RE: NVARCHAR

Posted in 1998

Lars Johansson — Fri, 19 Jun 1998 13:35:34 +0200 — source: Informix-list mailing list archive (1991-1998)

> ---------- > From: David Williams[SMTP:djw@smooth1.demon.co.uk] > Sent: 19. juni 1998 08:51 > To: informix-list@iiug.org > Subject: Re: NVARCHAR > > In article <3588AAEE.748D@dati.lv>, Leonid Vorontsov > <Leonids.Voroncovs@dati.lv> writes > >Hi All. > >Sometime we worked with version 7.22 and had performance problem with > >NVARCHAR fields. Answer from tech support was "...There is a theory that > >it could be a bug with VARCHAR which caused the server to always go to > >the SEQUENTIAL SCAN mode. If this is the case, the bug is fixed in > >7.3...". But real situation isn't good. I investigated Informix 7.3 > >behaviour and result is there: > > > One of the Informix SQL Manuals talks about optimizing your data model > and mentions than NVARCHAR is slower than VARCHAR is slower than CHAR. NCHAR, NVARCHAR is much slower than CHAR, VARCHAR, due to the fact that NCHAR columns sort i based on "rule-sort" where CHAR is sorted in codepage order. What it means is: Don't use NCHAR unless correct sorting is critical. GLS offers many good things, but perfomance isn't one of them. Lars Johansson