Hi All,
I was just wondering what are the implications of installing NLS on UniVerse (9.6.iforget on AIX 4.3.3).
Has anyone had any horrific problems?
Do we need to regression test existing code?
Will there be a performance hit for existing code?
Are there any problems interfacing existing code with new NLS code?
TIA
Craig Bennett
NLS Performance
>Has anyone had any horrific problems?
Works well as the engine for DataStage throughout Asia. Havent heard of any horrific problems, except where there were a couple of characters incorrectly mapped (in one case because Micro$oft didnt follow a standard they claimed to follow).
>Do we need to regression test existing code?
Need to? No. Should? Yes. Always.
>Will there be a performance hit for existing code?
Yes in some areas. Mainly in string handling. Inside the engine they cant move bulk bytes any more; have to check a byte at a time to determine whether the byte is the lead byte of a character or the trailing byte of a multi-byte character.
>Are there any problems interfacing existing code with new NLS code?
No. This was one of the original design goals, and they did it well.
Works well as the engine for DataStage throughout Asia. Havent heard of any horrific problems, except where there were a couple of characters incorrectly mapped (in one case because Micro$oft didnt follow a standard they claimed to follow).
>Do we need to regression test existing code?
Need to? No. Should? Yes. Always.
>Will there be a performance hit for existing code?
Yes in some areas. Mainly in string handling. Inside the engine they cant move bulk bytes any more; have to check a byte at a time to determine whether the byte is the lead byte of a character or the trailing byte of a multi-byte character.
>Are there any problems interfacing existing code with new NLS code?
No. This was one of the original design goals, and they did it well.
As always Ray your knowledge is amazing.
Craig
-----Original Message-----
From: Ray Wurlod [mailto:ray.wurlod@Informix.Com]
Sent: Friday, 19 October 2001 11:29
To: u2-users@oliver.com
Subject: UV: NLS Performance
>Has anyone had any horrific problems?
Works well as the engine for DataStage throughout Asia. Havent heard of any horrific problems, except where there were a couple of characters incorrectly mapped (in one case because Micro$oft didnt follow a standard they claimed to follow).
>Do we need to regression test existing code?
Need to? No. Should? Yes. Always.
>Will there be a performance hit for existing code?
Yes in some areas. Mainly in string handling. Inside the engine they cant move bulk bytes any more; have to check a byte at a time to determine whether the byte is the lead byte of a character or the trailing byte of a multi-byte character.
>Are there any problems interfacing existing code with new NLS code?
No. This was one of the original design goals, and they did it well.
Craig
-----Original Message-----
From: Ray Wurlod [mailto:ray.wurlod@Informix.Com]
Sent: Friday, 19 October 2001 11:29
To: u2-users@oliver.com
Subject: UV: NLS Performance
>Has anyone had any horrific problems?
Works well as the engine for DataStage throughout Asia. Havent heard of any horrific problems, except where there were a couple of characters incorrectly mapped (in one case because Micro$oft didnt follow a standard they claimed to follow).
>Do we need to regression test existing code?
Need to? No. Should? Yes. Always.
>Will there be a performance hit for existing code?
Yes in some areas. Mainly in string handling. Inside the engine they cant move bulk bytes any more; have to check a byte at a time to determine whether the byte is the lead byte of a character or the trailing byte of a multi-byte character.
>Are there any problems interfacing existing code with new NLS code?
No. This was one of the original design goals, and they did it well.
Actually, I have found one glitch.
If you use READBLK, SEEK or WRITEBLK, and expect to read/move/write a number of bytes, beware that they actually process a number of characters. This is not as documented, but did break a couple of programs I use in the UniVerse Internals class. There is a workaround, but its messy.
If you use READBLK, SEEK or WRITEBLK, and expect to read/move/write a number of bytes, beware that they actually process a number of characters. This is not as documented, but did break a couple of programs I use in the UniVerse Internals class. There is a workaround, but its messy.