SELECT StudentNo
FROM TStdCourses
WHERE (((DptID = 11) And (SubjectID = '27') And
(Grade in ('Ã','-Ã','+È','È','-È','+ÌÜ','ÌÜ','-ÌÜ','+Ï','Ï') ))) AND
StudentNo IN(
SELECT StudentNo
FROM TStdCourses
WHERE (((DptID = 11) And (SubjectID = '47') And
(Grade in ( 'åÜ','æ','') )))
EXCEPT(
SELECT StudentNo
FROM TStdCourses
WHERE (((DptID = 11) And (SubjectID = '47') And
(Grade = '' )) AND ((StudyYear = 2004) And
(StudySemester = 2)) ) OR (((DptID = 11) And
(SubjectID = '47') And
(Grade in ('Ã','-Ã','+È','È','-È','+ÌÜ','ÌÜ','-ÌÜ','+Ï','Ï') )))
)
)

---

 Comments

---

The problem was caused by two issues:

1) DBISAM was putting too many key fields in the temporary index that it was creating for optimizing the sub-query, and this was initially causing the problem.

2) However, this wasn't an issue for non-Arabic locales, so after further investigation it was found that the sorting of the ANSI characters for the Arabic locales was not working properly and sorting non-printable characters higher than printable characters. This messed up the building of the internal character map that is used by DBISAM to build buffer ranges for optimized filters and queries that can use an available index.

The solution was to exclude non-printable characters (#0-#31) from the sorting portion of building the internal character map.

---

 Resolution

---

Fixed Problem on 2/26/2005 in version 4.18 build 1