, are important when sorting a string using these functions. This includes the different behavior of these functions when a language driver is installed compared to when Windows's internal sort routine (English/American) is used.
List boxes that include the LBS_SORT style use lstrcmpi internally to perform the sort. Consequently, the information in this article applies to these list boxes also.
ALPHSORT is a sample that shows the results of these sort routines, and contains a dialog box with two list boxes. One list box contains the characters 32-255 and the option to display the characters in ANSI or sorted order; the other list box includes the LBS_SORT style and contains various strings. Changing the sort routine in use through the International Control Panel application illustrates the effects of the different language drivers and Windows's internal sort routine.
The following file is available for download from the Microsoft Software Library:
~ ALPHSORT.EXE (size: 27846 bytes)
ARTICLE-ID: Q119591
TITLE : How to Obtain Microsoft Support Files from
Online Services
1. Nonalphanumeric characters (in ASCII/ANSI order) 2. Numeric characters 3. Alphabetic characters
For performance reasons, the internal sort routine treats accented characters as nonalphanumeric. Therefore, when the internal routine is used, accented characters appear towards the beginning of the sort between punctuation and numbers. In contrast, when a language driver is used, accented characters appear near their unaccented equivalents because the language drivers sort accented characters as alphabetic characters.
The following illustrate the differences in character order ("..." signifies omitted characters):
ANSI Order:
!"#...0...9:;<...ABC...XYZ[\]...abc...xyz{|}...accented characters
!"#...:;<...[\]...{|}... accented characters 0...9AaBbCc...XxYyZz
...!"#...:;<...[\]...{|}...0...9A accented characters aBbCc...XxYyZz...
Primary and secondary weights of characters also affect the sort order when a language module is installed. In this case, sorting is done by primary weight for the entire length of the string, then by length, and lastly by secondary weight if the primary weights of all the characters and the lengths of the strings are equal. The secondary (diacritic) weights are important only when there is a tie in the entire string. The internal (English/American) sort routine does not sort extended characters; as mentioned above, they are sorted as punctuation rather than alphabetic characters. Therefore, the internal routines produce completely different results than the language routines in some cases.
Character weights are important with case-sensitive sorting also. For example, using lstrcmp will produce: A < a < B < b; it will also produce: Aaa < aaa < Aab. These examples use proper dictionary sort order, but the second example is not necessarily obvious because if A < a, then it seems Aab < aaa should also be true. In that case, it is said that A and a "collide" (that is, their primary weights are the same) and a delayed comparison must be performed if the remainders of the strings are equal. The strings continue to be compared character by character. Because a < b, then aaa < Aab and the comparison is complete. If the strings were equal all the way through (such as Aaa and aaa) then A and a would collide once again. The rest of the strings would be equal, and then the secondary weights of A and a would be checked to determine that Aaa < aaa.
Note: Special notation is used below to represent accented characters due to limitations in the distribution media for this article. For example <u umlaut> is used to represent the letter "u", which has an umlaut over it. Likewise, <a tilde> represents the letter "a" accented with a tilde.
The sort works on a character-by-character comparison, checking primary weights in a string. As soon as the primary (alphabetic) weights show one string greater than the other, the comparison stops. Therefore, sorted lists resemble the following:
a
aa
abba
ac
add
b
ba
baa
The secondary (diacritic) weights are important only when there is a tie in the string. For example, in the following sorted list, the characters in the first two strings have identical primary sort weights ("s", "a", "m", "e") and the strings are the same length. Because of the tie in primary weights and string lengths, the secondary weights are then compared. The first difference in secondary weight ("a" versus <a tilde>) breaks the tie. Secondary weights are also a factor when comparing strings 4 and 5 below:
1. same 2. s<a tilde>me 3. sandy 4. schon 5. sch<o umlaut>n 6. school
It is important to apply the primary weights to the whole string first, and only use the secondary weights in a tie. The primary weights must also carry more importance than the secondary weights as well. Otherwise, an incorrect sort would result (such as schon less than school less than sch<o umlaut>n). This type of weighting creates a sort that makes a distinction between "unique," hard-coded letters of a language and mere variants of a letter (which are only distinguished by diacritics).
Here is a sorted list using the English (International) driver:
a
<a grave>
apple
<a grave>pple
apples
<a grave>pples
lu
l<u umlaut>
l<u umlaut>b
lum
<a grave>
<a grave>pple
<a grave>pples
a
apple
apples
l<u umlaut>
l<u umlaut>b
lu
lum
For more information on the international sort, see the following:
Canadian Standards Association Z243.4.1-1990; "Canadian Alphabetic
Ordering Standard for Character Sets of CSA Z243.4 Standard," Alain
LaBont<e acute>:
DIN Standard 5007, "Orden von Schriftzeichenfolgen: ABC-Regeln,"
April 1991
IBM Document GG24-3516, "Keys to Sort and Search for Culturally
Expected Results," Denis Garneau
"R<e acute>gles du classement alphab<e acute>tique en langue
fran<c cedilla>aise...", Alain LaBont<e acute>
Keywords : kbsample kb16bitonly kbWinOS310 kbIntlDev kbWinOS300
Version : 3.00 3.10
Platform : WINDOWSLast Reviewed: December 12, 1998