Date: Mon, 23 Mar 1998 15:39:11 -0500
From: darrenk@insightdist.com (Darren King)
Message-Id: <9803232039.AA48906@ceodev>
To: pgsql-hackers@postgresql.org
Subject: Re: [HACKERS] char types gone.
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Md5: 1RjADdt3LjPz3njzSe5mkA==

> > Have managed to delete char2,4,8 and 16.
> > Three points now...regression tests, docs and interfaces.
> > 
> > 1. Regression tests make very extensive use of char16.  Seems to
> > me that these should be changed to a name.  Same for tutorial and
> > any other example references in the code.  Seem ok?
> 
> Any reason not to use "text" for these cases? Or even varchar()? "name"
> is probably really only an internal Postgres type. We should have a
> regression test for it if we don't already. Let me know if you want me
> to help...

I think the 'name' type should be considered an internal type only and
mentioned as such wherever it is documented.  It's only purpose seems
to be for a consistent length to internal names and it does that rather
well.  User's would be better served with the other, varlena-based, types.

Any examples I was going to convert from char16 to char(16).  There are a
couple of mentions of the char16 struct that will have to be integrated
or changed to reference the varlena struct.

Finally, I'm redoing the char16 regression test for name.

> > 2. I'll take my best shot at the docs.
> 
> Where? The man pages or the sgml sources? I'm shooting for getting the
> sgml sources right, and obsoleting the man page _sources_, regenerating
> them from the sgml instead.

Do you want patches against doc/src/sgml/*?  The man pages references were
few, so they were easy to change for consistency sake then til they drop
into the bit bucket.  Changed a few references in the userguide too.

Only have to tidy up the docs and I'll submit the patch.

darrenk